Computers in Biology and Medicine

Background: Transthoracic echocardiography (TTE) remains the gold standard for assessing left ventricular (LV) diastolic dysfunction (DD), but noninvasive alternatives are needed to overcome geometric and acoustic limitations. 4D-Flow MRI enables comprehensive assessment of intraventricular pressure differences (IVPDs) and flow propagation velocity (VP). We aimed to evaluate 4D-Flow-derived IVPDs and VP in patients with hypertrophic cardiomyopathy (HCM). Methods: Thirty-eight HCM patients and 38 age- and sex-matched healthy controls underwent TTE and 4D-Flow MRI. Spatiotemporal maps of velocity and IVPDs, decomposed into inertial and convective components, were derived along the main inflow streamline from the mitral valve to the apex. Peak early filling rate (PEFR) was used to normalize IVPDs and VP. Results: While absolute peak flow rates were comparable, HCM patients exhibited significantly faster VP (100.4 vs. 60.2 cm/s, p < 0.01) and higher positive early E IVPD (1.40 vs. 0.92 mmHg, p < 0.01) compared to controls. Normalizing IVPDs by PEFR amplified group differences, revealing a markedly increased flow-normalized E-IVPD (5.13 vs. 2.94 x 103 mmHg.ml1.s, p < 0.001) and deeper reverse ER IVPD (-6.06 vs. -4.61 x 103 mmHg.ml1.s, p < 0.01) in HCM. These differences were primarily driven by the inertial components. Multivariable analysis demonstrated that ER IVPD is independently determined by PEFR and the TTE E/E' ratio (r = 0.65, p < 0.001). Conclusions: 4D-Flow MRI-derived, flow-normalized IVPDs and propagation velocity provide robust, noninvasive markers of altered LV chamber stiffness and impaired diastolic filling in HCM, offering deeper 3D hemodynamic insights into diastolic dysfunction.

Patients with implantable cardioverter-defibrillators (ICDs) often are excluded from receiving CMR scans at 3.0T due to ICD-induced image artifacts. This work aims to transfer sequence protocols established for CMR imaging in the presence of ICDs at 1.5T to a first human application at 3.0T. All measurements were performed on a clinical 3T MRI scanner. Phantom measurements were used to establish and optimize wideband PSIR LGE imaging at 3.0T. Volunteer measurements with externally placed ICDs were used for optimization of CINE imaging and wideband PSIR imaging. Thirdly, patients with implanted ICD were examined with the optimized protocols. Human measurements showed that SA stacks obtained with spoiled GRE after contrast agent showed a markedly lower artifact burden than the conventional approach. For scar imaging, a wideband PSIR with a bandwidth of 3kHz and a frequency offset of +750Hz allowed a sufficient suppression of relevant artifacts in the majority of scans. No strong correlation between the image quality and the distance between the cardiac silhouette and the device was detected. The presented work shows the feasibility of clinical cardiac MRI at 3.0T including GRE-CINE and wideband PSIR-LGE for patients with ICDs.

Background: Self-supervised learning may improve segmentation when labelled data are limited, but its benefit may depend on model and initialisation strategy. This study evaluated in-domain masked-reconstruction pretraining across limited annotation budgets. Methods: The International Skin Imaging Collaboration (ISIC) 2016 development set was partitioned into 765 images for self-supervised pretraining, 94 for fine-tuning and 41 for validation. Five nested annotation budgets (12-94 images) and five seeds were evaluated on the 379-image test set. A custom U-Net compared random and in-domain self-supervised initialisation; a ResNet-34 U-Net compared random, ImageNet, in-domain self-supervised and sequential ImageNet-to-self-supervised initialisation. Models trained with 94 labelled images were evaluated without adaptation on ISIC 2017. A separate experiment examined restoration-based pretraining. Results: In the custom U-Net, self-supervised pretraining achieved higher mean Dice than random initialisation at all five budgets, including 0.664 versus 0.615 with 12 labels and 0.811 versus 0.768 with 94 labels. Seed-level 95% confidence intervals for paired differences excluded zero at 47 and 94 labels but not at the other budgets. With 47 labels, the self-supervised model achieved mean Dice of 0.789, compared with 0.768 for random initialisation with 94 labels. In ResNet-34, ImageNet initialisation achieved higher mean Dice than in-domain self-supervised pretraining at all five budgets, and adding self-supervised pretraining after ImageNet reduced mean Dice at four of five budgets. All six 94-label configurations achieved lower mean Dice on ISIC 2017 than ISIC 2016. In the restoration experiment, mean Dice was 0.686 with random initialisation and 0.638-0.642 following pretraining. Conclusions: In-domain masked-reconstruction pretraining improved mean Dice relative to random initialisation in the custom U-Net, but not consistently in the ResNet-34 U-Net, where ImageNet initialisation performed better at every budget. Adding self-supervised pretraining after ImageNet provided no consistent gain. These findings are specific to the dermoscopic lesion-segmentation experiments and do not establish melanoma-specific diagnostic or clinical performance.

The -Bonacci weighted sequence space {ell}(,) is introduced for [&ge;] 2, 1 [&le;] < {infty}, and > 0). It is proved to be a Banach space which is isometrically isomorphic to {ell}, with an unconditional Schauder basis. The space is also reflexive and uniformly convex. Its strict inclusion chain, continuous dual, and asymptotic truncation estimate of order - / are also established. Based on this framework, Scaled &#119898-BonacciNet is proposed as a lightweight CNN that rescales Fibonacci-Net channels using the dominant &#119898-Bonacci root and employs depth wise separable convolutions. Tribonacci-Net achieved 96.09% accuracy and a 0.9581 macro F1-score on the 15-class Brain MRI dataset using 0.85 million parameters and 0.366 GFLOPs. Pentanacci-Net attained 95.87% accuracy on the 44-class dataset while reducing parameters and FLOPs by 47.10% and 44.48%, respectively. On the CKD ultrasound dataset, Hexanacci-Net achieved 83.75% accuracy and a 0.7941 macro F1-score, with corresponding reductions of 50.37% and 45.10%. These findings support dominant-root scaling as a structured approach to lightweight CNN design.

Biomedical artificial intelligence (AI) increasingly relies on heterogeneous data sources, yet differences in data structure, terminology, evidence quality, and provenance complicate their integration into reliable knowledge resources. Here, we develop an evidence-aware framework for processing and integrating heterogeneous pharmacovigilance data from FDA Adverse Event Reporting System (FAERS) reports, biomedical literature, clinical trials, and electronic health records in Mass General Brigham. The framework combines large language model (LLM)-based information extraction, ontology-driven semantic normalization, cross-source evidence assessment, statistical and machine learning-based prioritization, and ingredient-based adverse drug event (ADE) relationship inference to transform heterogeneous evidence into a structured and evidence-based knowledge resource. The constructed knowledge system contains 11,588 drug nodes, 44,843 ingredient nodes, 3,275 ADE nodes, 993 indication nodes, and more than 2.47 million relationships, with each drug-ADE relationship assigned a confidence score. We further developed a drug-level web interface for accessing the knowledge resource and validated its utility in downstream AI tasks. Specifically, incorporating the developed knowledge resource improved the comprehensiveness, relevance, and overall quality of LLM-based drug-specific ADE synthesis and improved the precision and F1 score of clinical ADE relation extraction. These results demonstrate that AI-assisted processing combined with evidence-aware curation can transform heterogeneous biomedical evidence into reusable, AI-ready knowledge with explicit provenance and evidence strength information. The framework provides a general strategy for integrating heterogeneous biomedical data to support reproducible knowledge construction and downstream biomedical AI applications.

Background: Conventional performance metrics do not indicate whether artificial intelligence (AI)-platforms and radiologists are right or wrong on the same examinations, nor whether their non-agreement is more often beneficial or harmful. Objective: To develop, illustrate, and apply directional metrics quantifying non-agreement between radiologists and AI-platforms against a common reference standard, and to adapt a fragility-based framework for quantifying the statistical robustness of these metrics. Methods: We defined Positive Non-Agreement (PNA, AI-platform correct, radiologist incorrect), Negative Non-Agreement (NNA, radiologist correct, AI-platform incorrect), and Net Non-Agreement (NENA, their difference). Metrics were illustrated in three hypothetical scenarios of 1,000 examinations each and applied to three previously published AI-radiologist studies on pulmonary embolism and mammography. Bootstrap confidence intervals were calculated, and a reclassification fragility index (FI) quantified the minimum case reclassification needed to remove statistical support for NENA's direction. Results: In illustration, NENA was positive in two scenarios and negative in one. In the three published studies, NENA ranged from -4.18% to +10.74% and decreased consistently with increasing radiologist experience (junior residents, +6.27%; consultants, +1.45%), a gradient driven primarily by declining PNA with modest increase in NNA. The reclassification FI was defined for all three illustrative scenarios (range, 0.10%-1.70%) and for six of the seven comparisons from these studies (range, 0.34%-4.31%), but not for the weakest comparison, whose confidence interval for NENA included zero. Conclusions: PNA, NNA, and NENA provide a directional, case-level complement to conventional performance metrics for AI-radiologist non-agreement. The accompanying reclassification FI offers a statistically grounded, generalizable method for quantifying their robustness.

Computed tomography (CT) interpretation requires recognition of the anatomical level of each slice. Classification may facilitate preprocessing and AI-assisted fracture diagnosis. We developed deep learning models to classify craniofacial CT slices into six anatomical categories. A total of 11,769 images were collected from a multicenter database involving 10 Japanese university hospitals and the public Head-Neck-Radiomics-HN1 dataset. Patients were divided into training (70%) and evaluation (30%) sets. Five CNN architectures were trained from scratch, and hard- and soft-voting ensembles were evaluated. Individual-model accuracy ranged from 94.72% to 96.09%; the hard-voting ensemble achieved 96.46% accuracy (Cohen's k = 0.954). Although performance differed among individual models (Cochran's Q = 99.29, p < 0.001), the ensemble did not significantly outperform EfficientNetV2B1 (McNemar's p = 0.163). Errors occurred mainly between adjacent anatomical regions. Deep learning can classify craniofacial CT slices accurately and may support automated dataset organization and AI-assisted fracture diagnosis. External validation is needed.

Purpose: To evaluate whether visually assessed high-quality prostate MRI, as defined by the Prostate Imaging Quality (PI-QUAL) scoring system, is sufficient to ensure consistent AI model performance, or whether hidden acquisition-related technical effects introduce variability that cannot be detected through visual inspection. Methods: This retrospective study included 568 bpMRI scans from seven institutions, classified using PI-QUAL v2 into high-quality (scores 2-3) and low-quality (score 1) groups. Quantitative image quality features were extracted using MRQy from T2w and ADC sequences. Three experiments were conducted on high-quality scans: (1) unsupervised clustering to identify hidden technical batch effects; (2) V-Net models (M) for prostate capsule segmentation on T2w using single-cluster and balanced multi-cluster training strategies; and (3) VoxelMorph-based deformable registration (M2) to align ADC maps to T2w images. Performance metrics included DSC, precision, recall, volume ratio, HD95, ASSD, and TRE. Kruskal-Wallis tests assessed cross-cluster variability, Two One-Sided Tests (TOST) evaluated performance equivalence, and bootstrap resampling quantified training efficiency. Results: Clustering identified four clusters among 277 high-quality bpMRI scans. Single-cluster models showed variable performance (DSC=0.41-0.85) with significant batch effects (p<0.001), even with 140 training samples. The balanced multi-cluster model (n=60) eliminated batch effects (p=0.075), achieved equivalent performance (p<0.001), and demonstrated 2.23-fold greater training efficiency despite using 2.3-fold fewer samples. Registration performance also varied significantly across clusters (TRE=0.86-5.47 mm, p<0.001). Conclusion: Acquisition-related batch effects significantly impact AI generalizability in prostate MRI, even among high-quality scans. Quantitative clustering enables detection of technical variations and supports diversity-aware training strategies that achieve equivalent performance with 2.3-fold fewer samples while eliminating batch effects.

Large language models (LLMs) are increasingly evaluated for clinical tasks, but safe use requires reliable uncertainty estimates as well as high accuracy. We tested eight general-purpose LLMs on 100 validated neurology board-style multiple-choice questions and compared three confidence signals: self-reported confidence, critique-implied confidence from an independent model instance, and consistency-based confidence from repeated sampling. Accuracy varied from 45% (Llama 3.1 8B) to 92% (Gemini 3 Flash/Pro), but confidence was generally overestimated, particularly in lower-accuracy models. Consistency-based confidence provided the best discrimination between correct and incorrect answers (pooled AUROC 0.74), outperforming self-report, although calibration differences between methods were not significant. Calibration should therefore be reported alongside accuracy when clinical LLMs are evaluated.

Background: Artificial intelligence (AI) may improve electrocardiographic differentiation of Takotsubo syndrome from acute anterior ST-segment elevation myocardial infarction (STEMI), but its impact on human diagnostic performance, the added value of visual explanations, and differences by reader background and experience remain uncertain. Methods: Using 12-lead electrocardiogram (ECG) images from two hospitals, we developed and externally validated an ensemble of three deep learning models after self-supervised pretraining and fine-tuning. The diagnostic task used ECGs recorded within 14 days after symptom onset, regardless of reperfusion status. We conducted a four-phase reader study. Phases 1-3 compared unaided interpretation, a reference sheet, and additional AI assistance. In Phase 4, readers evaluated a new ECG set after randomization to reference-sheet support alone, with AI predictions, or with AI predictions plus gradient-weighted class activation mapping (Grad-CAM) heatmaps. Diagnostic correctness was analyzed using generalized linear mixed-effects models. Exploratory analyses examined differences by cardiologist status and experience. Results: Development and external-validation cohorts, including healthy controls, comprised 832 and 95 participants, respectively. For Takotsubo-STEMI discrimination, external-validation accuracy was 0.914, with areas under the receiver operating characteristic and precision-recall curves of 0.970 and 0.984. Seventy healthcare providers completed the reader study. The estimated probability of correct diagnosis increased from 0.632 without assistance to 0.864 with AI assistance (odds ratio [OR], 3.71; P<0.001). In randomized Phase 4, AI assistance improved accuracy vs reference-sheet support alone (OR, 4.28; 95% CI, 3.08-5.95; P<0.001), and Grad-CAM provided additional benefit (OR, 1.53; 95% CI, 1.04-2.25; P=0.033). Exploratory analyses showed greater improvement among less experienced readers; with AI assistance, non-cardiologists outperformed cardiologists. AI-assisted readers remained less accurate overall than the model alone. Conclusions: An ECG image-based ensemble differentiated Takotsubo syndrome from anterior STEMI and improved reader accuracy. Visual explanations provided additional benefit, while experience-related differences highlight the importance of effective human-AI interaction.

Aims: Myocardial strain enables sensitive assessment of cardiac function, but its clinical application has largely focused on individual chambers or selected patient populations. We evaluated the feasibility and clinical relevance of fully automated multi-chamber strain assessment in consecutive patients undergoing routine transthoracic echocardiography (TTE). Methods and results: This retrospective study included 8,350 consecutive adults undergoing TTE. A hospital-wide artificial intelligence (AI)-based system automatically measured left ventricular (LV) global longitudinal strain (LVGLS), left atrial (LA) reservoir strain (LASr), and right ventricular (RV) free-wall strain (RVFWS). AI-derived LVGLS was available from at least one apical view in 97.1% of patients and from all three views in 94.8%; LASr and RVFWS were available in 94.1% and 94.5%, respectively, resulting in complete multi-chamber strain assessment in 90.3%. Strain measures showed modest correlations with conventional chamber-specific echocardiographic parameters and differential associations with clinical outcomes. Gaussian mixture modeling identified three distinct clusters characterized by preserved multi-chamber function, RV-dominant dysfunction, and pan-chamber dysfunction. Compared with the preserved cluster, pan-chamber dysfunction was associated with the highest risk of all-cause death or heart failure hospitalization (adjusted hazard ratio, 3.44; 95% CI, 2.43-4.88; P<0.001). Prognostic associations remained evident in patients with LV ejection fraction (LVEF) [&ge;]50%, and cluster assignment provided incremental prognostic information beyond clinical factors, LVEF, and RV fractional area change. Conclusions: Fully automated multi-chamber strain assessment was feasible in a large consecutive TTE cohort. Integrating strain across cardiac chambers may provide complementary functional characterization and identify clinically relevant clusters with prognostic information beyond conventional assessment.

Introduction: Lichen sclerosus (LS) is a chronic inflammatory dermatosis predominantly affecting the anogenital region of both women and men. Sexual dysfunction remains among the most debilitating and overlooked consequences of the disease, as available patient-reported outcome measures are either restricted to one sex or capture sexual health only partially. The aim of the present work is to validate the Qualisex questionnaire (in its the Italian version) in patients with LS and to evaluate its psychometric properties, factor structure, convergent, discriminant validity, and sex symmetry. Methods: A cross-sectional study was conducted on 211 patients (167 women, 44 men; mean age 52.9 years) with a confirmed diagnosis of LS. Participants completed the Italian Qualisex, Skindex-29, DASS-21, and sexual-health-specific scales, the Female Sexual Function Index (FSFI) for women, and the International Index of Erectile Function (IIEF) for men. Construct validity was assessed via Principal Component Analysis (PCA), and reliability via Cronbach's alpha. Convergent and discriminant validities were evaluated against standard psychometric batteries. Results: Item 2 was removed due to low communality (0.167). PCA on the remaining 9 items revealed a two-factor structure accounting for 72.5% of total variance: Component 1 (Sexual Functioning and Satisfaction, 59.89%) and Component 2 (Partner Devaluation, 12.66%). Internal consistency was high (alpha= 0.911 overall; alpha= 0.841 in women; alpha= 0.899 in men). Qualisex scores correlated strongly with Skindex-29 Social Functioning (r = 0.582) and Emotions (r = 0.531) subscales, as well as FSFI (r = -0.391) and IIEF (alpha= -0.566). Discriminant validity was supported by low-to-moderate correlations with DASS-21 domains (r < 0.38). While FSFI and IIEF showed significant negative correlations with age (p = 0.002), Qualisex remained age-independent across sexes (p > 0.05). Conclusion. The 9-item Italian Qualisex is a reliable, sex-symmetrical, and age-independent patient-reported outcome measure for assessing sexual health impairment in LS, successfully isolating disease-specific distress from physiological aging and general distress.

Neurologic autoantibody cell-based assay images are difficult to interpret when fluorescence is weak, focal or dominated by background signal. We developed a visual artificial intelligence overlay that transfers detected cell boundaries from the structural channel to the antibody-staining image and evaluated standalone model concordance and its association with reader classifications. The analysis used 2 763 records for model development and 499 held-out records spanning 14 antibody targets. Four readers assessed the held-out records first without assistance and then with the boundary overlay and model category available. The selected model agreed with the reader-derived three-class reference in 85.2% of records, but recall was lowest for weak-positive or equivocal records. Reader agreement increased in the assisted session, and almost all revised ratings moved towards the model category. The combined display may support consistent visual review while also creating a risk of model-directed anchoring; its role as a second-read aid requires independent validation.

Flow cytometry is central to clinical hematology, generating extensive archives comprising billions of single-cell profiles linked to expert-derived diagnostic summaries. While these records represent a largely untapped repository of collective expert knowledge, the underlying single-cell annotations required for supervised machine learning are typically not recorded in a standardized, accessible format. We present IMPGAT (implicit gating), a weakly supervised multiple-instance learning framework that reconstructs single-cell gating strategies directly from existing sample-level cell-type frequency distributions. The model utilizes an induced set attention mechanism to map single-cell marker profiles to specific cell populations, requiring neither explicit cell-level labels nor unsupervised clustering during training. We evaluate the framework on acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) cohorts, totaling 711 measurements from 268 individuals and 240 million single cells. In ALL, the model's quantification of cell-type abundances shows no statistically significant difference from human inter-expert variability. In AML, the framework achieves blast quantification on par with a recently established supervised state-of-the-art model that relies on explicit single-cell annotations for training. Error analysis shows that mismatches between model predictions and expert gating primarily involve expert-labeled non-cellular events, reflecting the non-incorporation of any explicit physical prior knowledge (such as the geometric relationships traditionally used to identify doublets and debris in scatter cytograms) in model training. Conversely, model certainty correlates with phenotypic borderline cases where manual bisection is inherently difficult. Our approach enables the systematic utilization of retrospective clinical records for the development of automated, standardized gating logic and is applicable to broader high-throughput contexts, such as large-scale immune monitoring or phenotypic drug screening.

Objective: Case-finding models for underdiagnosed conditions are usually trained on diagnosis codes, which are missing from the records of many patients with a clinical diagnosis. We asked whether models trained on self-reported diagnoses rank uncoded self-reporters higher than models trained on later codes, and how a score fit only from released summary counts compares with a participant-level model. Methods: In the All of Us Research Program, we followed survey respondents with no qualifying diagnosis code by the survey date for later codes. For chronic fatigue, fibromyalgia, and endometriosis (participants recorded as female at birth), we fit a naive Bayes score with empirical-Bayes shrinkage, from released feature-by-label counts for the self-reported diagnosis and from exact counts inside the data enclave for a later code, and an elastic net on participant-level data for each label. In a random 20% validation split, we measured discrimination of self-report, of uncoded self-reporters (no qualifying code observed), and of later codes. Results: For chronic fatigue (149,860 respondents, including 7,770 self-reporters, 95% with no code observed), the summary-count score trained on self-report ranked uncoded self-reporters with an area under the receiver operating characteristic curve (AUROC) of 0.74 (95% CI, 0.72-0.75), against 0.67 (95% CI, 0.65-0.69) when trained on later codes. For later codes, the two scores were similar (0.69 vs 0.70). Fibromyalgia showed the same pattern (0.73 vs 0.64 and 0.74 vs 0.72). Elastic nets trained on self-report ranked uncoded self-reporters higher than the summary-count scores (0.80 vs 0.74 and 0.78 vs 0.73) but discriminated later codes less well than code-trained elastic nets. For endometriosis, the self-report score also ranked uncoded self-reporters above the code-trained score (0.61 vs 0.53), but discrimination was low and it did not discriminate later codes (0.50 vs 0.70). Conclusion: The choice of training label changes which patients a case-finding model ranks highly. For all 3 conditions, self-reported diagnoses were the more useful label for finding uncoded self-reporters. For chronic fatigue and fibromyalgia, the self-report score also discriminated later codes as well as the code-trained score. Independent clinical validation is needed before these models guide case finding.

BACKGROUND Timely cause-of-death (COD) information guides health policy, yet many deaths in low- and middle-income countries are not medically certified. In Africa, fewer than 10% of deaths are. Verbal autopsy (VA) yields a structured symptom questionnaire and a free-text narrative. Mortality surveillance increasingly records the narrative alongside the questionnaire, but the algorithms used to produce VA-based statistics were designed for structured responses only, so narratives inform physician COD assignment but do not contribute to algorithmic COD assignment. We evaluated whether supervised models could use both modalities to improve COD assignment. METHODS We analyzed 4,999 VA records for deaths at ages 12 years or older from a national sample of registered deaths in South Africa (RSA), using physician-assigned CODs as the reference standard. We trained unimodal and multimodal pipelines on the labelled records: pretrained language models (PLMs) fine-tuned on narratives alone, machine learning (ML) models on structured responses alone, and the two combined at the data, feature, and decision levels. We also prompted a locally deployed generative large language model (LLM) on the same unimodal and multimodal inputs, without parameter updates. All were compared with InSilicoVA, the operational comparator representing current automated practice. We evaluated individual-level assignment, population-level agreement, calibration, and generalization across provinces in RSA, reporting these metrics rather than ranking pipelines on any one of them. RESULTS Every supervised pipeline improved on InSilicoVA by 8.5-15.9 percentage points in top-1 accuracy and 3.5-7.1 points in cause-specific mortality fraction (CSMF) accuracy. The narrative and structured pipelines performed similarly in aggregate but differed by cause: 63.9% of deaths received identical predictions, 11.6% were correct only from the narrative pipeline, and 13.6% only from the questionnaire pipeline, with the narrative pipeline more accurate for noncommunicable causes and the questionnaire pipeline for communicable causes. Against the unimodal approaches, the decision-fusion pipelines added 4.2-5.4 percentage points in top-1 accuracy, concentrated in the most common causes, whereas feature fusion with cross-modal attention improved the less common causes, adding 6.6 points (95% CI, 1.7-11.5) in balanced accuracy and 5.5 points (95% CI, 0.6-10.3) in macro F1. Feature fusion reached the highest CSMF accuracy (95.7%). Paired differences from the better unimodal representative ranged from -0.2 to +3.4 points, with all intervals including zero. The LLM exceeded InSilicoVA on individual-level metrics and reached the highest balanced accuracy of any pipeline (60.6%), while the supervised pipelines retained higher CSMF accuracy. Supervised pipelines had lower expected calibration error than InSilicoVA. Supervised gains held across all nine held-out provinces, including every top-1 comparison and nearly every macro F1 and CSMF comparison. CONCLUSIONS PLMs using narratives alone substantially improved on current best practice and matched ML pipelines using structured responses, while multiple multimodal strategies produced further gains in individual COD assignment. The zero-shot LLM showed particular strength on rare causes, whereas supervised pipelines better recovered the population cause distribution. These findings support incorporating narratives as a core input to automated VA systems and evaluating representative fusion strategies for the intended output. External validation across settings and age groups will support broader implementation.

High attrition rates across clinical drug development impose immense clinical and financial burdens, yet prospectively forecasting whether an individual trial will achieve its primary endpoint remains challenging. Here we present a machine-learning system that predicts clinical trial success across phases I-IV, independently modeling studies with efficacy versus safety/toxicity primary endpoints. To prevent lookahead bias and outcome leakage, models are trained on registration-time features, including target and mechanism-of-action precedent, drug modality and novelty, trial design and operational characteristics, and frozen literature-derived target-choice embeddings. Each prediction yields a continuous Predictive Trial Index (PTI) between 0 and 1 reflecting success probability. On a test set of trials initiated in 2021 or later, the efficacy model achieved an area under the receiver operating characteristic curve (AUROC) of 0.828 (95% CI 0.812-0.845; n = 2,347) and the safety/toxicity model reached 0.809 (95% CI 0.787-0.831; n = 1,485). These models are embedded in inClinico Agent, an agentic AI system integrating frontier large language models (LLMs) with deterministic scoring engines, retrieval tools and specialized clinical skills. We also introduce ClinBench, a prospective benchmark of 2,040 ongoing trials evaluated by inClinico Agent and five frontier LLMs, establishing a leakage-free foundation for tracking prospective predictive accuracy as trials reach completion.

Background and Objective: The volume of patient portal messages has grown substantially and artificial intelligence (AI) can support clinical staff and improve the timeliness of patient care. Our objective is to test if an AI tool for managing the electronic inbox In-Basket reduces response times to patient portal messages in neurology. Methods: We conducted an implementation study from January to May 2026 in the neurology department at two clinical sites within a regional medical system. A single AI tool was implemented to auto-categorize and auto-route all patient messages received in the electronic inbox. The tool also generates optional draft replies to messages. Primary process measures were the median weekly time (hours) from message creation to the first action and to closure. Impact was analyzed using a controlled interrupted time-series (ITS) design, comparing message-volume-matched control divisions. Results: Among 24,142 messages analyzed, pre-intervention median weekly time to first action was 11.3 hours [IQR: 5.6, 15.1] (Site A) and 16.1 hours [IQR: 13.8, 19.0] (Site B), falling post-intervention to 1.0 hour [IQR: 0, 3.1] (Site A) and 0.6 hours [IQR: 0.2, 2.4] (Site B). Controlled interrupted time-series analysis showed a significant immediate differential level change in median weekly hours to first action at Site A (-92.5%, 95% CI: -97.5% to -78.0%), and a significant downward trend at Site B (-43.5%, 95% CI: -65.2% to -8.2%). Changes in time to closure were directionally favorable but not statistically significant at either site. There was low use of the generated draft replies, only 0.5% at Site A and 3% at Site B. Discussion: Implementation of AI-assisted message categorization and routing was associated with a substantially reduced initial action time for patient In-Basket messages in Epilepsy. The utilization of AI-generated draft-replies to messages was low. These findings support the feasibility and effectiveness of AI-augmented workflows to improve inbox management.

Background: Conventional discrimination metrics, such as the area under the receiver operating characteristic curve (AUROC), do not characterize the composition of outcomes among patients flagged as high risk by clinical prediction models. This limitation is particularly relevant in critical care, where competing deaths and incomplete follow-up can complicate the interpretation of prediction alerts. We developed the Balanced Risk Integrity Index (BRII), a stratum-level metric defined as the proportion of target events among target events, competing deaths, and early-censored observations in a flagged high-risk stratum. BRII is algebraically related to positive predictive value (PPV) and is reported alongside its component decomposition to characterize the outcome composition of flagged patients. Methods: We evaluated BRII using seven acute kidney injury (AKI) prediction models developed in MIMIC-IV and externally validated in MIMIC-III and eICU-CRD (116,010 patients in total). Additional simulations examined the effects of competing deaths and early censoring on BRII and conventional model performance metrics. Results: For 48-hour AKI prediction in the top 10% risk stratum, BRII ranged from 0.90 to 0.96 in MIMIC-III, compared with less than 0.26 across all models in eICU-CRD, where 61.5% of evaluated patients were censored before the prediction horizon. In the eICU-CRD top 10% strata for the same target, 37.0% to 50.2% of patients were censored early, 12.2% to 15.5% had AKI, and 3.4% to 8.1% died before AKI. BRII and PPV ordered at least one model pair differently in 34 of 48 cohort-outcome-top-k groups. Under simulated random censoring for 48-hour AKI prediction, BRII declined by up to 0.80 while binary-label AUROC changed by less than 0.02. Conclusions: BRII provided complementary information on high-risk stratum composition, yielding different model rankings from PPV and revealing substantial differences across external validation cohorts. Interpreted alongside PPV and its component decomposition, BRII offers a standardized approach to evaluating clinical prediction alerts across models and observation environments.

Background. Real-world 12-lead electrocardiograms (ECGs) often have missing or corrupted leads. A deep-learning classifier can be prepared for this by random lead masking during training, or by injecting clinical knowledge, for example by pulling its attention towards the leads that guidelines use for each diagnosis. Whether this knowledge adds robustness beyond masking is unknown. Methods. Pre-registered comparison of three arms sharing one network (per-lead 1D-ResNet encoder and per-diagnosis attention over the 12 leads): standard training (M0); random whole-lead masking (M1); masking plus a Kullback-Leibler penalty pulling attention towards a fixed guideline-derived key-lead map (M2, guideline-based attention guidance). Six diagnoses; five seeds per arm. Development on PTB-XL (21,799 ECGs); single sealed evaluation on PTB-XL fold 10 (2,198 ECGs) and on CODE-15% (233,204 exams, external validation). Primary endpoint: AUROC difference M2 - M1, averaged over first-degree AV block, right and left bundle branch block and atrial fibrillation and over 1 to 6 lost (zeroed) leads, with a patient-level paired BCa bootstrap 95% confidence interval. Results. The guidance did not change robustness: {Delta}AUROC -0.0010 (95% CI -0.0033 to +0.0006), an interval entirely within the pre-specified +-0.02 margin; the result was unchanged with another bootstrap seed, other interval methods and leave-one-diagnosis-out analyses. Masking improved AUROC: M1 - M0 = +0.0134 (95% CI +0.0102 to +0.0172) when one key lead was lost, and the number of lost leads tolerated before AUROC fell by more than 0.02 rose from 7 (M0) to 10 (M1). M2 attention matched the map and M2 was non-inferior on intact ECGs, but its calibration slope was lower for all six diagnoses in both datasets. M2 lost slightly more when all key leads were lost together (-0.0067) and gained only when one to three leads remained (exploratory; AUROC 0.9408 vs 0.9210 with 11 leads lost). CODE-15% reproduced every pattern (primary -0.0007, 95% CI -0.0011 to -0.0003). Conclusions. Random lead masking suffices to make an ECG classifier robust to missing leads; adding guideline knowledge aligns attention with clinical criteria but adds no robustness and worsens calibration. Guideline-aligned attention should not be taken as evidence of robustness.