Because the work was retrospective, conducted at one center and used clinical determination as the reference standard, its estimates should not be interpreted as proof that Doppler can independently establish the diagnosis in every setting. The researchers reviewed 512 children admitted to the pediatric intensive care unit between 2016 and 2025. All had a Glasgow Coma Scale score of 3T, indicating no eye opening, no motor response and an unassessable verbal score because of endotracheal intubation, and all underwent Doppler on the same day.

Their median age was 81 months, with an interquartile range of 33 to 130 months and a range from one month to 17.6 years; 304 children, or 59.4%, were male. Major causes of severe brain injury included infection-related encephalopathy, which accounted for 21.9% of the cohort, hypoxic-ischemic brain injury, at 20.9%, and traumatic brain injury, at 13.7%. The children were divided into 47 who met Chinese criteria for brain death or death by neurologic criteria, 266 in whom the condition was excluded because preserved brain function meant no evaluation was initiated, and 199 whose evaluation remained incomplete.

Chinese pediatric criteria required a devastating brain injury, exclusion of confounders and two sets of examinations consistent with brain death or death by neurologic criteria. Each set combined clinical examination with confirmatory testing. The clinical assessment evaluated coma, brainstem reflexes and apnea, while the permitted ancillary tests included transcranial Doppler, electroencephalography and short-latency somatosensory evoked potentials.

At least two confirmatory tests had to support the diagnosis, or all three when a brainstem reflex could not be tested. The age-dependent interval between the two sets was at least 24 hours for children from 29 days to younger than one year and at least 12 hours for those aged one to 18 years. All 47 children in the confirmed group fulfilled the clinical criteria without confounders, and the final clinical determination was used as the reference standard for evaluating Doppler performance.

Doppler findings were considered consistent with the diagnosis when the anterior and posterior circulations on both sides showed reverberating flow, systolic spikes or loss of previously detectable intracranial flow signals on serial examinations. Reverberating flow consisted of forward systolic flow accompanied by reverse diastolic flow within one cardiac cycle. Systolic spikes were isolated systolic signals lasting less than 200 milliseconds and with peak systolic velocity below 50 centimeters per second.

The middle cerebral artery or terminal internal carotid artery was used to assess the anterior circulation, while the basilar or intracranial vertebral artery was used for the posterior circulation. Examinations used temporal, suboccipital and submandibular windows, with the carotid siphon assessed through the orbital window when the temporal window was inadequate. Two certified technicians performed the examinations, and a third certified investigator independently reinterpreted the stored waveforms while blinded to all other clinical and ancillary-test information; the independent interpretations were fully concordant with the original assessments of whether findings were consistent with brain death.

Among the 47 children who met the clinical criteria, 44 underwent Doppler before the first examination met those criteria. Thirty-four of them, or 77.3%, already had Doppler findings consistent with brain death a median of 28.4 hours before the clinical determination point. One additional child developed consistent waveforms on a repeat examination five hours later.

Three children were first examined two to 16 hours after the clinical criteria were met, and all had consistent Doppler findings. Four children retained inconsistent waveforms on follow-up scans, while five received no further Doppler examination. The researchers also noted that prior reports have found the rate of Doppler detection of cerebral circulatory arrest rising from 58% within six hours of clinical diagnosis to 100% after 24 to 36 hours.

In the present cohort, most Doppler examinations were performed early in the diagnostic sequence because Chinese protocol places confirmatory testing before apnea testing. The nine children with clinically confirmed brain death but Doppler findings inconsistent with the diagnosis had detectable intracranial arterial flow. The most common potential explanation was disruption of the normally closed intracranial system: four children had ventricular drainage, a patent anterior fontanelle or multiple skull fractures.

Such openings can reduce the rise in intracranial pressure, potentially delaying the characteristic Doppler pattern of cerebral circulatory arrest. One child had a suspected cerebrovascular malformation, which may have provided a low-resistance route for blood to bypass areas of critically elevated pressure. One child received venoarterial extracorporeal membrane oxygenation, whose driving pressure could maintain pulsatile or continuous flow despite absent brain function.

Another examination was delayed by 28 days after severe impairment of consciousness began, when resolution of cerebral edema may have reduced intracranial pressure. No definite explanation was found for the remaining two children. The researchers described these as potential contributing factors rather than experimentally established mechanisms.

Several observations illustrate why waveform character matters. In children with disrupted intracranial systems, Doppler sometimes detected continuous forward diastolic flow rather than the expected reverberating flow or systolic spikes. The child with suspected arteriovenous malformation had asymmetric perfusion, with residual forward flow on the affected side and markedly reduced or absent flow on the opposite side.

The child receiving venoarterial extracorporeal membrane oxygenation initially showed nonpulsatile flow and later pulsatile flow, but forward diastolic flow remained present; computed tomography angiography and perfusion imaging performed on day 13 of severe impairment of consciousness indicated cerebral circulatory arrest. Because no contemporaneous Doppler examination was available for that later imaging, the investigators could not determine whether the flow pattern had subsequently evolved. Another child’s scan was performed 28 days after severe impairment of consciousness and showed only reduced diastolic blood velocity, a pattern that may represent a transitional stage or residual perfusion below the level required to sustain brain function.

The study did not include consistent serial imaging or concurrent cerebral blood-flow measurements for every case. The incomplete-determination group requires a separate interpretation. Doppler findings consistent with brain death or death by neurologic criteria were observed in 117 of 199 children, or 58.8%, whose determination was not completed.

This included 83 of 157 whose evaluation was not initiated, 16 of 24 whose apnea testing was not completed and all 18 whose repeat examination was not completed. The inconsistent group included 74 children whose evaluation was not initiated, eight whose apnea testing was not completed and none whose repeat examination was not completed. Because these children did not receive a complete clinical determination, the distribution of Doppler findings in this group cannot establish diagnostic performance or a false-positive rate.

Doppler may also have influenced whether clinicians initiated or continued the determination process, introducing verification bias. The data are therefore best read as a description of the results in an unresolved clinical population, not as evidence that Doppler alone identified brain death in 117 children. The study has important limitations.

Doppler timing and frequency were not standardized, and missing repeat examinations may have underestimated how many children would eventually develop waveforms consistent with cerebral circulatory arrest. Clinical instability or family preference prevented some evaluations from being initiated or completed, leaving a large, uncertain population whose final status was unknown. The lack of concurrent validation with angiography or nuclear-medicine flow studies, together with incomplete longitudinal imaging, limited investigation of why residual flow persisted.

The nine discordant children were also a small subgroup, and their characteristics could not establish which factor was decisive in an individual case. The results support Doppler as an adjunct that is noninvasive, repeatable and feasible at the bedside, not as a substitute for a full clinical assessment and other required confirmatory evidence. When Doppler conflicts with strong clinical suspicion or established criteria, clinicians must examine the waveform and consider drainage devices, skull fractures, patent fontanelles, vascular malformations, mechanical circulatory support and examination timing rather than treating residual flow as an automatic reversal of the diagnosis.

Across the full cohort, the apparent strength of a Doppler result depended on what it showed. In the group whose evaluation remained incomplete, 58.8% also had consistent findings, but the absence of a final reference diagnosis prevents those observations from being converted into sensitivity or false-positive estimates. The study therefore adds evidence that Doppler can reinforce a pediatric brain-death or death-by-neurologic-criteria assessment when it shows the expected arrest pattern, while preserving a clinically important warning: residual arterial flow can remain detectable after the clinical criteria have been met.

That warning is most relevant in children with disrupted intracranial systems, suspected cerebrovascular malformations, extracorporeal circulatory support or delayed testing, circumstances in which the Doppler waveform may not tell the same story as the complete clinical evaluation.

Subject of Research: Diagnostic performance and limitations of transcranial Doppler for pediatric brain death or death by neurologic criteria

Article Title: Clinical utility and limitations of transcranial Doppler in evaluating pediatric brain death or death by neurologic criteria: a 10-year retrospective study

Article References: Tan, N., Li, Y., Xiao, N., Wang, Q., Wu, J., & Qian, S. (2026). Clinical utility and limitations of transcranial Doppler in evaluating pediatric brain death or death by neurologic criteria: a 10-year retrospective study. Frontiers in Pediatrics, 14, Article 1931129. https://doi.org/10.3389/fped.2026.1931129

Image Credits: AI Generated

DOI: 10.3389/fped.2026.1931129

Keywords: pediatric brain death, death by neurologic criteria, transcranial Doppler, cerebral circulatory arrest, intracranial arterial flow, pediatric intensive care, ancillary testing, intracranial pressure, extracorporeal membrane oxygenation, Clinical, utility, limitations

The patient initially experienced prolonged constitutional symptoms, including fever, weight loss and malaise. These symptoms were followed by progressive neurological deterioration, with headache, altered mental status and signs of raised intracranial pressure. The report notes that tuberculous meningitis and cryptococcal meningitis have overlapping clinical features, creating diagnostic and therapeutic challenges when both infections are present.

Cerebrospinal fluid analysis supported a dual diagnosis. A positive GeneXpert result identified Mycobacterium tuberculosis, while cryptococcal antigen testing was consistent with Cryptococcus neoformans infection. The patient was described as HIV-negative and had no known immunosuppressive condition. The report characterizes simultaneous central nervous system infections caused by these pathogens as exceedingly rare, particularly in immunocompetent individuals.

Following diagnosis, appropriate antituberculous therapy and antifungal treatment were initiated. Nevertheless, the woman’s clinical condition continued to worsen. The case report does not provide sufficient information in the supplied account to determine whether a particular treatment regimen, the concurrent infections or other factors contributed to the fatal outcome. It therefore should not be read as evidence that either infection was untreatable or that any specific treatment caused the outcome.

During the clinical course, the patient developed communicating hydrocephalus, accompanied by intracranial pressure that remained elevated. Her neurological deterioration ultimately progressed to brain herniation and death. Hydrocephalus and raised intracranial pressure are therefore documented features of this presentation, but the supplied report does not detail the imaging or clinical measurements used to characterize them.

The authors emphasize the need for a high index of suspicion for dual infections in settings where tuberculosis and cryptococcosis occur. The case illustrates that symptoms compatible with either form of meningitis may accompany coinfection, making laboratory evidence important for establishing the diagnosis. In this patient, cerebrospinal fluid testing supported the presence of both pathogens rather than a single infectious cause.

The report also identifies multidisciplinary management and timely neurosurgical intervention as potentially important in cases accompanied by hydrocephalus and raised intracranial pressure. This is presented as a general implication of the case, not as evidence that a particular coordinated team or intervention was delivered in this patient. The supplied account does not establish the timing, type or outcome of any neurosurgical procedure.

As a single-patient case report, the research has a narrow evidentiary scope. It cannot determine the incidence of concurrent tuberculous and cryptococcal meningitis, identify the immune mechanisms or predisposing factors that permit coinfection, or compare management strategies. Its principal reported value is to document severe dual CNS infection with fatal hydrocephalus in an HIV-negative patient without a known immunosuppressive condition, while showing the diagnostic difficulty created by overlapping clinical features.

Subject of Research: Fatal concurrent tuberculous and cryptococcal meningitis with hydrocephalus in an HIV-negative patient

Article Title: Concurrent tuberculous and cryptococcal meningitis with hydrocephalus in an HIV- negative patient: a fatal case report

Article References: Tesfaye, M. Z., Gebretekle, F. T., Mered, E. F., Talemahu, A. A., & Kassie, H. T. (2026). Concurrent tuberculous and cryptococcal meningitis with hydrocephalus in an HIV- negative patient: a fatal case report. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14455-4

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14455-4

Keywords: tuberculous meningitis, cryptococcal meningitis, concurrent infections, HIV-negative patient, hydrocephalus, intracranial pressure, cerebrospinal fluid, case report, Concurrent, tuberculous, cryptococcal, meningitis

The research, published in the journal Neurocritical Care, was conducted by a team of critical care physicians, neurologists, radiologists, and clinical pharmacists affiliated with Kasr Al-Ainy Medical School at Cairo University Hospital and collaborating institutions in Egypt. The investigators enrolled 100 neurocritical intensive care unit patients with acute, nontraumatic brain injury, either ischemic stroke caused by a blocked vessel or intracerebral hemorrhage driven by hypertension. All patients had severely depressed consciousness, scoring below 9 on the Glasgow Coma Scale, the standard measure of awareness that ranges from 3 for unresponsive to 15 for fully alert. Each patient underwent a computed tomography scan of the brain on admission and a bedside ultrasound measurement of the optic nerve sheath diameter.

The logic behind the technique rests on a quirk of anatomy. The optic nerve is not a free-floating cable; it is wrapped in a sleeve of dura mater, the same tough membrane that encases the brain, and this sleeve communicates with the space around the brain where cerebrospinal fluid circulates. When intracranial pressure rises, fluid is pushed along the nerve, and the sheath distends like a pressurized hose. Because the sheath is most distensible just behind the eyeball, a few millimeters of measurement through the orbit can serve as a proxy for pressure deep inside the skull. The probe emits high-frequency sound waves, receives the echoes, and displays the nerve and its sheath in cross-section, allowing clinicians to measure the diameter in seconds at the bedside.

To establish what a normal reading looks like, the researchers also recruited 60 age- and sex-matched healthy volunteers. Importantly, these controls were included only to provide descriptive reference values for physiological optic nerve sheath diameter; they were not part of the primary diagnostic or prognostic analyses. The results formed a striking staircase. Healthy volunteers averaged 4.62 millimeters, with a standard deviation of 0.33. Stroke patients whose CT scans showed no radiologic signs of elevated pressure averaged 5.52 millimeters. Patients whose scans did show CT-based surrogates of raised intracranial pressure averaged 6.50 millimeters, nearly two millimeters wider than the healthy cohort.

Because the study used CT-based radiologic surrogates as the reference standard rather than invasive pressure monitors, the team could calculate how well the ultrasound measurement discriminated between patients with and without imaging evidence of elevated pressure. Using a cutoff derived from their own cohort of 6.00 millimeters or greater, the measurement achieved 86.7 percent sensitivity, meaning it correctly identified nearly nine out of ten patients with CT-positive elevated pressure, and 85.0 percent specificity, correctly clearing about nine out of ten patients without it. The positive predictive value was 89.7 percent, the negative predictive value 81.0 percent, and overall accuracy 86.0 percent. The area under the receiver operating characteristic curve, a summary of discrimination that ranges from 0.5, no better than chance, to 1.0, perfect, reached 0.901, a figure that places the technique in the range of genuinely useful clinical tests.

Diagnostic accuracy alone would be noteworthy, but the prognostic findings may carry even greater weight. In an exploratory survival analysis, patients whose optic nerve sheath diameter exceeded the 6.00 millimeter threshold had significantly higher in-hospital mortality than those below it, with a log-rank p-value of 4.77 multiplied by 10 to the power of minus 13, an extraordinarily small probability of observing such a survival difference by chance. In other words, a single number obtained with a handheld probe in the first hours of admission carried information about who would live or die during the hospitalization.

The researchers did not stop at raw associations. Elevated intracranial pressure tends to accompany other adverse features, including deeper coma and lower blood pressure, so a crude link between sheath diameter and outcome could simply reflect that patients with big hemorrhages have both wide sheaths and poor prognoses. To address this, the team ran exploratory multivariable models adjusting for age, body mass index, Glasgow Coma Scale score, and systolic blood pressure. Even after these adjustments, an optic nerve sheath diameter above the threshold remained independently associated with CT-positive elevated intracranial pressure, supporting the interpretation that the ultrasound signal reflects pressure itself rather than merely the severity of the underlying brain insult.

Why does this matter globally? Invasive intracranial pressure monitoring requires neurosurgical expertise, sterile operating conditions, and equipment that many hospitals in low- and middle-income settings simply cannot guarantee around the clock. Egypt, like much of the world, faces a substantial burden of hypertensive intracerebral hemorrhage, a condition in which pressure dynamics evolve rapidly and every hour of unrecognized elevation worsens outcomes. A noninvasive tool that any trained clinician can deploy with a portable ultrasound machine changes the calculus of triage: it can identify patients who need aggressive pressure-lowering therapy, transfer to a facility capable of invasive monitoring, or repeated serial examinations to track an evolving crisis. The authors specifically highlight its potential for early triage and longitudinal bedside surveillance, particularly when invasive monitoring is unavailable or delayed.

The technique is not without caveats. The reference standard here was radiologic rather than a directly measured intracranial pressure in millimeters of mercury, which means the study validates ONSD against imaging surrogates rather than against the gold standard itself. The cutoff of 6.00 millimeters was derived from this specific cohort, and different populations or different ultrasound protocols may warrant different thresholds; indeed, systematic reviews and meta-analyses across traumatic brain injury, pediatric populations, and other neurological conditions have reported a spread of proposed cutoffs, reflecting variation in technique, anatomy, and patient mix. Measurement of the sheath also demands careful technique and operator training, which is why international consensus groups have published quality criteria for image acquisition and measurement. Furthermore, the survival and adjustment analyses are described by the authors as exploratory, and the prognostic signal, however striking, would need confirmation in independent cohorts before sheath diameter becomes a standard element of stroke prognostication.

Nevertheless, the study adds an important piece of evidence in a field long dominated by data from traumatic brain injury. Nontraumatic brain injury patients, those with spontaneous strokes and hemorrhages, have been underrepresented in the ultrasound literature, and this prospective cohort of 100 severely affected patients provides some of the clearest data yet that the eye can serve as a window on the injured brain. The stepwise increase from healthy controls to CT-negative patients to CT-positive patients is internally consistent, the discrimination statistics are strong, and the mortality signal is dramatic. If larger, multicenter studies reproduce these findings, a probe, a dollop of gel, and a closed eyelid could become part of the standard first assessment of every comatose stroke patient, turning a routine ultrasound exam into a rapid, radiation-free estimate of one of neurocritical care’s most feared variables.

Subject of Research: Optic nerve sheath diameter ultrasonography for detecting elevated intracranial pressure and prognosis in ischemic stroke and hypertensive intracerebral hemorrhage

Article Title: Diagnostic and Prognostic Value of Sonographic Optic Nerve Sheath Diameter in Patients with Ischemic Stroke or Hypertensive Intracerebral Hemorrhage

Article References: Alasar, H., Kandeel, A., Hamdi, M., Fotooh, A. M. A., Amin, F., Abdeltawab, M., & Kenawy, M. (2026). Diagnostic and Prognostic Value of Sonographic Optic Nerve Sheath Diameter in Patients with Ischemic Stroke or Hypertensive Intracerebral Hemorrhage. Neurocritical Care. https://doi.org/10.1007/s12028-026-02697-8

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02697-8

Keywords: optic nerve sheath diameter, intracranial pressure, ultrasonography, ischemic stroke, intracerebral hemorrhage, neurocritical care, diagnostic accuracy, prognosis, point-of-care ultrasound, Glasgow Coma Scale, ICU monitoring, computed tomography

The research, led by Peter Galos and Ludvig Hult of Uppsala University together with colleagues in Sweden and Italy and the CENTER-TBI High-Resolution ICU investigators, set out to answer a question that haunts almost every machine learning study in medicine: does a model trained on data from one hospital actually work anywhere else? The team had previously developed an XGBoost model, a gradient-boosted decision tree algorithm prized for combining strong predictive accuracy with modest computational demands, using a rich dataset of high-frequency physiological recordings from 602 patients treated at Uppsala University Hospital between 2008 and 2019. That model achieved an area under the receiver operating characteristic curve of 98.7 percent in its home setting, but single-centre success is a notoriously unreliable predictor of real-world performance.

To test generalisability, the researchers turned to the Collaborative European NeuroTrauma Effectiveness Research in TBI project, known as CENTER-TBI, a landmark European initiative that collected high-resolution physiological monitoring data, sampled at 100 hertz or higher, from patients with traumatic brain injury under standardised conditions. After rigorous data integrity checks removed patients with missing demographics or inconsistent recordings, the validation cohort comprised 255 patients from 21 different centres, treated between 2014 and 2017. The two cohorts were broadly similar in age and sex distribution, with roughly four out of five patients male in both, but they were not identical: the CENTER-TBI patients had more severe injuries on average, with a median admission Glasgow Coma Scale score of 5 compared with 7 in Uppsala, and anticoagulant use was four times more common, at 16 percent versus 4 percent.

The predictive task itself was defined with clinical precision. An intracranial pressure insult was declared to begin when pressure exceeded 20 millimetres of mercury for five consecutive minutes, a threshold consistent with internationally accepted neurointensive care guidelines, and to end when pressure remained at or below that level for five consecutive minutes. The model was asked to examine a four-hour historical window of minute-level averages of intracranial pressure and invasive arterial blood pressure, and to predict whether an insult would begin within the next thirty minutes. Only features computable from these two pressure signals were used, deliberately excluding demographics and treatment data, which kept the model lightweight and focused on the physiological trajectory itself.

The results were striking. When the model trained on the Uppsala cohort was tested on the multicentre CENTER-TBI data, it achieved an AUROC of 97.4 percent, with sensitivity of 93.8 percent and specificity of 98.6 percent, meaning it correctly identified nearly 94 percent of impending pressure insults while raising false alarms on fewer than 2 percent of safe windows. Balanced accuracy exceeded 90 percent. When the direction was reversed, with the model trained on CENTER-TBI and tested on the Uppsala data, performance remained high, with an AUROC of 96.1 percent and specificity of 98.7 percent, though sensitivity dropped to 81.2 percent. In both directions, balanced accuracy stayed above 88 percent, a level of cross-centre robustness that few clinical prediction models achieve.

That asymmetry between the two validation directions is scientifically revealing rather than a simple failure. The authors interpret it as evidence of domain shift, the phenomenon whereby the statistical relationship between physiological patterns and outcomes differs subtly between populations. The CENTER-TBI cohort experienced intracranial pressure insults in 16.8 percent of prediction windows, nearly triple the 6.4 percent rate seen in Uppsala, reflecting both a more severely injured patient mix and possibly different management practices. A model trained on the higher-prevalence, more severe cohort appears to miss a larger share of insults when applied to the milder Uppsala population, while the reverse direction preserved sensitivity almost perfectly. The finding underscores a central truth of clinical machine learning: external validation is essential, and even strong AUROC figures do not guarantee identical behaviour across settings.

The study also grapples with a subtler statistical trap: class imbalance. Because dangerous pressure surges are relatively rare events at the level of individual time windows, metrics such as positive predictive value, accuracy and F1-score depend heavily on how common insults are in the test population. The same model, making equally good predictions instance by instance, can look better or worse simply because the prevalence of the event has changed. This is why the positive predictive value ranged from 58 to 70 percent across configurations even as discrimination remained excellent. The authors caution that these prevalence-dependent metrics must be interpreted in context, and that the apparent differences between cohorts partly reflect this statistical artefact rather than a genuine collapse in model quality.

What would deployment actually look like? In a future clinical setting, an early warning from the model would prompt clinicians to hunt for the suboptimal physiological conditions that typically precede an intracranial pressure crisis: insufficient sedation, hyponatraemia, impaired venous outflow from the head, altered carbon dioxide levels, fever, or seizures. Caught early, these can often be corrected before pressure ever crosses the danger threshold, shifting care from reactive to preventive. But the researchers flag a fascinating feedback problem: if clinicians successfully intervene, they change the very physiological trajectory the model was trained on, potentially altering its performance over time. Real-time deployment studies will therefore be needed to assess clinical utility, alarm burden and how staff respond to alerts in practice.

The path to the bedside will also require local calibration. The decision threshold that converts the model’s risk score into an alarm involves an unavoidable trade-off: a low threshold maximises sensitivity and catches more true insults but floods the unit with false alarms, while a higher threshold reduces alarm fatigue at the cost of missed events. Because data characteristics, monitoring equipment and patient populations vary between hospitals, the authors recommend that each centre evaluate and, where necessary, retrain the model locally before implementation. They point to adaptive strategies such as site-specific retraining and transfer learning, in which a model pre-trained on large multicentre data is fine-tuned on a smaller local dataset, as promising directions for future work. Including treatment variables and analysing periods during and between active insults separately may further sharpen performance and illuminate the underlying pathophysiology.

The study has limitations worth noting. The feature set was restricted to physiological data, and while the earlier Uppsala work showed that adding baseline variables such as age, sex and Glasgow Coma Scale score yields only marginal improvement, treatment variables remain unexplored. Excluding incomplete recordings may have introduced selection bias, and the low prevalence of insults at the window level limited the statistical power of subgroup analyses. Yet the headline conclusion stands firm: a computationally lightweight XGBoost model, fed nothing more than minute-by-minute intracranial and arterial blood pressure readings, retained strong discrimination for impending intracranial hypertension across 21 European centres and two very different patient cohorts. The finding that pre-insult physiological patterns are shared across hospitals suggests that the brain signals its distress in a recognisable language, and that machines, given enough high-quality data, are learning to read it. For the roughly hundreds of thousands of people worldwide who sustain severe traumatic brain injuries each year, an AI early warning system that buys clinicians thirty extra minutes could one day mean the difference between recovery and irreversible damage.

Subject of Research: External multicentre validation of a machine learning model for early prediction of intracranial hypertension after traumatic brain injury

Article Title: Artificial Intelligence for Early Prediction of Intracranial Hypertension: From Single-Centre Development to Multicentre Validation

Article References: Galos, P., Hult, L., Citerio, G., Lewén, A., Hånell, A., Schön, T. B., Enblad, P., In collaboration with The CENTER-TBI High-Resolution ICU (HR ICU) Sub-Study Participants and Investigators, Anke, A., Beer, R., Bellander, B.-M., Beqiri, E., Buki, A., Cabeleira, M., Carbonara, M., Chieregato, A., Clusmann, H., Czeiter, E., Czosnyka, M., … Zeiler, F. A. (2026). Artificial Intelligence for Early Prediction of Intracranial Hypertension: From Single-Centre Development to Multicentre Validation. Neurocritical Care. https://doi.org/10.1007/s12028-026-02633-w

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02633-w

Keywords: traumatic brain injury, intracranial pressure, intracranial hypertension, XGBoost, machine learning, artificial intelligence, external validation, neurointensive care, CENTER-TBI, predictive medicine, early warning system, clinical decision support

The research team, led by Annika Meyer of University Hospital Cologne together with colleagues from several German institutions, conducted the review in accordance with PRISMA 2020 guidelines and registered the protocol prospectively before data collection began. In January and February 2026, the investigators searched PubMed, Web of Science, Dimensions AI, The Lens, and the Cochrane Library, supplemented by screening of Google Scholar records and, notably, by artificial intelligence-assisted literature discovery tools including ChatGPT 5.2, OpenEvidence, and SciSpace, which were used in a supplementary role to catch studies that conventional database queries might have missed. The search yielded 778 records, from which ten studies encompassing 177 neurocritical care patients in Europe and China ultimately met the inclusion criteria. Eight of those studies, covering 120 patients, provided extractable data for the meta-analysis of the primary outcome: the change in intracranial pressure from a supine baseline to the prone position.

The pooled results showed a statistically significant but modest rise in intracranial pressure of 5.06 millimeters of mercury when patients were turned prone, with a 95 percent confidence interval spanning 2.05 to 8.06 millimeters of mercury. Between-study heterogeneity was substantial, with an I-squared statistic of 76.7 percent, reflecting the wide variety of protocols, patient populations, and positioning techniques across the included studies. In descriptive terms, pooled mean intracranial pressure increased from 11.10 millimeters of mercury at baseline to 16.13 millimeters of mercury during proning. Crucially, both values remained below the treatment thresholds commonly used to define intracranial hypertension, which typically begin around 20 to 22 millimeters of mercury. The increase, in other words, was real but generally did not push patients into the danger zone.

Perhaps the most clinically important finding concerned what happened after patients were turned back. In five studies with 45 patients, the pooled change in intracranial pressure between the supine baseline and the period after returning to supine was minus 0.41 millimeters of mercury, with a confidence interval of minus 2.34 to 1.51, meaning there was no meaningful difference from pre-proning values. The pressure elevation during proning was therefore transient and fully reversible, a position-dependent effect rather than a sustained deterioration of intracranial dynamics. Cerebral perfusion pressure, the gradient that actually drives blood flow to brain tissue, showed no significant change either during proning or afterward, with mean differences close to zero and wide confidence intervals. Only one included study reported vasopressor adjustments, and no changes in vasopressor requirements were observed during the prone phase.

Against this modest hemodynamic cost, the respiratory benefits were consistent and striking. Every one of the ten studies that reported systemic oxygenation parameters, whether the partial pressure of arterial oxygen alone or the PaO2-to-FiO2 ratio, documented improvement during prone positioning. Four studies went further and measured invasive brain tissue oxygenation directly, using parenchymal sensors, and found that brain tissue oxygen levels also rose while patients were prone. This detail carries particular weight because, in three of the four studies, baseline brain tissue oxygenation values fell within ranges previously associated with cerebral ischemia and unfavorable neurological outcomes. In one study, brain tissue oxygenation improved during proning even as cerebral perfusion pressure declined slightly, suggesting that better oxygenation of arterial blood can translate into better oxygen delivery to tissue even when perfusion pressure shifts modestly.

The physiological mechanism behind the transient pressure rise is well understood from decades of neurosurgical anesthesia practice. Prone positioning increases intrathoracic and intraabdominal pressures, which can impede cerebral venous outflow, and suboptimal head and neck alignment can compromise jugular venous drainage. Because the skull is a rigid compartment governed by the Monro-Kellie doctrine, any increase in venous blood volume must be offset by displacement of cerebrospinal fluid or arterial blood, and patients with limited intracranial compliance have less reserve to absorb the change. Consistent with this framework, several included studies found that tolerance of proning was usually apparent within the first hour and was associated with lower baseline intracranial pressure. One study observed larger pressure increases in patients without external ventricular drainage, underscoring the role of cerebrospinal fluid diversion in preserving intracranial reserve.

The clinical relevance of a five-millimeter-of-mercury rise is therefore highly patient dependent. In individuals with preserved compensatory reserve, the transient elevation can be buffered by normal autoregulatory displacement of fluid compartments. In patients already hovering near the threshold of intracranial hypertension, however, the same increase may exhaust the remaining reserve and trigger an exponential rise in pressure. This helps explain why, across six studies that reported it, roughly 8.89 percent of patients had prone positioning prematurely discontinued because of intracranial pressure elevation, with a confidence interval of 4.21 to 17.78 percent. The authors suggest a pragmatic interpretation: rather than treating proning as a binary yes-or-no decision, clinicians should implement it as a monitored therapeutic trial with an explicit early reassessment point, typically within the first hour, when vulnerability is most likely to declare itself.

The evidence base behind these conclusions comes with important caveats that the authors address candidly. Only two of the ten studies were randomized trials, and one of those was judged to have a high risk of bias due to problems with the randomization process and missing outcome data, while the other raised some concerns. The eight nonrandomized studies carried an overall moderate risk of bias, driven primarily by serious confounding and selective reporting. Using the GRADE framework, the overall certainty of evidence was rated very low. Notably, patients with intracranial pressure monitoring have been excluded from roughly 75 percent of prior trials evaluating prone positioning for acute respiratory distress syndrome, which is precisely why this population has remained so poorly characterized. Additional limitations include inconsistent control of arterial carbon dioxide levels, a major determinant of cerebral blood flow, and the fact that brain tissue oxygenation reflects only regional rather than global oxygen delivery.

The heterogeneity across studies also extended to technique, and the authors argue that positioning details are far from trivial technicalities. Whether patients were placed fully prone or semi-prone, whether the head was elevated or horizontal, and whether the head was kept neutral or rotated are likely to be major determinants of intracranial tolerance. Turning maneuvers themselves can provoke coughing, ventilator desynchrony, and sympathetic activation, all of which acutely raise intracranial pressure, while concomitant spinal injuries, which are relatively common in traumatic brain injury, can complicate mobilization. Excessive sedation stabilizes physiology but delays neurological assessment, a tradeoff that cannot be eliminated but can be managed through protocolized practice.

The bottom line for clinicians is a cautious green light. The findings support considering prone positioning in selected neurocritical care patients with moderate-to-severe hypoxemic respiratory failure when invasive intracranial pressure monitoring is in place and timely de-escalation pathways are available, in line with European Society of Intensive Care Medicine consensus recommendations that already endorse proning when intracranial pressure is not elevated. Because hypoxemia is a major driver of secondary brain injury, and pulmonary complications affect an estimated 5 to 30 percent of acute brain injury patients while worsening outcomes and prolonging intensive care stays, the stakes of leaving respiratory failure untreated are considerable. The authors emphasize that current evidence supports a cautious approach but not definitive thresholds or protocols, and they call for larger prospective studies using standardized proning protocols, time-resolved intracranial pressure data, stratification by baseline intracranial reserve, and patient-centered neurological outcomes. Until such trials arrive, the message from this analysis is that the feared catastrophe of proning the injured brain is, in most monitored patients, a manageable and reversible bump rather than a cliff edge.

Subject of Research: Effects of prone positioning on intracranial pressure and oxygenation in neurocritical care patients with invasive intracranial pressure monitoring

Article Title: Prone Positioning in Neurocritical Care Patients with Invasive Intracranial Pressure Monitoring: A Systematic Review and Meta-analysis

Article References: Meyer, A., Overbeek, R., Mühlhausen, L., Küchler, J., Ditz, C., Foit, A., Hof, M., Goldbrunner, R., & Steinbicker, A. U. (2026). Prone Positioning in Neurocritical Care Patients with Invasive Intracranial Pressure Monitoring: A Systematic Review and Meta-analysis. Neurocritical Care. https://doi.org/10.1007/s12028-026-02667-0

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02667-0

Keywords: prone positioning, intracranial pressure, neurocritical care, acute brain injury, ARDS, cerebral perfusion pressure, brain tissue oxygenation, systematic review, meta-analysis, traumatic brain injury, subarachnoid hemorrhage, intensive care medicine

The study, led by Rhona C. B. Muyise and Samuel Jjunju of the Infectious Diseases Institute at Makerere University, together with collaborators at the University of Minnesota and other institutions, was designed as a post-hoc analysis of a prospective cohort assembled from the ASTRO, AMBITION, ENACT, and FLOOR clinical trials. Of 1,055 participants diagnosed with cryptococcal meningitis across these trials, 787 survived their initial hospitalization and could be followed for more than ten weeks after discharge. This large, carefully characterized population allowed the investigators to measure not only how often rehospitalization occurred, but why it occurred and what it meant for survival—questions that have remained largely unanswered despite substantial advances in the treatment of cryptococcal disease itself.

The technical details of the cohort underscore how profoundly immunocompromised these patients were. The median CD4 cell count at baseline was just 21 cells per microliter of blood, with an interquartile range of 8 to 58—a level of immune depletion at which the body’s defenses against fungal pathogens are essentially absent. Only 45 percent of participants had prior experience with antiretroviral therapy, meaning that more than half were starting HIV treatment for the first time while simultaneously battling a life-threatening fungal infection of the brain and its surrounding membranes. Cryptococcal meningitis is caused by Cryptococcus neoformans and related species, environmental yeasts that people inhale routinely but that cause disease almost exclusively in severely immunosuppressed individuals.

The headline numbers are stark. Within ten weeks of discharge, 17 percent of the 787 survivors—133 patients—were rehospitalized at least once, and 13 percent—101 patients—died. These categories overlapped in important ways: among the 654 participants who did not report rehospitalization, 11.6 percent, or 76 patients, nevertheless died by the ten-week mark without ever returning to hospital. Among the 133 who were rehospitalized, 18 percent, or 25 patients, died during or after that readmission. The median time from initial discharge to rehospitalization was 20 days, with an interquartile range of 10 to 36 days, meaning that half of all readmissions occurred within roughly three weeks of discharge—a period when many patients in resource-limited settings have limited access to follow-up care.

Perhaps the most clinically valuable contribution of the study is its systematic accounting of why patients returned to hospital. The researchers classified the reasons for all 133 rehospitalizations and found that the largest share—92 cases—stemmed from meningitis-related causes: culture-positive relapse of the original cryptococcal infection, paradoxical immune reconstitution inflammatory syndrome, or elevated intracranial pressure. Each of these mechanisms reflects a distinct biological problem. Relapse occurs when viable fungi persist despite antifungal therapy, often because induction treatment was inadequate or adherence lapsed. Paradoxical immune reconstitution inflammatory syndrome, or IRIS, is the inflammatory storm that can erupt when antiretroviral therapy restores immune function and the revived immune system mounts an exuberant response to residual fungal antigens in the central nervous system. Elevated intracranial pressure, a hallmark of cryptococcal meningitis, can persist or recur even after microbiological cure, causing severe headaches, vision loss, and neurological deterioration that require repeated lumbar punctures to relieve.

Beyond the brain, the readmission diagnoses read like a catalogue of advanced AIDS and its complications. Tuberculosis accounted for 26 rehospitalizations, other opportunistic infections for 39, and malignancy for 9. Hematologic disorders, including anemia and cytopenias, drove 21 admissions, while kidney and electrolyte disorders—conditions that can arise from both HIV-associated nephropathy and the nephrotoxic effects of amphotericin-based antifungal regimens—accounted for 29. Another 23 admissions were attributed to other conditions. The breadth of this list is itself a finding: patients surviving cryptococcal meningitis are not merely at risk of their index infection recurring, but remain extraordinarily vulnerable to the full spectrum of AIDS-defining and treatment-related complications during the fragile weeks after discharge.

In an effort to identify who was most likely to be readmitted, the investigators performed statistical analyses of potential risk factors, and the results were unexpectedly null. No baseline characteristic reliably predicted rehospitalization. This absence of identifiable risk factors carries its own clinical message: readmission risk appears to be distributed broadly across the survivor population rather than concentrated in a recognizable high-risk subgroup. In practical terms, clinicians cannot easily single out which discharged patients warrant the most intensive follow-up on the basis of demographic or clinical features alone, which argues for structured post-discharge monitoring for all survivors rather than targeted surveillance of a presumed vulnerable minority.

Where the analysis did find signal was in mortality. Older age was associated with an increased risk of death during the post-discharge period, while two factors were protective: higher body weight and higher Glasgow Coma Scale scores at baseline. The Glasgow Coma Scale, a standard measure of consciousness and neurological function, is already used to stratify severity in cryptococcal meningitis, and its protective association here confirms that patients discharged with better neurological status fare better. The protective effect of higher weight likely reflects overall nutritional reserve and illness severity—patients who are more wasted by advanced HIV disease and prolonged infection have less physiological buffer against subsequent complications. These mortality associations, unlike the rehospitalization analysis, offer concrete variables that clinicians can incorporate into discharge counseling and triage decisions.

The study’s implications reach well beyond Uganda. Cryptococcal meningitis kills an estimated 180,000 people annually worldwide, with the overwhelming burden in sub-Saharan Africa, and even with newer, more effective and better-tolerated regimens such as the single-dose amphotericin strategy validated in the AMBITION trial, survival after hospital discharge has remained poorly characterized. By demonstrating that a quarter of survivors experience rehospitalization or death within ten weeks, the study reframes the post-discharge period as a distinct phase of care deserving its own clinical infrastructure—structured follow-up visits, early antiretroviral therapy initiation with careful IRIS monitoring, screening for tuberculosis and other opportunistic infections, and management of intracranial pressure. The work also highlights the value of linking large clinical trial cohorts to longitudinal outcomes: because the four trials enrolled participants prospectively with standardized follow-up, the researchers could capture readmissions and deaths with a completeness that routine hospital records rarely allow.

For a disease that has seen genuine therapeutic progress over the past decade, the persistence of this post-discharge crisis is a reminder that antifungal drug regimens, however improved, address only part of the problem. The patients in this cohort left the hospital with CD4 counts in the low double digits, often newly started on antiretroviral therapy, and facing a landscape of competing infectious, inflammatory, renal, hematologic, and malignant threats. The Ugandan team’s data make clear that the ten weeks after discharge are not a quiet convalescence but a high-stakes transition, and that health systems serving populations with advanced HIV disease must plan for it accordingly. Whether interventions such as enhanced follow-up clinics, pre-emptive treatment of residual fungal burden, or earlier detection of IRIS and tuberculosis can bend this curve is now the pressing question the study leaves for the field.

Subject of Research: Rehospitalization and post-discharge outcomes among adults with HIV-related cryptococcal meningitis in Uganda

Article Title: Incidence, reasons, and outcomes of rehospitalization among adults with HIV-related cryptococcal meningitis in Uganda: a post-hoc analysis of a prospective cohort

Article References: Muyise, R. C. B., Jjunju, S., Kwizera, R., Dai, B., Adzemovic, T., Mufti, S. S., Meya, D. B., & Boulware, D. R. (2026). Incidence, reasons, and outcomes of rehospitalization among adults with HIV-related cryptococcal meningitis in Uganda: a post-hoc analysis of a prospective cohort. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14594-8

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14594-8

Keywords: cryptococcal meningitis, HIV, AIDS, rehospitalization, mortality, Uganda, immune reconstitution inflammatory syndrome, tuberculosis, intracranial pressure, CD4 count, antiretroviral therapy, BMC Infectious Diseases

The device, known as brain4care or B4C, is built on a principle first reported in 2012. The skull is not as rigid as it appears. With every heartbeat, pulsations of intracranial pressure produce microscopic deformations of the cranial bones, and a highly sensitive mechanical extensometer placed on the temporal region can detect these nanometer-scale movements. The sensor transmits the resulting waveforms over Bluetooth to a mobile device, where cloud processing filters and analyzes the signals in real time, delivering a continuous report that clinicians can interpret at the bedside. Because the technique reads cranial pulsations synchronized with the cardiac cycle, it captures the same fundamental pressure wave that an invasive catheter records inside the head, without ever breaching the skull.

The physiological story centers on the shape of that wave. Each pulse contains characteristic peaks labeled P1, P2, and P3. P1 primarily reflects arterial pulsatility transmitted into the intracranial compartment, while P2 is shaped by the brain’s capacity to accommodate pulsatile volume changes. Intracranial compliance, the brain’s intrinsic ability to absorb volume shifts within the rigid confines of the skull without a dangerous rise in pressure, is the key variable. As compliance deteriorates, P2 grows progressively taller relative to P1, and in severe cases the dicrotic wave P3 may even exceed P1 in amplitude. A rising P2/P1 ratio therefore signals that the compensatory reserve is running out, often before the absolute pressure number crosses a treatment threshold.

To assess how well this noninvasive approach has actually been validated, an international team led by Juliana Caldas and Sérgio Brasil systematically reviewed the clinical literature following PRISMA-ScR guidelines, with the protocol prospectively registered in PROSPERO. The researchers searched MEDLINE, Embase, and the Scientific Electronic Library Online from 2016, when the first clinical study using the device appeared, through May 2025. Out of 986 records identified, 660 duplicates were removed, and after screening 326 articles, 134 abstracts met criteria for full analysis. Case reports, reviews, experimental studies, editorials, and studies in healthy subjects were excluded. Ultimately, 16 studies encompassing 826 patients with acute brain injury made the final cut, conducted across three institutions in Brazil, one in the United States, and one in Poland.

The headline finding is that the P2/P1 ratio, the most consistently evaluated metric, performed respectably as a detector of intracranial hypertension, defined in most studies as an invasive pressure above 20 mm Hg. Across the seven studies reporting diagnostic statistics, cutoff values for the ratio ranged from 1.06 to 1.4, with sensitivity estimates spanning 26.8 to 100 percent and specificity from 45.5 to 96.5 percent. Discriminative performance, quantified by the area under the receiver operating characteristic curve, fell between 0.71 and 0.83, which is generally regarded as moderate to good. Notably, one study found that higher P2/P1 values combined with elevated mean intracranial pressure were associated with early mortality, giving the waveform parameter prognostic as well as diagnostic weight.

The second derived parameter, time to peak, or TTP, measures the interval from the onset of the waveform to its highest point. Shorter values indicate preserved compliance and rapid buffering; prolonged values suggest a stiff, poorly compensating brain. Here the evidence was thinner and more heterogeneous. Reported sensitivities ranged from 33.3 to 85.7 percent with specificities between 50.0 and 73.7 percent, and area under the curve values hovered around a modest 0.69. Yet the largest study identified a TTP threshold of 0.3 as the optimal cutoff, achieving a striking 92 percent specificity, a figure supported by normative data showing that healthy individuals across age groups typically remain below that value. Every one of the 16 studies also demonstrated qualitative morphological concordance between the waveforms recorded noninvasively at the skull and those measured by invasive catheters.

Some of the most intriguing results came from studies that pushed the technology beyond simple threshold detection. One investigation tested a composite Intracranial Compliance Scale built from the P2/P1 ratio and TTP in 72 monitored patients, and the combined score outperformed either parameter alone, reaching an area under the curve of 0.83. A score of zero carried a 100 percent negative predictive value, while a score of three carried a 100 percent positive predictive value for intracranial hypertension. In another study, pairing the B4C waveform analysis with transcranial Doppler ultrasound produced the highest diagnostic accuracy of any combination tested, 0.802, for detecting intracranial hypertension. Machine learning has entered the picture as well: a histogram gradient boosting regression algorithm fed with B4C and Doppler inputs estimated invasive pressure with a strong Spearman correlation of 0.76 and a mean absolute error of just 3 mm Hg, while Doppler alone failed to achieve significant correlation.

The review’s authors are careful about what these numbers mean clinically, and their interpretation is arguably the most interesting part of the story. A P2/P1 ratio of 0.8 or lower showed 92 percent sensitivity for ruling out intracranial hypertension, meaning a low reading could help clinicians avoid unnecessary invasive procedures in patients who are likely safe. Conversely, a ratio of 1.4 or higher strongly suggests dangerous pressure is present. The gray zone in between does not confirm or exclude anything, but it is valuable for tracking how compliance changes over time. Crucially, one study found that patients with pressures near conventional treatment thresholds, around 18 to 22 mm Hg, could still have favorable outcomes when their P2/P1 ratios stayed below 1.2, while similar pressure values paired with higher ratios predicted early death. The shape of the wave, in other words, adds context that a single pressure number cannot, supporting a shift toward individualized, physiology-driven precision medicine in neurocritical care.

None of this makes the device a replacement for the catheter, and the review is candid about the caveats. Methodological quality across the included studies was only moderate on the Newcastle–Ottawa Scale, with scores from 3 to 8 points. No study included a true healthy control group, few adjusted for confounders such as age, comorbidities, or injury severity, and most used cross-sectional or short-term designs that precluded long-term follow-up. The device itself is susceptible to motion artifacts, limiting its use in agitated or pediatric patients, and cardiac arrhythmias such as atrial fibrillation can disrupt the regular pulsatile signal and corrupt derived parameters. Accuracy drops in patients with compromised cranial integrity, such as those who have undergone decompressive craniectomy, and direct equivalence with invasive waveforms is not expected because the two methods measure at different locations with different physical properties. Regulatory approval currently extends only to Brazil and the United States, which constrains broader adoption.

Still, the overall picture is one of a technology that has quietly accumulated a coherent validation record. The B4C sensor was well tolerated in every study, with no reported discomfort or adverse effects, and its continuous, up-to-24-hour monitoring sessions offer far more than the snapshot provided by techniques like optic nerve sheath diameter ultrasound. The authors position it as an adjunct within multimodal neuromonitoring, consistent with the Brussels B-ICONIC consensus on noninvasive intracranial pressure assessment when invasive systems are unavailable, serving as a screening tool that guides escalation to invasive monitoring when compliance begins to fail. What is needed next, they argue, are adequately powered, prospective, multicenter trials with standardized protocols, homogeneous brain injury populations, and longitudinal outcomes to define exactly where the device performs best. If those trials succeed, the humble act of listening to the skull’s silent pulse could become a routine part of keeping dangerously swollen brains alive, in ICUs from São Paulo to settings where a neurosurgeon and a catheter are nowhere to be found.

Subject of Research: Noninvasive intracranial pressure and compliance monitoring using the brain4care skull deformation sensor in acute brain injury

Article Title: Intracranial Pressure Estimation Using B4C Device: A Systematic Review on the Validation Studies

Article References: Caldas, J., Alves, F., Rynkowski, C., Solla, D., Godoy, D. A., Robba, C., Taccone, F. S., & Brasil, S. (2026). Intracranial Pressure Estimation Using B4C Device: A Systematic Review on the Validation Studies. Neurocritical Care. https://doi.org/10.1007/s12028-026-02658-1

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02658-1

Keywords: intracranial pressure, intracranial compliance, brain4care, noninvasive neuromonitoring, acute brain injury, neurocritical care, ICP waveform, P2/P1 ratio, systematic review, intracranial hypertension, skull deformation sensor, machine learning

The research, led by Josef Škola and colleagues from Bulovka University Hospital, Charles University, and collaborating Czech institutions, set out to answer a deceptively simple question. When different clinicians examine the same stretch of physiological waveform, do they agree on what counts as an anomaly? The answer matters enormously because, despite rapid progress in automated signal cleaning, human annotations remain the de facto reference standard against which machine learning detectors are trained and judged. If the humans disagree, the algorithm inherits their confusion, and the ceiling of achievable model performance is set not by computing power but by the consistency of the labels themselves.

To measure that consistency, the team assembled a dataset of extraordinary granularity. From a single-center neurocritical care database of 39 patients with acute brain injury, they selected 100 hours of synchronized ICP and ABP recordings, slicing them into 32,760 ten-second segments. The ten-second window was no arbitrary choice: it matches the standard averaging interval used to compute the pressure reactivity index (PRx), a cornerstone measure of cerebrovascular autoregulation that is notoriously sensitive to waveform contamination. Six annotators spanning the full clinical spectrum, three senior neurocritical care consultants with more than a decade of experience, two junior residents, and one final-year medical student, independently labeled each segment using custom-built software that displayed the target segment within a minute of surrounding context, in fully randomized order.

The scale of the effort is worth pausing on. Each of the 32,760 segments was reviewed by all six annotators for both signals, yielding roughly 393,000 individual judgments. Annotators worked in isolation, unable to communicate about their progress or impressions, with every action logged to preserve procedural integrity. Crucially, the team adopted a deliberately broad definition of “anomaly” rather than the narrower, wildly variable “artifact” definitions that have fragmented the literature. An anomaly was any segment in which the signal departed from the expected physiological-technical pattern in a way that masked or distorted clinically relevant information, a framing borrowed from data science that the authors argue will generalize better across centers and event types.

The headline finding is reassuring at first glance. Anomalies were rare, just 3.99 percent of ICP segments and 5.83 percent of ABP segments, yet overall agreement was superb, with Gwet’s AC2, a statistic robust to rare events, exceeding 0.98 in every configuration. But the devil, as so often, lives in the details. When the researchers separated agreement on anomalies specifically, rather than agreement on the abundant normal segments, cracks appeared. For ICP waveforms, senior clinicians achieved substantially higher anomaly-specific agreement than the mixed-experience team: positive agreement rose by 12.8 percentage points, and Fleiss’ kappa by 13.5. For ABP, by contrast, experience barely mattered at all once the analysis accounted for the fact that segments from the same patient are statistically entangled.

That statistical subtlety deserves attention, because it is where much of the study’s methodological rigor lies. Ten-second segments are not independent observations; each of the 16 contributing patients supplied many consecutive segments, and anomalies cluster within individuals. Treating segments as independent would artificially inflate confidence. The team therefore reported every estimate twice, once at the segment level and once with patient-level cluster-robust confidence intervals derived from bootstrap resampling of whole patients, and anchored all inferential claims to the more conservative patient-level analysis. Under that stricter lens, the experience effect for ICP remained robustly significant, while most nominal differences for ABP dissolved.

Why would expertise matter so much more for intracranial pressure than for arterial blood pressure? The authors point to the nature of the distortions themselves. ABP anomalies tend to be loud and stereotyped: flushing artifacts, damped traces, sudden dropouts that even a medical student can spot instantly. ICP anomalies, by contrast, often hide in the fine structure of the pulsatile waveform, shifts in the relative heights of the P1 and P2 peaks, subtle changes in pulse morphology that blur the line between technical glitch and genuine physiological change in cerebral compliance. Reading those nuances fluently requires years of exposure to both the monitoring technology and the underlying physiology, exposure that routine clinical practice concentrates in specialized neurocritical care units.

The study also delivers a cautionary tale about statistics. Fleiss’ kappa, one of the most widely used reliability measures, suggested only moderate agreement (0.61 to 0.74) even as Gwet’s AC2 reported near-perfect consistency. This is the well-known kappa paradox: when the event of interest is rare, chance agreement inflates and kappa values sink, systematically understating reliability. No single metric tells the whole story, the authors argue, which is why they triangulated across Gwet’s AC2, Fleiss’ kappa, prevalence- and bias-adjusted kappa, positive and negative agreement, and the intraclass correlation coefficient. Only the combination reveals that the high overall agreement coexists with meaningful, modality-specific differences in how anomalies are actually identified.

The implications for the AI pipeline are direct and practical. Because human inconsistency injects noise into training labels, the reliability figures reported here effectively define the performance ceiling any detector trained on such data can reach. For ABP, that ceiling is high and reachable by mixed teams, meaning datasets can be expanded cheaply with junior annotators under light oversight. For ICP, the authors recommend expert consensus or multi-rater agreement to suppress label noise, higher confidence thresholds for ICP-based clinical predictions, and active-learning strategies that route genuinely ambiguous segments to senior review. They also note a subtle trade-off: senior annotators flagged fewer segments overall, suggesting stricter criteria that produce fewer false positives but may discard borderline events a broader pool would capture.

Limitations remain, and the authors are candid about them. The 100 hours of data, though vast in segment count, came from a single center with standardized equipment; the cohort was dominated by physiology within recommended target ranges, leaving extreme states underexplored; and the binary labeling scheme ignores anomaly severity and mechanism. Future work, they suggest, should purposively sample unstable physiology, dissect subtype-specific reliability, and pursue consensus methods such as Delphi studies to sharpen anomaly definitions. But the core message stands as a milestone for the field: reliable human annotation of brain and blood pressure waveforms is achievable at scale in routine neurocritical care, yet it is not a one-size-fits-all enterprise. As hospitals race to build AI on top of their monitoring streams, this study is a reminder that the most important component of any machine learning system may still be the trained human eye, and knowing exactly when that eye needs a decade of experience behind it.

Subject of Research: Interrater reliability of human anomaly annotation in intracranial pressure and arterial blood pressure waveforms for training machine learning models in neurocritical care

Article Title: Data for Machine Learning Models: Clinical Experience Determines Interrater Reliability in Waveform Anomaly Detection

Article References: Škola, J., Posel, Z., Waldauf, P., Falta, P., Soukup, M., Horáková, L., & Rožánek, M. (2026). Data for Machine Learning Models: Clinical Experience Determines Interrater Reliability in Waveform Anomaly Detection. Neurocritical Care. https://doi.org/10.1007/s12028-026-02655-4

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02655-4

Keywords: neurocritical care, intracranial pressure, arterial blood pressure, waveform anomaly detection, interrater reliability, machine learning, artifact detection, clinical expertise, ground truth annotation, pressure reactivity index, acute brain injury, Gwet's AC2

Researchers analysing the high-resolution dataset from the Collaborative European Neuro Trauma Effectiveness Research in Traumatic Brain Injury study, known as CENTER-TBI, developed and tested a three-step framework for evaluating what they call annotation plausibility. Their work, published in the journal Neurocritical Care, examined more than 15,000 annotated interventions across 205 patients and found that a striking proportion of manually entered treatment records were physiologically implausible, a problem with serious consequences for the artificial intelligence tools increasingly built on such data.

The core difficulty is deceptively simple. High-frequency neuromonitoring has transformed neurocritical care research, enabling individualised treatment strategies and predictive models that detect rapid physiological changes. Yet the annotations that contextualise these signals are usually entered manually, often under pressure, and sometimes hours after the event. Staff at the 21 European centres participating in the CENTER-TBI high-resolution sub-study used a touch-screen interface in the ICM+ software to log interventions from nine predefined categories, ranging from osmotherapy and suctioning to physiotherapy and sedation changes. The fixed categories were designed to standardise documentation across centres, but the process remained vulnerable to human error, inconsistency and temporal imprecision.

Those weaknesses matter enormously for modern data science. Supervised machine learning models assume that their training labels are accurate, a condition rarely met in clinical settings. As previous research has highlighted, inconsistent human annotations can compromise model performance, producing incorrect predictions and undermining clinical decision-support systems. Most studies have simply treated annotation imprecision as a limitation to acknowledge rather than a problem to solve. The CENTER-TBI team set out to close that methodological gap with a pragmatic framework that does not require an impossible gold standard, since no definitive ground truth exists for retrospective bedside notes.

The first step of the framework relies on human eyes. Because suctioning and physiotherapy leave short, recognisable fingerprints in the physiological record, such as transient spikes in arterial blood pressure and intracranial pressure, they serve as reference events. A reviewer visually inspected each patient file, tolerating timing errors of up to twenty minutes, and classified the file as showing high, moderate or low evidence of a genuine relationship between annotations and signals. Files with fewer than ten total annotations were automatically deemed low evidence. A second independent reviewer then checked the classifications, achieving substantial agreement, with a Cohen’s kappa of 0.69, and only seven of the 205 recordings required reclassification after consensus review.

The results of this visual screen were broadly reassuring: 60 percent of files showed high evidence, 30.2 percent moderate evidence and 9.8 percent low evidence. Within the high-evidence files, an independent event-level review of 1,769 suctioning annotations found that 91.3 percent were physiologically plausible. In other words, where bedside teams documented consistently, their notes generally matched what the waveforms showed. The framework’s authors argue that this kind of file-level screening can inject a degree of trust into interventions, such as fluid boluses or sedation changes, for which algorithmic validation is not feasible.

The second and third steps turned to automation, using osmotherapy, the administration of mannitol or hypertonic saline to lower dangerous intracranial pressure, as a test case. The algorithm searched a forty-minute baseline window around each of the 388 osmotherapy annotations for sustained intracranial hypertension, defined as pressure above 20 mm Hg for at least five consecutive minutes. Events lacking such a baseline were rejected as implausible. Of the 248 events that passed this filter, the post-intervention window was then assessed: 67.7 percent were classified as effective, showing either a pressure drop of at least 10 mm Hg or normalisation, while 32.3 percent were ineffective. The rejected events, 36.1 percent of the total, typically showed intracranial pressure stably below the treatment threshold both before and after the recorded time stamp, strongly suggesting the annotations did not reflect genuine treatment responses.

The most compelling evidence for the framework came from the convergence of its two independent strategies. Among osmotherapy events from high-evidence files, only 18 percent were rejected by the algorithm, compared with 44.1 percent from moderate-evidence files and 57.3 percent from low-evidence files, a highly significant difference that persisted across alternative pressure thresholds of 15 and 25 mm Hg. Two methods, one qualitative and human, one quantitative and automated, arrived at consistent verdicts, supporting the credibility of both. The study also found that annotation counts bore no significant relationship to patients’ six-month functional outcomes, and that documentation density peaked in the first 24 hours of intensive care before declining.

The researchers are careful about what their classifications mean. Treatment response is used here to separate plausible non-responders from annotations lacking a credible physiological context, not to judge the clinical efficacy of osmotherapy itself. Most ineffective events came from well-annotated files and showed prolonged pressure elevation, consistent with genuine treatment failure, whereas rejected events showed consistently low pressures. The team also acknowledges limitations: the framework cannot detect interventions performed but never documented, concurrent procedures can mimic reference signatures, and patients monitored exclusively with external ventricular drains were excluded because drain-open periods interrupt signal acquisition.

Beyond its immediate findings, the study carries a broader message for the era of clinical artificial intelligence. Until intensive care units systematically and automatically record medication types, doses and precise timings, manual annotations will remain the main bridge between physiological signals and clinical meaning, and that bridge needs inspection. The three-step framework, demonstrated for osmotherapy but applicable in principle to any intervention with definable physiological criteria, offers a practical audit tool. The authors call for external validation in other datasets and populations, but their conclusion is clear: in high-stakes neurocritical care, the data feeding tomorrow’s algorithms deserve the same scrutiny as the science built upon them.

Subject of Research: A methodological framework for evaluating the physiological plausibility of time-stamped clinical annotations in high-frequency neuromonitoring data from traumatic brain injury patients

Article Title: Time-Stamped Annotations in High-Frequency Physiological Data: Evaluating an Approach for Assessing Annotation Plausibility in the CENTER-TBI Dataset

Article References: Turella, S., Beqiri, E., Bögli, S. Y., Ianosi, B., Olakorede, I., Zoerle, T., Tas, J., Helbok, R., Smielewski, P., the Smart Neuromonitoring to Support Precision Medicine in Acute Central Nervous Injury (SOPRANI) collaborators and the CENTER-TBI High-Resolution Sub-Study Participants and Investigators, Anke, A., Beer, R., Helbok, R., Bellander, B.-M., Nelson, D., Buki, A., Chevallard, G., Chieregato, A., Citerio, G., … Rehber, C. (2026). Time-Stamped Annotations in High-Frequency Physiological Data: Evaluating an Approach for Assessing Annotation Plausibility in the CENTER-TBI Dataset. Neurocritical Care. https://doi.org/10.1007/s12028-026-02637-6

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02637-6

Keywords: traumatic brain injury, CENTER-TBI, neurocritical care, neuromonitoring, clinical data annotation, osmotherapy, intracranial pressure, machine learning, data quality, ICU, physiological signals, artificial intelligence

The scale of the physiological transformation is striking. The most common technique, decompressive hemicraniectomy, involves removing a bone flap averaging 14 by 14 centimeters and opening the dura widely, which enlarges the cranial cavity by an estimated 120 to 200 milliliters. A meta-analysis cited in the viewpoint quantified the immediate effect: intracranial pressure drops by an average of 17.59 millimeters of mercury, a reduction sustained at 24 and 48 hours, while cerebral perfusion pressure rises by 7.37 millimeters of mercury. Cerebral blood flow and brain tissue oxygenation typically improve as well. But the surgery also produces a cascade of adverse changes: rapid outward brain expansion, impaired cerebrovascular autoregulation, disrupted cerebrospinal fluid dynamics that predispose patients to hydrocephalus, altered intracranial pressure waveform morphology, and even cardiovascular effects mediated through autonomic mechanisms.

The central paradox the authors highlight is that decompressive craniectomy is a temporizing hemodynamic intervention. It relieves the mechanical consequences of high intracranial pressure but does nothing to reverse the underlying structural brain injury. This means the rationale for continued invasive neuromonitoring remains as strong as ever, despite lingering skepticism in some centers that clinical assessment and neuroimaging suffice once the skull is open. Two major consensus statements support continued monitoring: the Milan Consensus Conference recommended intracranial pressure monitoring after secondary decompressive craniectomy to assess the effectiveness of the surgery and guide further therapy, and an international neurosurgical consensus stated that monitoring and treatment should continue until pressure values remain controlled and stable.

The empirical data make a compelling case. Observational studies show that intracranial hypertension persists or recurs in 36 to 76.5 percent of traumatic brain injury patients after decompressive craniectomy, with similar rates reported after surgery for malignant middle cerebral artery infarction. The rationale for continued monitoring includes persistence of the underlying pathophysiology, the risk of recurrent intracranial hypertension or de novo hematomas, and the need to rationalize therapies such as osmotherapy, sedation, and cerebrospinal fluid drainage. The authors also draw a critical distinction between primary decompressive craniectomy, performed as the first intervention, and secondary surgery after conventional pressure management fails. The greatest lesion evolution occurs in the first 8 to 12 hours after the initial insult, a window that coincides with the early postoperative period in primary cases where no prior monitoring was in place. Contralateral epidural hematoma, a well-recognized complication, occurs predominantly after primary surgery, providing an additional argument for sustained surveillance.

Where exactly to place the pressure sensor turns out to be a question with no evidence-based answer. Neither current consensus statements nor major decompressive craniectomy trials report data on sensor implantation site. The limited published series favor ipsilateral placement, on the reasoning that intracranial pressure is compartmentalized, interhemispheric pressure gradients may develop, and regional hypertension can persist or emerge despite decompression. The authors stress that this recommendation rests on pathophysiological reasoning rather than high-quality prospective evidence. Contralateral placement risks underestimating true pressure and brain tissue oxygenation, leaving focal hypoxia or hypertension undetected, though specific scenarios may warrant it: extensive hemispheric parenchymal destruction, large hemorrhages with marked midline shift, diffuse bilateral injury, a nonviable ipsilateral surgical trajectory, infection of pre-existing sensors, or post-primary surgery where no prior monitoring exists.

The choice of monitoring modality adds another layer of complexity. External ventricular drains offer both pressure measurement and therapeutic cerebrospinal fluid diversion but can be technically challenging after surgery because of ventricular distortion. Intraparenchymal probes are easier to deploy but provide only regional measurements without drainage capability. The authors suggest a staged approach, beginning with an ipsilateral parenchymal probe and adding a contralateral ventricular drain in selected patients. Meanwhile, the evidence base for noninvasive monitoring in the decompressed brain remains sparse and contradictory. Optic nerve sheath diameter ultrasonography has shown that postoperative reductions correlate with outcome in malignant stroke and traumatic brain injury, yet other studies report poor performance of both this technique and transcranial Doppler for estimating pressure and perfusion after surgery. The discrepancies likely reflect heterogeneous reference standards, variable timing of measurements, and the fact that optic nerve sheath measurements reflect global rather than regional pressure dynamics.

Even electroencephalography becomes harder to interpret in these patients. The skull defect produces the so-called breach rhythm, high-voltage physiological waveforms that can mimic epileptiform discharges, while magnetoencephalography is less affected by the missing bone. Near-infrared spectroscopy demonstrated significant bilateral improvement in cerebral oxygen saturation after surgery for malignant infarction, though only contralateral changes correlated with 30-day outcome. Novel noninvasive devices that detect cranial vault micro-expansions to reconstruct pressure waveforms are particularly controversial in the absence of a bone flap, and an artificial intelligence-based pulse shape index for classifying intracranial compliance requires especially cautious interpretation in this population. A systematic review and meta-analysis concluded that invasive and noninvasive methods show similar effectiveness after decompressive craniectomy, and a prospective study by Alhamdan and colleagues demonstrated that invasive monitoring provides significant prognostic information after malignant infarction, with autonomic indices such as heart rate variability, baroreflex sensitivity, and signal complexity potentially complementing standard pressure data.

Perhaps the most provocative section of the viewpoint concerns treatment thresholds. Current Brain Trauma Foundation and Seattle International consensus guidelines recommend initiating treatment for intracranial hypertension at pressures above 22 millimeters of mercury, recommendations based on low-level evidence and extrapolated to decompressed patients despite fundamental physiological differences. Multiple observational studies now challenge this. A large prospective analysis identified 19 millimeters of mercury as the value most strongly associated with outcome, a finding maintained in patients undergoing decompressive craniectomy. After surgery for malignant infarction, Hernández-Durán and colleagues found 10 millimeters of mercury as the threshold most strongly associated with mortality, and in a comparable population pressures above 15 millimeters of mercury correlated with unfavorable outcomes, as did a pressure reactivity index near zero or perfusion pressure below 80 millimeters of mercury. A retrospective traumatic brain injury analysis using a 15-millimeter treatment threshold found Glasgow Outcome Scale Extended scores of 1 versus 5 depending on whether pressures exceeded 25 millimeters, and identified 40 millimeters of mercury as the functional lower limit of cerebral perfusion pressure.

A German retrospective cohort that monitored pressures for seven days after surgery adds further weight. Favorable outcomes corresponded to pressures that never exceeded 14 millimeters of mercury, with a mean of 11.5, while unfavorable outcomes corresponded to a mean of 17.5 millimeters, with statistical significance for both traumatic brain injury and malignant infarction independently. Kaplan-Meier analysis showed 75 percent versus 40 percent survival probability respectively. The authors are careful to note that all these data are exclusively retrospective, small-sample, and heterogeneous, and must be considered hypothesis-generating rather than practice-defining. Still, the pattern is consistent: clinically relevant deterioration appears to occur at substantially lower pressures in the decompressed brain than conventional guidelines assume.

The viewpoint distills the problem into six key physiological differences that undermine extrapolation from intact-skull physiology: abolition of the closed-box condition negates the Monro-Kellie doctrine and profoundly alters the pressure-volume relationship; cerebrospinal fluid dynamics are disrupted with predisposition to hydrocephalus; pressure waveform morphology and pulse amplitude characteristics are fundamentally changed; interhemispheric pressure gradients may render single-site measurements unrepresentative; autoregulatory indices require reinterpretation in the altered compliance environment; and noninvasive tools relying on acoustic or mechanical coupling through an intact skull lose reliability without a bone flap. The authors call for prospective, standardized research to fill these gaps and suggest that, in the interim, a structured consensus-based approach could harmonize clinical practice and identify research priorities. For a procedure performed thousands of times each year worldwide, the message is sobering: the brain without its skull is, in monitoring terms, an almost entirely uncharted organ.

Subject of Research: Neuromonitoring physiology and intracranial pressure thresholds after decompressive craniectomy

Article Title: Neuromonitoring in the Decompressed Brain: Rethinking Physiology, Monitoring, and Thresholds After Decompressive Craniectomy

Article References: Godoy, D. A., Paiva, W. S., de Amorim, R. L. O., Rovegno, M., & Rubiano, A. M. (2026). Neuromonitoring in the Decompressed Brain: Rethinking Physiology, Monitoring, and Thresholds After Decompressive Craniectomy. Neurocritical Care. https://doi.org/10.1007/s12028-026-02622-z

Image Credits: AI Generated

DOI: 10.1007/s12028-026-02622-z

Keywords: decompressive craniectomy, intracranial pressure, neuromonitoring, cerebral perfusion pressure, traumatic brain injury, malignant cerebral infarction, neurocritical care, ICP thresholds, sensor placement, noninvasive monitoring, cerebral autoregulation, Monro-Kellie doctrine