The study used resting-state magnetoencephalography, or MEG, to measure the timing of neuronal activity with greater temporal resolution than blood-oxygen-level-dependent functional magnetic resonance imaging. Participants’ MEG signals were projected onto 400 brain parcels defined by a publicly available atlas, and the researchers calculated parcel-to-parcel connectivity within three canonical networks: the default mode network, executive control network and salience network. This approach allowed the analysis to move beyond a single score for an entire network and test whether particular regional connections behaved differently during acute risk. The researchers described the work as a way to examine both relatively stable vulnerability and state-dependent changes associated with a recent crisis.

Participants came from the National Institute of Mental Health’s Neurobiology of Suicide protocol and were divided into four clinically defined groups. The high-risk group included eight people who had attempted suicide or experienced suicidal ideation with intent to act within the previous two weeks. Twenty-six participants formed a lower-risk group with a history of suicide attempts but no attempt or ideation during the past year. A clinical-control group of 22 had anxiety or mood disorders but no history of suicide attempts or ideation, while 21 minimal-risk participants had no personal or family history of psychiatric disorders or suicidal behavior. The final sample included 47 women and 30 men, with a mean age of 38.41 years.

Groups did not differ significantly in age, sex or ethnicity, but they did differ in mood and anxiety symptoms, medication history and suicidal-ideation scores. The high-risk group had substantially higher current ideation scores than every comparison group and also had greater depression and anxiety severity. A multivariate analytical method called DIABLO—Data Integration Analysis for Biomarker discovery using Latent variable approaches for Omics studies—integrated 2,628 default mode connectivity features, 1,225 executive control features and 1,176 salience features. A discrimination design, selected because the network measures were highly correlated, significantly separated the groups at p = 0.018, whereas an alternative sparse partial least-squares design did not at p = 0.319.

Follow-up analyses indicated that the high-risk group generally had reduced within-network connectivity across all three systems. Reduced default mode connectivity involved the dorsolateral prefrontal cortex in pairs with the retrosplenial or medial prefrontal cortex and with a combined precuneus and posterior cingulate region. Lower executive control connectivity appeared among frontal and parietal areas, including the precuneus, while weaker salience connections joined medial or lateral frontal regions with sensory, insular and parietal areas. One exception was stronger connectivity between the lateral prefrontal cortex and parietal operculum than in the clinical-control group. These were group-level associations, and the analysis identified influential features rather than demonstrating that any single connection could independently predict an individual’s risk.

The relationship between recent ideation and connectivity was more specific than the overall group differences. Among participants in the high-risk group, stronger default mode connectivity between the precuneus or posterior cingulate cortex and the dorsolateral prefrontal cortex, and between the dorsolateral and medial prefrontal cortices, was associated with more severe current suicidal thoughts. The researchers reported interaction effects for these relationships, along with a negative interaction for another posterior default mode connection. In the salience network, stronger connections between the medial frontal and medial parietal cortices and between the parietal operculum and lateral prefrontal cortex were likewise associated with greater ideation severity in the recent-crisis group.

By contrast, executive control connectivity between the precuneus and posterior medial prefrontal cortex showed an inverse association with recent ideation in the high-risk group relative to lower-risk and minimal-risk participants. The investigators used Bayesian linear mixed models to accommodate non-normal data and the limited sample. Models were run with four Markov chain Monte Carlo chains and 5,000 iterations, with additional iterations used when convergence was inadequate. They reported convergence with R-hat below 1.1 and generally acceptable effective sample sizes. These models describe associations with ideation severity; they did not prospectively predict whether a participant would attempt suicide or validate the signatures in an independent cohort.

The results suggest that a recent suicidal crisis is associated with a reorganization of coordinated activity spanning cognitive control, internal thought and detection of salient stimuli. The default mode and salience relationships were generally positive with ideation severity, while the executive control relationship was negative, demonstrating that regions grouped within the same broad network can contribute in different directions. The authors suggested that stronger default mode coupling may reflect intensified self-referential and emotional processing, weaker precuneus–posterior medial prefrontal coupling may relate to impaired cognitive integration, and stronger salience coupling may reflect active emotion regulation or amplified processing of emotionally relevant information. These interpretations were drawn from prior literature rather than tested directly in the present experiment, which did not manipulate death-related cues or measure the cognitive processes proposed as explanations.

Sensitivity analyses complicated the most straightforward interpretation. When depression severity, anxiety severity and medication history were included as covariates, several executive control and salience associations were attenuated, whereas most selected default mode features remained comparatively robust. Because the high-risk group had more severe mood and anxiety symptoms and a more extensive medication history than the comparison groups, the neural patterns cannot yet be cleanly attributed to suicidal crisis alone. Participants were also medication-free or receiving clinically indicated prescriptions, and the sample included both major depressive disorder and bipolar disorder, introducing additional variability. The timing between clinical assessments and MEG collection varied modestly and was longer for some high-risk participants because of scheduling constraints.

Major limitations were the small high-risk sample, the absence of an external validation dataset and possible sensitivity to the chosen brain-parcellation scheme. The group categories also captured different temporal states—recent crisis, historical risk, clinical control and minimal risk—rather than tracking the same individuals through changing levels of risk. The results therefore support the hypothesis that acute suicidal states have electrophysiological signatures that differ from historical vulnerability, but they do not show that those signatures are unique to suicide, are stable over time or can be used in routine screening. Larger longitudinal studies, external replication and samples with closely matched symptom severity and medication histories will be needed to determine whether the network connections retain their value as objective biomarkers of acute risk.

Subject of Research: Magnetoencephalography assessment of brain-network connectivity in recent suicidal crisis and historical suicide risk

Article Title: Brain network-driven predictive neural markers for suicide risk

Article References: Lee, Y., Gilbert, J. R., Chin Fatt, C. R., Stout, J. D., Waldman, L. R., Lamontagne, S. J., Zarate, C. A., Jr., & Ballard, E. D. (2026). Brain network-driven predictive neural markers for suicide risk. Frontiers in Psychiatry, 17, Article 1950068. https://doi.org/10.3389/fpsyt.2026.1950068

Image Credits: AI Generated

DOI: 10.3389/fpsyt.2026.1950068

Keywords: suicide risk, magnetoencephalography, suicidal ideation, brain connectivity, default mode network, executive control network, salience network, DIABLO, electrophysiology, predictive biomarkers, Brain, network-driven

The research team, led by Xianglan Zhou and Lei Shi of the hospital’s radiology department, developed a machine learning approach that combines two types of information: standard clinical characteristics and a technique called delta-radiomics, which captures how the texture and appearance of a tumor changes over time on imaging. Their goal was ambitious but practical. Rather than waiting for a tumor to progress and then performing an invasive biopsy to test for the T790M mutation, they wanted to predict the risk of that mutation emerging before first-line EGFR-TKI therapy was ever started.

The study included 285 patients with non-small cell lung cancer whose disease had progressed after first-line treatment with an EGFR tyrosine kinase inhibitor between January 2013 and September 2019. These patients were randomly divided into a training cohort of 213 individuals and a separate test cohort of 72, allowing the researchers to build their predictive models on one group and evaluate them honestly on another. Four distinct models were constructed and compared: one based purely on clinical characteristics, one on radiomics features extracted from non-contrast-enhanced CT scans, one on delta-radiomics features, and a fourth that combined clinical data with delta-radiomics.

Radiomics itself is a rapidly growing field that converts medical images into vast arrays of quantitative data. Rather than simply measuring the size of a tumor, radiomic algorithms extract hundreds of mathematical descriptors that characterize texture, shape, intensity patterns, and spatial relationships within the imaged tissue. Features derived from grey-level co-occurrence matrices, grey-level run-length matrices, and grey-level dependence matrices capture subtle patterns of pixel intensity that the human eye cannot perceive but that may reflect underlying biological properties such as cell density, necrosis, and vascular structure.

What makes delta-radiomics particularly powerful is its temporal dimension. Instead of analyzing a single scan, it computes the difference between radiomic features extracted from two time points, essentially quantifying how a tumor’s imaging phenotype evolves. This change over time may encode information about tumor heterogeneity and treatment response that a static snapshot cannot. In the context of this study, the delta approach allowed the machine learning models to detect imaging signatures associated with tumors that were biologically primed to develop the T790M resistance mechanism.

The results, measured on the independent test set, told a clear story. The clinical model alone achieved an area under the receiver operating characteristic curve, or AUC, of 0.70, with an accuracy of 66.7 percent. The NECT radiomics model performed worse, with an AUC of 0.63 and accuracy of 57.0 percent. The delta-radiomics model improved on both, reaching an AUC of 0.75 and accuracy of 72.2 percent. But the combined model, integrating clinical characteristics with delta-radiomics features, outperformed them all, achieving an AUC of 0.82 and an accuracy of 79.2 percent. Decision curve analysis, which evaluates the clinical net benefit of a predictive model across different threshold probabilities, further supported the combined model as the optimal choice.

One of the most significant contributions of this work lies not in the raw predictive numbers but in the effort to make the machine’s reasoning transparent. The researchers applied the Shapley additive explanations framework, commonly abbreviated as SHAP, to visualize and interpret how the model arrived at its predictions. SHAP, rooted in cooperative game theory, assigns each input feature a contribution value for every individual prediction, revealing which factors pushed the model toward predicting the presence or absence of the T790M mutation. This interpretability is critical in medicine, where clinicians are understandably reluctant to trust black-box algorithms that cannot explain themselves.

Through the SHAP analysis, the team identified ten significant features driving the model’s decisions: two clinical features and eight radiomic features. While the specific identities of these features reflect the particular dataset and modeling choices of this study, the broader implication is that measurable imaging characteristics, particularly those that change over the course of treatment, carry genuine predictive signal about the evolutionary trajectory of the tumor’s resistance mechanisms. This finding aligns with a growing body of evidence suggesting that the radiographic appearance of a tumor is a window into its genomic and microenvironmental state.

The clinical implications, if validated, could be substantial. Patients identified as high risk for developing the T790M mutation might be candidates for closer surveillance, earlier liquid biopsy testing, or alternative treatment sequencing strategies. Conversely, those predicted to be at low risk might avoid unnecessary interventions. The ability to anticipate resistance rather than react to it represents a meaningful shift in the management of EGFR-mutant lung cancer, where the emergence of T790M has historically been discovered only after disease progression has already occurred.

The authors are careful to note that their work represents a preliminary step rather than a finished clinical tool. The retrospective design, the single-center cohort, and the modest size of the test set all warrant caution, and the researchers explicitly state that further prospective multicentre validation is needed before the model can be applied in clinical practice. Nevertheless, the study demonstrates that the fusion of temporal imaging data with clinical variables, interpreted through explainable artificial intelligence, offers a viable path toward anticipating one of the most consequential resistance mechanisms in lung cancer. As radiomics and machine learning continue to mature, the humble CT scan may become not just a diagnostic record but a predictive instrument, reading the future of a tumor before that future arrives.

Subject of Research: Predicting acquired EGFR T790M resistance mutations in non-small cell lung cancer using delta-radiomics and clinical features with SHAP interpretability

Article Title: Predicting acquired T790M mutations in non-small cell lung cancer after EGFR-TKI therapy: integration of delta-radiomics and clinical characteristics with SHAP interpretation

Article References: Zhou, X., Shi, L., Lin, C., Zou, M., Sun, J., Song, Z., Wang, W., Ren, H., & Shi, L. (2026). Predicting acquired T790M mutations in non-small cell lung cancer after EGFR-TKI therapy: integration of delta-radiomics and clinical characteristics with SHAP interpretation. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02748-8

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02748-8

Keywords: non-small cell lung cancer, EGFR-TKI, T790M mutation, delta-radiomics, machine learning, SHAP, computed tomography, drug resistance, predictive biomarkers, tumor heterogeneity, radiomics, precision oncology

Nucleos(t)ide analogues such as entecavir and tenofovir suppress hepatitis B virus replication with remarkable potency, but they rarely eradicate the reservoir of covalently closed circular DNA that persists in hepatocytes. For HBeAg-negative patients without cirrhosis, treatment guidelines therefore allow, and in some settings encourage, a finite course of therapy followed by careful monitoring. The rationale is that stopping drugs can provoke an immune-mediated flare that, in a minority of patients, leads to durable viral control or even loss of hepatitis B surface antigen, the closest available surrogate of a functional cure. The challenge is that relapse is common, flares can be severe, and clinicians have lacked reliable tools to identify who will benefit and who will simply rebound.

Ethnicity has repeatedly been proposed as one such tool. Observational studies across Europe and Asia have suggested that patients of Asian descent relapse more frequently after treatment cessation than those of European background, and some guidelines have quietly incorporated this notion into patient selection. The problem, as the COIN-B investigators argue, is that ethnicity is a subjective and inconsistently applied construct, blending self-identification, place of birth, language and social factors into a category whose biological content is undefined. To test whether the apparent ethnic effect is real, the team designed a trial that would measure it prospectively and then dissect it with more rigorous instruments: genome-wide ancestry markers and viral genotype.

The trial enrolled long-term nucleos(t)ide analogue-treated patients who were HBeAg-negative at the start of their original treatment and who had no advanced fibrosis, a population considered safe enough to undergo supervised withdrawal of therapy. All participants stopped their antiviral drug under protocol and were followed intensively for 72 weeks, with serial measurements of HBV DNA, hepatitis B surface antigen, liver biochemistry and, where applicable, retreatment when predefined safety or virological thresholds were crossed. Of the 91 patients who discontinued treatment, 84 completed the full 72-week protocol, providing a substantial prospective dataset for the prespecified analyses.

The primary endpoint, off-treatment viral control at week 72, did not differ between the two prespecified ethnic groups. Among 41 patients classified as Caucasian, 41.5 percent maintained viral control, compared with 30.2 percent among 43 non-Caucasian patients, a difference that fell well short of statistical significance with a p value of 0.28. On its face, this negative result might seem to close the door on the ethnicity hypothesis. But the exploratory analyses that followed revealed a more interesting picture, one in which the blunt ethnic dichotomy had been masking genuine heterogeneity that finer-grained variables could detect.

When the researchers replaced self-reported ethnicity with genomic ancestry, inferred from genome-wide markers and quantified as the proportion of DNA attributable to reference populations, patterns emerged that the ethnic categories had obscured. Hepatitis B surface antigen loss, the secondary endpoint and the most desirable outcome short of eradication, occurred in 7.1 percent of participants overall, and every single case arose in a patient with predominantly European or North African and Middle Eastern ancestry. Conversely, a decline of at least one log in surface antigen levels, an accepted on-treatment marker of a favorable immune trajectory, was significantly less frequent among patients with Asian or Sub-Saharan African ancestry, with a p value of 0.01. These findings suggest that ancestry-related factors, whether immunogenetic, viral or both, do influence the probability of a functional cure after treatment withdrawal.

The viral side of the equation proved equally informative. Hepatitis B virus is classified into at least ten genotypes with distinct geographic distributions, and genotype has long been suspected to modulate disease course and treatment response. In the COIN-B analysis, lower hepatitis B surface antigen levels at the end of treatment independently predicted viral control, surface antigen loss and meaningful surface antigen decline, confirming that this biomarker remains the strongest single predictor available to clinicians. But even after adjusting for end-of-treatment surface antigen, patients infected with genotype C or genotype D achieved less viral control and required retreatment more often than those infected with genotype A. The result aligns with a growing body of evidence that viral genotype is not a passive phylogenetic label but a biologically active determinant of how the infection behaves once drug pressure is removed.

Methodologically, the trial illustrates why prospective design and molecular characterization matter in this field. Previous reports linking ethnicity to relapse were largely retrospective, drawn from heterogeneous cohorts treated with different drugs for different durations, and they relied on ethnic labels assigned inconsistently across centers. By registering the comparison in advance, standardizing the cessation protocol across Belgian centers, and applying genome-wide ancestry inference alongside viral sequencing, the COIN-B team was able to separate what the coarse label conflates. The trial is registered as NCT04779970 with ClinicalTrials.gov and as EudraCT 2021-001003-32 with the European clinical trials register, and it was funded in full by the Research Foundation Flanders under its Applied Biomedical Research program, with the funder having no role in the design, analysis or reporting.

The clinical implications are twofold. First, the study cautions against using ethnicity as a proxy for biological risk in treatment cessation decisions. A category that fails to reach statistical significance as a predictor, while its genomic and viral components do, is a poor foundation for patient selection, and its use risks embedding social classification into medical algorithms without justification. Second, the results argue for more granular characterization of patients considered for treatment withdrawal: quantified genomic ancestry, viral genotype and, above all, carefully measured end-of-treatment surface antigen levels. None of these variables alone is sufficient, but together they begin to explain why two patients with identical biochemical profiles can follow sharply divergent courses after the same drug is stopped.

For the field at large, COIN-B reframes a familiar question in more precise terms. The goal is no longer to ask whether a patient’s ethnicity predicts relapse, but to identify the specific host and viral factors that ancestry and genotype merely tag. Genotype-specific viral biology, human leukocyte antigen variation, and ancestry-linked immune gene polymorphisms are all plausible mediators, and disentangling them will require the kind of prospective, molecularly annotated cohorts this trial exemplifies. In the meantime, the study offers clinicians a sober and practical message: the decision to stop nucleos(t)ide analogue therapy should rest on measurable viral and serological variables rather than on broad demographic labels, and even the best current predictors leave most patients still infected, still monitored, and still waiting for a true functional cure.

Subject of Research: Determinants of off-treatment viral control after cessation of nucleos(t)ide analogue therapy in chronic hepatitis B

Article Title: Ethnicity, genomic ancestry, and HBV genotype in off-treatment outcomes following nucleos(t)ide analogue cessation: a prospective multicentre trial (COIN-B)

Article References: Furquim d’Almeida, A., Vanderlinden, A., Ralf, A., Bourgeois, S., Mulkay, J.-P., Sersté, T., Struyve, M., Deressa, B., Sprengers, D., de Vos, M., Callens, J., Shihao, B., François, S., Deltenre, P., Janssens, F., Negrin Dastis, S., Stärkel, P., Orlent, H., Van Roey, G., … Vanwolleghem, T. (2026). Ethnicity, genomic ancestry, and HBV genotype in off-treatment outcomes following nucleos(t)ide analogue cessation: a prospective multicentre trial (COIN-B). Nature Communications. https://doi.org/10.1038/s41467-026-78490-0

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78490-0

Keywords: hepatitis B, HBV genotype, genomic ancestry, nucleos(t)ide analogue cessation, HBsAg loss, functional cure, viral relapse, COIN-B trial, ethnicity, predictive biomarkers, HBeAg-negative, retreatment

The stakes are considerable. Hypertensive disorders of pregnancy, which include gestational hypertension, preeclampsia, chronic hypertension with superimposed preeclampsia and eclampsia, affect between 2 and 15 percent of pregnancies and rank as the second leading cause of maternal mortality worldwide. They sharply raise the risk of acute maternal complications such as renal dysfunction, pulmonary edema, stroke and death, and they contribute to adverse fetal outcomes including small-for-gestational-age infants, preterm delivery and stillbirth. The consequences reach far beyond delivery: women who experience these disorders carry elevated lifetime risks of hypertension, coronary artery disease, stroke and vascular dementia, and their children face higher risks of cardiovascular and neurodevelopmental problems. In the United States alone, hypertensive disorders of pregnancy were estimated to cost the healthcare system 2.18 billion dollars per year as of 2012, largely driven by complications of preterm birth.

What makes the new work compelling is its biological rationale. Preeclampsia is believed to arise in part from maternal vascular inflammation and thrombosis that disrupt placental development, impairing the deep invasion of placental tissue into the maternal decidua between roughly 10 and 16 weeks of gestation. These early deviations from normal placentation precede clinical symptoms by weeks or months, and they share features with the vascular pathology seen in cardiovascular disease. Because the retina is the only place in the body where capillaries, arterioles and venules can be directly and noninvasively visualized, the researchers reasoned that the maternal eye could serve as a living biosensor of the systemic vascular stress that accompanies early placental dysfunction. Prior studies had already found increased retinal arterial tortuosity in women with severe preeclampsia, narrower vessels and reduced capillary density in women who later developed placental insufficiency, and lower vessel densities in both the superficial and deep capillary plexuses of women with the disease.

Current screening, however, remains blunt. Preeclampsia is typically diagnosed only after 20 weeks of gestation, when hypertension, proteinuria or organ dysfunction become clinically apparent. Known risk factors such as prior hypertensive disorders, diabetes, obesity and advanced maternal age provide some predictive value but lack individual-level accuracy. The most validated biomarker, the ratio between the antiangiogenic factor sFlt-1 and placental growth factor, is primarily useful later in pregnancy and in women already suspected of having the disease. First-trimester tools such as the Fetal Medicine Foundation risk calculator combine clinical variables with biomarkers or ultrasound measurements, but these protocols are expensive, logistically complex and inconsistently implemented. In routine practice, prevention-oriented risk stratification often reduces to deciding who qualifies for low-dose aspirin, even though randomized trials show that starting aspirin before 16 weeks can reduce the risk of preterm delivery with preeclampsia by up to 62 percent. Accurate, early identification is the bottleneck that prevents that intervention from reaching the women who need it most.

Visionary AI attacks the problem with an architecture designed for interpretability rather than brute-force pattern recognition. The pipeline begins with ultrawidefield retinal images, captured by scanning laser ophthalmoscopes that cover roughly 80 percent of the retinal surface. A dedicated vessel segmentation model, inspired by generative adversarial network architectures and fine-tuned on hand-annotated ultrawidefield images, converts each photograph into a binary map of the vascular tree. When multiple images of the same eye exist, the segmentations are merged using keypoint matching and deformable registration to recover regions obscured by eyelashes or shadows. The resulting vessel maps are then transformed into three complementary families of features: graph-based measures of branching and connectivity, geometric measures of vessel length, caliber, curvature and tortuosity, and higher-order topological descriptors, including box counting, fractal organization and topological data analysis, that quantify loop structures and hierarchical network organization across scales.

These feature sets feed a stacked ensemble of machine learning models. Each semantically grouped feature set trains independent base learners built on logistic regression, random forests and extreme gradient boosting, and their out-of-fold predictions are integrated by a metalearner. The team evaluated the framework across increasingly stringent settings. In a high-contrast analysis comparing 54 preeclampsia cases against 82 carefully screened healthy controls, the model achieved an AUC of 0.94 and a positive predictive value of 0.85. When the comparison was broadened to a clinically heterogeneous population-wide control pool that deliberately retained women with obesity, diabetes and chronic hypertension, performance remained strong at an AUC of 0.91 plus or minus 0.05. Notably, adding the Fetal Medicine Foundation calculator’s probabilities to the retinal predictions did not materially improve performance, suggesting that the retinal vasculature captures most of the relevant risk signal on its own.

The most demanding test was external. The researchers built a stability-optimized model that prioritized reproducibility over peak performance, using stability-based hyperparameter selection, recursive feature elimination and feature-set deduplication to shrink an initial search space of 96 candidate base learners down to just eight nonredundant learners per fold. This deliberately constrained model was then applied, frozen and untouched, to an independent cohort of 79 pregnant individuals recruited at NYU Langone Health, a population that differed from the development cohort in race and ethnicity distribution, maternal age, in vitro fertilization prevalence and imaging hardware. Restricted to first-trimester images acquired between 6 and 13 weeks of gestation, the model achieved an AUC of 0.81 and an average precision of 0.68, outperforming the official Fetal Medicine Foundation calculator, which reached an AUC of 0.73 and an average precision of 0.50 on the same cohort.

The benchmark against routine clinical practice is perhaps the most striking result. At NYU, aspirin eligibility based on standard clinical criteria identified preeclampsia cases with a sensitivity of 0.62 but at a false positive rate of 0.47 and a prevalence-adjusted positive predictive value of just 0.1. Visionary AI achieved a higher sensitivity of 0.69 at a false positive rate of only 0.13, with a prevalence-adjusted positive predictive value of 0.31. Even within the subgroup of women already recommended aspirin, a clinically enriched high-risk group, the retinal model maintained an AUC of 0.95. The framework also outperformed two direct image-level deep-learning approaches, the retinal foundation model RETFound and a convolutional network, both of which performed near chance in external validation, a finding the authors interpret cautiously but which supports the idea that biologically informed, lower-dimensional vascular representations are better suited to clinical cohorts with limited numbers of outcome events.

Beyond prediction, the features themselves tell a coherent biological story. Recurrently selected features involved retinal vascular topology, geometry, network complexity and hierarchical organization, and model-independent analyses confirmed preeclampsia-associated differences in branch count, topological length, tortuosity, vessel thickness, fractal organization, loop count and vascular asymmetry. The signatures point to vascular simplification through two complementary processes: capillary rarefaction, the loss of small vessels and reduced branching complexity, and arteriolar wall thickening that narrows luminal diameter. Intriguingly, gestational and chronic hypertension produced distinct retinal fingerprints, with localized simplification of hierarchical vascular organization, whereas preeclampsia showed more globally extensive sparsity and disorganization. The authors stress that these associations are hypothesis generating rather than causal, and that the validation cohort, while genuinely independent, remains modest in size with a higher observed preeclampsia prevalence than the institutional average.

The practical appeal is hard to overstate. Retinal fundus imaging requires no laboratory infrastructure, takes less than two minutes, and can be performed by trained nonspecialists, making the approach potentially viable in both high-resource and low-resource settings. The computational pipeline is fully automated, eliminating subjective image grading. Before Visionary AI reaches the clinic, larger and more geographically diverse prospective studies will be needed to establish calibration, optimal thresholds and real-world screening performance, and ongoing work aims to integrate the retinal signal with circulating biomarkers and placental pathology. But if the findings hold, the humble eye photograph, already a staple of ophthalmology and increasingly a window into diabetes and Alzheimer’s disease, may become one of the most consequential tools in prenatal medicine: a two-minute scan that flags silent vascular danger early enough for aspirin, surveillance and time to change the outcome.

Subject of Research: Retinal imaging and interpretable AI for predicting hypertensive disorders of pregnancy

Article Title: Decoding systemic vascular health and hypertensive disorders in pregnancy through retinal imaging and Visionary AI

Article References: Bearelly, S., Areff, C., Hoffman, M. K., Sunko, D., Laine, A. F., Choi, B., Coleman, H. R., Cordazzo Vargas, B., Li, C. Y., Dawkins, J. J., Amir, E., Paz, S., Hark, L. A., Booker, W. A., Jauhal, A. A., Brandt, J. S., Wapner, R. J., & Shenhav, L. (2026). Decoding systemic vascular health and hypertensive disorders in pregnancy through retinal imaging and Visionary AI. Nature Biotechnology. https://doi.org/10.1038/s41587-026-03303-0

Image Credits: AI Generated

DOI: 10.1038/s41587-026-03303-0

Keywords: preeclampsia, hypertensive disorders of pregnancy, retinal imaging, artificial intelligence, Visionary AI, vascular topology, predictive biomarkers, maternal health, machine learning, prenatal screening, placental dysfunction, Nature Biotechnology

Bulimia nervosa is a serious psychiatric condition characterized by recurrent episodes of binge eating followed by compensatory behaviors such as purging, along with intense preoccupation with body shape and weight. The disorder carries substantial medical and psychological burden, and while enhanced cognitive behavioral therapy, known as CBT-E, remains the first-line intervention, remission rates remain stubbornly modest. This treatment gap has driven researchers to ask a fundamental question: what changes in the brain when therapy succeeds, and can those changes, or the brain’s starting configuration, tell us who is likely to respond?

To investigate, researchers affiliated with the Eating Disorder Neuroimaging Initiative recruited thirty Japanese women with bulimia nervosa, with an average age of 33.6 years. The participants received one of three CBT-based treatments delivered across multiple sites: twenty-four received enhanced CBT in person, four received CBT via videoconference, and two completed guided internet-based CBT. Each woman underwent resting-state functional magnetic resonance imaging and completed symptom assessments, including the Eating Disorder Examination Questionnaire, both before and after treatment, with an average interval of 5.6 months between scans.

Resting-state functional connectivity, the technique at the heart of the study, measures how spontaneously synchronized activity is between brain regions when a person is simply lying still in the scanner. Rather than probing task-driven brain responses, it captures the intrinsic wiring diagram of the brain, offering a window into the circuit organization that persists across mental states. The researchers computed connectivity between pairs of predefined regions of interest across the whole brain and used statistical harmonization procedures to remove site-specific differences introduced by the multi-center design, a critical step for ensuring that scanner effects did not masquerade as clinical findings.

The analysis then asked two complementary questions. First, did changes in connectivity between the pre- and post-treatment scans track the degree of symptom improvement measured by the questionnaire? Second, did the strength of specific connections before treatment predict how much a patient would improve? The team used multiple regression models that adjusted for age and pretreatment psychotropic medication, and they applied false discovery rate correction at two levels: a seed-level correction across all connections emanating from each region, and a stricter analysis-level correction across every tested connection in the study.

Under the most conservative statistical threshold, no individual connection survived correction, a sobering but honest result that the authors emphasize throughout. However, at the exploratory seed level, a coherent pattern emerged. Greater improvement in eating disorder symptoms was associated with decreased connectivity between the right orbitofrontal cortex, a region central to reward valuation and decision-making, and the right posterior supramarginal gyrus, which participates in integrating sensory and bodily information. Improvement was also linked to increased connectivity between the right anterior temporal fusiform cortex and cerebellar regions, and between the right caudate nucleus, a key node of the basal ganglia reward system, and the right cerebellar hemisphere IX.

These circuits are not random. The orbitofrontal cortex and caudate nucleus have long been implicated in the aberrant reward processing thought to drive binge eating and purging, while the cerebellum is increasingly recognized as far more than a motor structure, contributing to cognitive and affective regulation. The finding that symptom improvement accompanied a loosening of orbitofrontal connections and a strengthening of striatal-cerebellar coupling suggests that successful therapy may reorganize how reward signals are computed and propagated through the brain, potentially dampening the pathological valuation of food and body shape that sustains the disorder.

Equally intriguing were the pretreatment predictors. Women who improved more showed stronger baseline connectivity between the right superior parietal lobule and inferior frontal and temporo-occipital regions, as well as between the left postcentral gyrus and bilateral inferior temporal gyri. The postcentral gyrus houses the primary somatosensory cortex, which represents the body’s physical form, while the superior parietal lobule supports spatial attention and cognitive control. Stronger connectivity in these networks before therapy may reflect a neural readiness, a capacity for body-awareness integration and self-regulation that allows patients to engage more fully with the cognitive restructuring work that CBT demands.

The researchers are careful to frame these results as preliminary. The sample of thirty participants is small, there was no healthy control group, and the findings did not survive the strictest multiple-comparison correction. Resting-state connectivity measures are also sensitive to head motion and physiological noise, and the mixed treatment modalities, while all CBT-based, introduce heterogeneity. Replication in larger, controlled cohorts is essential before these connectivity signatures could inform clinical practice. Still, the study’s multi-site harmonized design and longitudinal structure represent methodological advances over the cross-sectional snapshots that have dominated the field.

The broader significance lies in what this line of research could eventually deliver: biomarkers that identify patients at risk of treatment failure before therapy begins, and mechanistic targets for interventions that directly modulate the circuits involved. Brain stimulation techniques, neurofeedback, and digitally delivered therapies could all be refined by a clearer map of the neural substrate of recovery. For now, the study offers a cautious but compelling glimpse of the brain rewiring itself as patients heal, suggesting that the path out of bulimia nervosa may be written, in part, in the changing conversation between the reward system, the body’s sensory maps, and the mind’s capacity for control.

Subject of Research: Resting-state functional connectivity changes and pretreatment predictors of cognitive behavioral therapy response in bulimia nervosa

Article Title: Resting-state functional connectivity changes and pretreatment features associated with symptom improvement after cognitive behavioral therapy for bulimia nervosa

Article References: Sudo, Y., Chhatkuli, R. B., Takamura, T., Kamashita, R., Matsumoto, K., Numata, N., Hamatani, S., Moriguchi, Y., Ino, K., Ando, T., Ohara, C., Funaba, M., Isobe, M., Noda, T., Tose, K., Murai, T., Sunada, M., Morimoto, K., Taniguchi, M., … Futatsuya, K. (2026). Resting-state functional connectivity changes and pretreatment features associated with symptom improvement after cognitive behavioral therapy for bulimia nervosa. Journal of Eating Disorders. https://doi.org/10.1186/s40337-026-01754-z

Image Credits: AI Generated

DOI: 10.1186/s40337-026-01754-z

Keywords: bulimia nervosa, cognitive behavioral therapy, resting-state functional connectivity, fMRI, reward processing, somatosensory cortex, orbitofrontal cortex, cerebellum, treatment response, eating disorders, neuroimaging, predictive biomarkers

The study, published in BMC Medical Imaging, was a retrospective dual-center investigation led by Yi Dai of Peking University Shenzhen Hospital together with colleagues at Guangdong Provincial People’s Hospital. The researchers enrolled 156 patients who had undergone pretreatment dynamic contrast-enhanced MRI, the standard contrast-enhanced breast imaging technique, and who had also received MammaPrint testing on their surgical or biopsy specimens. Their goal was ambitious: to see whether the subtle texture patterns hidden inside routine clinical scans could stand in for a molecular assay that requires actual tumor tissue.

The approach rested on two complementary computational techniques. The first is radiomics, a method that converts medical images into hundreds of quantitative features describing shape, intensity, and texture characteristics that the human eye cannot reliably perceive. The second is habitat analysis, a technique that acknowledges a fundamental biological truth: tumors are not uniform. Instead of treating a tumor as a single lump of tissue, habitat imaging partitions it into distinct subregions, or habitats, based on how different areas take up contrast agent on MRI. These subregions are thought to reflect genuine biological differences, such as variations in cell density, blood supply, and oxygen levels, that drive aggressive behavior.

Crucially, the team did not stop at the tumor’s visible border. They extracted radiomic features from both the intratumoral region, the tumor itself, and the peritumoral region, the surrounding tissue that radiologists typically discard. This decision reflects growing evidence that the tumor microenvironment, including the stromal reaction and immune infiltration in adjacent tissue, carries information about a cancer’s biology. By combining features from inside and outside the tumor with habitat-based heterogeneity measures, the researchers built what they call a combined radiomics-habitat model.

Feature selection followed a rigorous pipeline. Radiomic features were standardized according to the Image Biomarker Standardisation Initiative, an international effort to make radiomic features reproducible across centers and scanners. The team then applied the least absolute shrinkage and selection operator, known as LASSO, together with minimum redundancy maximum relevance filtering to whittle down the initial feature set to the most informative and least redundant variables. Multiple machine learning classifiers were then trained to distinguish MammaPrint low-risk from high-risk tumors, and their performance was evaluated with receiver operating characteristic analysis, calibration plots, and decision curve analysis.

The ExtraTrees classifier, an ensemble method that builds many randomized decision trees and averages their votes, emerged as the strongest performer. The combined model achieved an area under the receiver operating characteristic curve of 0.931 in the training set, 0.896 in the internal validation set, and 0.877 in the external test set drawn from the second hospital. An AUC of 0.877 on unseen external data is particularly noteworthy, because it demonstrates that the model’s performance was not an artifact of overfitting to a single institution’s scanner or patient population.

Calibration, which measures whether the model’s predicted probabilities match actual observed risk, was also strong. Hosmer-Lemeshow tests yielded P values of 0.267, 0.491, and 0.892 across the training, internal validation, and external test sets respectively, indicating no statistically significant departure from ideal calibration. Decision curve analysis, a technique that quantifies the net clinical benefit of acting on a model’s predictions across a range of risk thresholds, confirmed that using the model would add value compared with treating all patients the same way or treating none.

The analysis also identified two independent clinical predictors. Mass margin, the boundary characteristic of the tumor on imaging, carried an odds ratio of 0.854, while tumor size carried an odds ratio of 1.014, both statistically significant. In plain terms, tumors with more irregular margins and larger dimensions were more likely to fall into the high-risk genomic category, which aligns with established understanding that infiltrative growth patterns and tumor burden correlate with aggressive biology.

The implications extend beyond a single genomic assay. MammaPrint’s clinical utility was established in the landmark MINDACT trial, and it is incorporated into guidelines from the National Comprehensive Cancer Network and other bodies for guiding chemotherapy decisions in node-negative and select node-positive breast cancer. Yet genomic testing requires adequate tumor tissue, adds cost, and introduces turnaround delays. An imaging surrogate that could be computed from MRI scans already acquired during diagnostic workup would allow risk stratification to begin earlier, potentially informing surgical planning and neoadjuvant treatment discussions before pathology and genomics are finalized.

The authors caution that this remains a retrospective study in hormone receptor-positive, HER2-negative disease, the population in which MammaPrint is most commonly used, and that prospective validation in larger and more diverse cohorts will be needed before clinical deployment. Still, the convergence of habitat imaging, standardized radiomics, and interpretable machine learning represents a tangible step toward what the researchers describe as a noninvasive imaging surrogate for genomic risk stratification. If future trials confirm these results, the MRI scan a patient already receives could quietly become the first word in deciding whether chemotherapy is worth its toll.

Subject of Research: Noninvasive prediction of MammaPrint genomic risk in breast cancer using MRI-based radiomics and habitat analysis

Article Title: Prediction of mammaprint risk in breast cancer using a combined radiomics-habitat model integrating intratumoral and peritumoral MRI features: a retrospective dual-center study

Article References: Dai, Y., Lin, J., Chen, K., Lian, C., Aishanjiang, D., Li, G., Chen, M., Lu, Y., Fang, Z., Cui, J., & Wu, L. (2026). Prediction of mammaprint risk in breast cancer using a combined radiomics-habitat model integrating intratumoral and peritumoral MRI features: a retrospective dual-center study. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02862-7

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02862-7

Keywords: breast cancer, MammaPrint, radiomics, habitat imaging, MRI, machine learning, tumor heterogeneity, ExtraTrees, peritumoral features, genomic risk stratification, DCE-MRI, predictive biomarkers