The investigators examined 115 consecutive patients with a mean age of 57.2 ± 10.7 years, of whom 95 were men. Every patient had undergone preoperative MR imaging and surgical resection of a hepatocellular carcinoma tumor measuring less than 3 cm. The dataset was randomly partitioned into training and testing cohorts at an 8:2 ratio.

The imaging analysis encoded how features varied across the arterial, venous, and delayed phases of contrast-enhanced MRI. Investigators manually segmented intratumoral regions of interest for each phase and created expanded regions at distances of 3 mm, 5 mm, and 10 mm around the tumor. Five temporal-encoding methods were evaluated with four machine-learning classifiers in models that combined radiomics features and clinical characteristics.

Across the study’s comparisons, models incorporating multi-phase temporal encoding consistently performed better than single-phase models and models relying only on clinical characteristics. The Fourier-inspired temporal encoding method combined with ExtraTrees produced the strongest reported diagnostic performance when applied to the original tumoral region with a 5-mm expansion, yielding an AUC of 0.850 and a 95% confidence interval of 0.675–0.978. The reported result was a performance estimate from the study’s internal training and testing split rather than evidence from prospective or independent external validation.

The findings show that variations across MRI contrast phases contained predictive information for microvascular invasion in this surgically treated population. By encoding changes among arterial, venous, and delayed images rather than analyzing a single phase in isolation, the framework converted cross-phase imaging variation into measurements that could be evaluated by machine-learning classifiers. The study therefore supports the use of multi-phase MRI radiomics as a candidate quantitative approach to preoperative risk stratification, although the evidence reported here comes from a single retrospective cohort of 115 patients with tumors smaller than 3 cm.

The wide confidence interval around the best AUC also indicates substantial uncertainty in the model’s estimated performance.

Subject of Research: Preoperative prediction of microvascular invasion in small hepatocellular carcinoma using multi-phase MRI radiomics and machine learning

Article Title: Preoperative prediction of microvascular invasion in small hepatocellular carcinoma (≤ 3 cm) by leveraging cross-phase variations of MRI contrast phases

Article References: Dou, Y., Liu, Y., Yuan, W., Wu, S., & Yao, Z. (2026). Preoperative prediction of microvascular invasion in small hepatocellular carcinoma (≤ 3 cm) by leveraging cross-phase variations of MRI contrast phases. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02984-y

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02984-y

Keywords: microvascular invasion, hepatocellular carcinoma, contrast-enhanced MRI, multi-phase imaging, radiomics, machine learning, temporal encoding, ExtraTrees, preoperative prediction, liver cancer, Preoperative, prediction

The study’s design was straightforward but rigorous. The researchers collected clinical data and contrast-enhanced cranial MRI images from all patients admitted with intracranial tuberculosis between January 2018 and March 2026. Each patient underwent follow-up lasting at least 12 months, after which outcomes were graded using the modified Rankin Scale, a standard clinical measure of functional disability. Patients scoring 0 to 2, meaning little to no disability, formed the favorable prognosis group of 67 individuals, while those scoring 3 to 6, ranging from moderate disability to death, formed the poor prognosis group of 52 patients.

Contrast-enhanced MRI was central to the investigation because it reveals pathological changes that plain CT and non-contrast imaging often miss. Gadolinium-based contrast agents leak across disrupted blood-brain barriers, highlighting thickened, inflamed meninges, tuberculomas, abscesses, and the characteristic basal enhancement pattern of tuberculous meningitis. This technique allowed the team to classify patients into imaging subtypes and to document complications such as cerebral infarction, hydrocephalus, and cerebral arteritis with far greater sensitivity than conventional approaches.

When the two prognosis groups were compared, striking differences emerged. Patients in the poor prognosis group were older and more frequently male. They more often presented with meningeal-type or mixed-type intracranial tuberculosis rather than the parenchymal form, and they showed significantly higher rates of meningeal thickening, involvement of the basal and pericallosal cisterns, multi-site meningeal involvement, acute cerebral infarction, hydrocephalus, and cerebral arteritis. Conversely, parenchymal-type disease and tuberculomas were less common in the poor outcome group, suggesting that disease confined to brain tissue with discrete granulomas may carry a better outlook than diffuse meningeal inflammation.

To move beyond simple associations, the researchers performed multivariate logistic regression, a statistical technique that isolates the independent contribution of each variable while controlling for all others. Three factors emerged as independent risk factors for poor long-term prognosis: age, acute cerebral infarction, and hydrocephalus. This finding is clinically meaningful because it separates true prognostic drivers from markers that merely travel alongside them. Meningeal thickening and arteritis, for example, were more common in poor-outcome patients but did not retain independent predictive value once infarction and hydrocephalus were accounted for.

The predictive model built from these three variables performed impressively well. Evaluated with receiver operating characteristic analysis, it achieved an area under the curve of 0.861, sensitivity of 70.1 percent, specificity of 89.6 percent, and overall accuracy of 80.7 percent. In practical terms, the model correctly identified nearly nine out of ten patients who would ultimately do well, while capturing seven out of ten of those destined for poor outcomes. An AUC above 0.85 is generally considered strong discrimination for a clinical prediction model, particularly one built on routinely available imaging and demographic data.

Each of the three predictors tells a coherent biological story. Older patients have less physiological reserve, more comorbidities, and blunted immune responses that impair clearance of infection and repair of tissue damage. Acute cerebral infarction in tuberculous meningitis typically results from inflammatory occlusion of small perforating arteries at the base of the brain, where exudate accumulates in the subarachnoid space and triggers vasculitis. Once brain tissue is lost to infarction, recovery is limited, and the infarct itself serves as a proxy for both disease severity and vascular involvement. Hydrocephalus, meanwhile, arises when inflammatory debris obstructs cerebrospinal fluid flow through the ventricular system or impairs its resorption at the arachnoid granulations, raising intracranial pressure and compressing neural tissue.

The Evans index, a ratio comparing the width of the frontal horns of the lateral ventricles to the maximal internal diameter of the skull, was among the imaging variables recorded, providing a quantitative measure of ventricular enlargement. Its inclusion alongside direct assessments of hydrocephalus reflects the study’s effort to capture both qualitative radiological judgments and reproducible morphometric markers. Similarly, the initial Glasgow Coma Scale score was documented for each patient, anchoring the clinical assessment at admission and allowing the investigators to account for baseline neurological status.

The implications for clinical practice are substantial. Patients with intracranial tuberculosis who present with advanced age, evidence of acute infarction on contrast-enhanced MRI, or ventricular enlargement can now be flagged early as high risk, prompting intensified monitoring, earlier consideration of surgical intervention for hydrocephalus such as ventriculoperitoneal shunting, and closer attention to antitubercular therapy optimization. Conversely, patients with parenchymal-type disease and tuberculomas, who showed more favorable outcomes in this cohort, may be candidates for less intensive follow-up protocols, though the authors caution that individualized assessment remains essential.

Limitations of the study are worth noting. As a single-center retrospective analysis, its findings require validation in larger, multicenter cohorts before widespread adoption of the risk model. The retrospective design also means that imaging protocols and clinical management may have varied over the eight-year enrollment period. Nevertheless, the study demonstrates that contrast-enhanced cranial MRI, already a mainstay of diagnosis in intracranial tuberculosis, can be repurposed as a prognostic tool, transforming a routine diagnostic scan into a roadmap for risk stratification and early intervention. As tuberculosis continues to pose a global burden, particularly in regions with limited neurosurgical resources, simple imaging-based predictors of outcome could help clinicians allocate intensive care where it matters most.

Subject of Research: Risk stratification of intracranial tuberculosis patients using contrast-enhanced MRI predictors of long-term prognosis

Article Title: Risk stratification in intracranial tuberculosis patients based on contrast-enhanced MRI

Article References: Wang, Y.-J., Qiu, X.-W., Wang, A.-L., Hu, Y.-N., Li, W.-Q., & Huang, X.-S. (2026). Risk stratification in intracranial tuberculosis patients based on contrast-enhanced MRI. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14479-w

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14479-w

Keywords: intracranial tuberculosis, contrast-enhanced MRI, hydrocephalus, cerebral infarction, tuberculous meningitis, prognosis, risk stratification, neuroradiology, modified Rankin Scale, cerebral arteritis, BMC Infectious Diseases, Risk

The research team, led by Sergio Valencia and Michael S. Gee, analyzed 2,149 pediatric MRI examinations performed under propofol anesthesia between January 2020 and May 2025. The children had a median age of just five years, and roughly six in ten were female. By linking scanner-level operational variables — magnetic field strength, the use of gadolinium-based contrast agents, and whether the protocol covered a single body region or multiple regions — to total examination time and cumulative propofol dose in milligrams per kilogram of body weight, the investigators were able to disentangle which factors genuinely drive anesthetic burden and which are merely correlated with it.

The headline finding is stark. Multi-region MRI protocols, in which the scanner images several anatomical areas in a single session, lengthened examination time by 45.4 percent compared with single-region protocols, with a 95 percent confidence interval of 40.2 to 50.7 percent. Contrast-enhanced examinations, which require the injection of a gadolinium contrast agent and additional imaging sequences, added 18.6 percent to scan duration. Both associations were highly statistically significant, with p-values below 0.001. In practical terms, a child scheduled for a multi-region, contrast-enhanced study could spend well over an hour on the table — and under anesthesia — where a streamlined single-region scan might have taken half as long.

Not every operational factor pushed in the same direction, however. Imaging at 3 Tesla, the higher field strength now standard in many pediatric radiology departments, was actually associated with a 3.9 percent reduction in MRI duration compared with 1.5 Tesla. That modest but statistically significant advantage — the confidence interval ranged from minus 6.8 to minus 0.9 percent, with a p-value of 0.011 — likely reflects the higher signal-to-noise ratio available at 3 T, which allows faster sequences and shorter acquisition times for a given image quality target. The finding is a useful counterweight to the perception that stronger magnets necessarily complicate pediatric workflows; when protocols are designed to exploit the speed advantage, higher field strength can shorten the anesthetic window.

The most consequential number in the study may be the simplest one. Each additional ten minutes of MRI duration was associated with an increase of 0.71 milligrams per kilogram in total propofol exposure, with a 95 percent confidence interval of 0.59 to 0.84 and a p-value below 0.001. Propofol is typically administered as a bolus for induction followed by a continuous infusion, so the cumulative dose scales almost linearly with time under sedation. That linear relationship means every minute trimmed from a protocol translates directly into less drug delivered to a developing brain — a relationship that holds regardless of why the scan took longer.

Critically, when the researchers adjusted their statistical models for MRI duration, the associations between the protocol-related operational factors and propofol exposure were substantially attenuated. This pattern tells a coherent causal story: multi-region protocols and contrast administration do not independently increase anesthetic dose through some pharmacological mechanism of their own; rather, they increase anesthetic exposure because they make the examination longer. Duration, in other words, is the mediator — the pipeline through which operational decisions flow into anesthetic burden. That insight matters for quality improvement, because it identifies a single, modifiable lever: if departments can shorten scans, they can reduce drug exposure without changing anything about how anesthesia itself is delivered.

Why does reducing propofol exposure matter? Propofol is widely regarded as one of the safest and most controllable sedative agents in pediatric practice, and large registry studies have documented its general safety for procedural sedation in children. Yet no anesthetic is entirely free of risk. Sedation and general anesthesia for MRI carry well-documented risks of airway events, respiratory depression, and hemodynamic changes, and large multicenter analyses from the Pediatric Sedation Research Consortium have shown that adverse events, while uncommon, are a persistent feature of out-of-operating-room sedation. Shorter exposure means a narrower window in which such events can occur, fewer drug-related side effects, and faster recovery times in the post-anesthesia care unit.

There is also the unresolved question of neurotoxicity. Preclinical studies have raised concerns that prolonged exposure to general anesthetics during critical periods of brain development can affect neuronal function, prompting a series of large clinical trials. The GAS trial and the Mayo Anesthesia Safety in Kids study, along with a landmark JAMA analysis of a single early anesthetic exposure, have largely been reassuring for brief, single exposures in infancy. But the question of whether cumulative or prolonged exposure carries risk — particularly in children who require repeated anesthetics, such as those undergoing serial imaging for cancer — remains an active area of research. A recent Children’s Oncology Group study examined the impact of propofol exposure on neurocognitive outcomes in children with high-risk acute lymphoblastic leukemia, underscoring that the concern is not theoretical. In that context, minimizing anesthetic time is a prudent default even in the absence of definitive harm data.

The economic argument points in the same direction. Anesthetic-related adverse events impose a measurable financial burden on health systems, and every minute of scanner time occupied by a sedated child is a minute unavailable for other patients. Pediatric MRI demand has grown steadily, and studies of MRI utilization in pediatric accountable care organizations have highlighted sedation and anesthesia as a major operational bottleneck. Faster protocols therefore offer a rare win-win-win: lower anesthetic exposure for the child, reduced risk and cost for the institution, and increased throughput for the waiting list. Previous work by some of the same investigators, including studies of accelerated and abbreviated imaging protocols and of MRI techniques to decrease imaging times in children, has already demonstrated that substantial time savings are technically feasible without compromising diagnostic quality.

The authors are careful to note the limitations inherent in a single-center retrospective design, and the study’s associations — however well-adjusted — cannot fully exclude residual confounding by case complexity. Still, the central message is difficult to escape. The anesthetic burden of pediatric MRI is not an fixed cost of the technology; it is, to a meaningful degree, a design choice embedded in the protocol. When radiologists decide whether to image two regions or five, whether to add contrast, and which field strength to use, they are also, implicitly, deciding how long a five-year-old will lie anesthetized inside a magnet. The new data give that decision a number: roughly 0.71 milligrams per kilogram of propofol for every ten minutes of scan time. As machine learning reconstruction, compressed sensing, and other acceleration techniques continue to compress acquisition times, the operational levers identified in this study suggest that the safest anesthetic for a child’s MRI may be the one that simply ends sooner.

Subject of Research: The association between MRI protocol-related operational factors, examination duration, and propofol anesthetic exposure in pediatric imaging

Article Title: Associations between protocol-related operational factors, MRI duration, and propofol exposure in pediatric MRI

Article References: Valencia, S., Fazio Ferracioli, S., Nichols, J. H., Jaimes, C., & Gee, M. S. (2026). Associations between protocol-related operational factors, MRI duration, and propofol exposure in pediatric MRI. Pediatric Radiology. https://doi.org/10.1007/s00247-026-06766-0

Image Credits: AI Generated

DOI: 10.1007/s00247-026-06766-0

Keywords: pediatric MRI, propofol, anesthesia, sedation, MRI duration, scan protocols, 3 Tesla, contrast-enhanced MRI, radiology, patient safety, neurotoxicity, workflow efficiency