IEEE Transactions on Biomedical Engineering

Minimally invasive endovascular procedures offer reduced surgical trauma, shorter recovery times, and improved outcomes, but rely on 2D fluoroscopic X-ray imaging, which provides limited depth perception and exposes patients and clinicians to ionizing radiation. Here we present an augmented reality (AR) system that fuses intravascular ultrasound (IVUS) and electromagnetic (EM) position tracking with preoperative computed tomography (CT) to produce an anatomically accurate, deformation-corrected navigational reference. A robotic device performs ECG-gated pullback of the IVUS probe, capturing 4D aortic motion across the cardiac cycle. We introduce a deep learning architecture for extracting vascular lumen boundaries and side-branch orifices from artifact-prone IVUS streams, and a semantically driven non-rigid CT-IVUS fusion pipeline robust to false positive landmarks. We evaluate the platform with trained surgeons in benchtop phantom studies and in-vivo ovine models, and demonstrate its application to fenestrated endovascular aneurysm repair (FEVAR). Compared to fluoroscopy alone, AR guidance significantly reduces cannulation time, radiation exposure, and cognitive workload, while improving procedural efficiency and safety. Our IVUS-EM and CT aortic datasets are released open source.

Distal radius fractures (DRFs) can delay return to activities of daily living and social participation because of postoperative pain, temporary joint immobilization, and limited wrist and forearm range of motion. The Ghost System developed at Saitama Prefectural University, Japan, combines visual action observation with tendon vibration stimulation and has shown potential as an adjunct to conventional rehabilitation. This Study Protocol describes a modified Ghost system intended to improve clinical implementation by replacing the head-mounted virtual reality display with iPad-based action observation and by using a wristband-type vibrator. This single-center, single-arm, open-label feasibility trial will enroll 10 adults after palmar locking plate fixation for DRF. The intervention will be delivered twice weekly during outpatient rehabilitation follow-up sessions from the early postoperative period (postoperative days 2-10 after enrollment) through the approved early postoperative rehabilitation period (generally up to postoperative week 8), in parallel with standard rehabilitation practices. Primary feasibility and preliminary clinical outcomes include device fit and acceptability, pain assessed using a 100-mm Visual Analog Scale, and wrist/forearm range of motion. Secondary implementation and safety outcomes include Disabilities of the Arm, Shoulder and Hand (DASH), Patient-Rated Wrist Evaluation (PRWE), Hand20 Questionnaire (HANDS-20), EuroQol 5 Dimensions 5 Levels (EQ-5D-5L), body ownership and hand-illusion questionnaires, setup time, setup errors, adherence, adverse events, and device incidents. We hypothesize that the modified Ghost system will be feasible and acceptable for early postoperative outpatient rehabilitation and will be delivered without serious device-related adverse events. Clinical outcomes will be summarized descriptively to inform a future controlled study rather than to establish efficacy.

Cognitive impairment is a prevalent symptom extending from physiological ageing to disease. It commonly manifests itself in initial memory problems, progressing and co-occurring in more severe conditions such as Mild Cognitive Impairment, Alzheimer's Disease and Major Depressive Disorder. However, current non-invasive screening assessments either lack biological information or are invasive and restricted to specialized centers with complex and cost-intensive set-ups. Here, we conducted an initial validation of mobile pupillometry with Virtual Reality (VR) under experimental conditions as a digital biomarker for cognitive impairment by testing required biomarker-specific properties. For this purpose, we first assessed its construct validity by testing healthy participants (n=43) on an n-back task in VR while pupil size was measured. Mixed effects models revealed that similar to lab-based eye-tracking systems, pupil size increased in a sensible and distinguishable fashion as a function of working memory load. Second, to test the signal's reliability, the same participants were tested on the identical set-up two to three months after their first visit. We observed that the pupil response profile was highly stable over this period. Third, for its clinical validity, we examined patients (n=89) from three different cohorts with varying degrees of cognitive impairment and compared them to healthy control participants (n=81). Mixed-effects models indicated that pupil size was reduced as a function of cognitive impairment levels at higher cognitive load and that this effect was stronger pronounced with increasing age. In conclusion, we provide initial evidence for mobile pupillometry being a sensitive, reliable and clinically valid digital biomarker for cognitive functioning and impairment, which offers desirable properties due to its quick, automatized and location-independent set-up. Keywords: digital biomarker, mobile pupillometry, Virtual Reality, cognition, , Major Depressive Disorder, Mild Cognitive Impairment, Alzheimer's Disease

As machine learning and artificial intelligence systems are increasingly used in healthcare, rigorous evaluation of their classification performance has become critical. The F1 and F{beta} scores are widely adopted metrics for assessing performance in imbalanced biomedical data. Recently, we introduced psF1, a unified statistical framework for inference and study design for single and comparative F1 and F{beta} scores under the assumption of independent classifiers. In practice, however, benchmarking two classifiers on the same dataset creates a correlated paired setting. Ignoring this intrinsic dependency leads to overestimation of the standard error and a substantial loss of statistical power. To address this, we develop psF1pair, an advanced framework for statistical inference and power analysis that explicitly accounts for correlations between classifier pairs. Extensive simulation studies demonstrate the performance of psF1pair, and its utility is further illustrated through application to a real-world imaging classification system. As expected, higher correlation between classifiers yields narrower confidence intervals and enhanced statistical power. A freely available R package is provided to facilitate implementation, supporting accurate evaluation and study design for predictive and classification models in biomedical research.

The aim of this work is to validate patient-specific EEG baseline establishment using the e-norms method as a screening and retrospective-review tool for seizure detection in pediatric epilepsy. The method was applied to 247 seizure-free EEG recordings (263.92 hours) from 10 patients in the CHB-MIT Scalp EEG Database (ages 3-18). A composite stability metric combining first-derivative dynamics, spectral entropy, variance, and line length was computed per 2-second epoch across 23 channels. Patient-specific detection thresholds were derived from each patient's seizure-free baseline using a weighted statistical procedure. Performance was validated against 72 expert-annotated seizures (2,705 epochs) across 62 seizure files, with durations spanning 6 to 264 seconds (44-fold range). The results show that detection achieved 94.4% event-level sensitivity (68 of 72 seizures; 95% CI 86.6-97.8%) and 81.5% epoch-level sensitivity (2,204 of 2,705 epochs; 95% CI 80.0-82.9%). Eight of ten patients achieved 100% event-level sensitivity with epoch-level sensitivity ranging from 58.7% to 100.0%. Two patients showed partial event-level failures (CHB-15: 17 of 20; CHB-18: 5 of 6), with the four missed events attributable to two characterizable failure modes. Patient-specific thresholds ranged from 4.06 to 4.81 (mean 4.51 +/- 0.25); threshold variation did not correlate reliably with age or sex, confirming that no universal threshold could achieve comparable performance. Detection margins ranged from 0.88 to 1.24 times. Patient-specific e-norms achieves 94.4% event-level sensitivity for pediatric EEG seizure detection without requiring labeled seizure training data, exceeding published human expert inter-rater agreement (50-76%) and recent automated approaches in adult cohorts using behind-the-ear EEG and wearable ECG. Two characterizable failure modes account for the four missed events and inform appropriate clinical use. As a high-sensitivity screening tool complementary to real-time alarm systems, the method is ready for adult validation, prospective deployment, and head-to-head benchmarking.

Background: Female chronic pelvic pain disorders (CPPDs) are highly prevalent and frequently accompanied by sleep disturbance and autonomic nervous system (ANS) dysregulation. Heart rate variability (HRV), a non-invasive index of ANS function, may provide an objective, physiological correlate of sleep health and can be monitored using wearable devices, enabling a continuous, scalable approach. Objectives: This study examined whether wearable-derived daily HRV metrics are associated with self-reported sleep disturbance in women with CPPD(s) compared with healthy controls, using epoch-level data and generalized additive models. Methods: We conducted a retrospective observational study using up to 90 days of data from a mobile health research app. Participants were 128 women with CPPD(s) and 63 demographically matched healthy controls, who completed a daily PROMIS-based 3-item sleep disturbance questionnaire and wore Fitbit devices that provided 5-minute HRV epochs. Primary predictors were high frequency (HF) and low frequency (LF) power and root mean square of successive differences (RMSSD), with group (CPPD vs control), daily pain severity, and menstrual status as covariates. We fit separate generalized additive mixed models (GAMMs) for each HRV metric with a nonlinear smooth term and an HRV x Group interaction. Results: Higher HF and RMSSD were associated with lower sleep disturbance scores, and these associations were stronger in controls than in the CPPD group (HF x group B {approx} -1.59, p < 0.00010; RMSSD x group B {approx} -0.58, p < 0.0001). LF showed a more complex pattern but also differed by group (B {approx} -0.531, p < 0.0001). HRV smooth terms were highly nonlinear, and models explained ~8-9% of deviance in sleep disturbances. Pain severity and menstrual bleeding were strongly associated with worse sleep. Conclusion: These findings indicate small but consistent associations between wearable-derived HRV metrics and daily sleep disturbances in women with CPPD(s) and healthy controls, with weaker associations in CPPD(s). Integrating continuous HRV with symptom tracking could support low-burden and multimodal monitoring of sleep health in chronic pelvic pain, but prospective validation is needed before HRV can be used for diagnostic or treatment response decision making.

We present FootNet, a 453-image multi-view smartphone foot dataset for binary foot segmentation, with expertannotated masks across six anatomical views (dorsal, medial, and plantar, both left and right). We benchmark four segmentation models under a controlled protocol: U-Net with a MobileNetV2 encoder achieves the best performance (IoU 0.9268, Dice 0.9608, 95 % CI [0.9209, 0.9320]); DeepLabV3 with MobileNetV3-Large scores IoU 0.8984 (Dice 0.9449); UNet++ with MobileNetV2 scores IoU 0.8913 (Dice 0.9391); and SAM ViT-B with oracle boundingbox prompt scores IoU 0.9219 on the matched 191-image subset. Bonferroni-corrected Wilcoxon signed-rank tests (k = 6 comparisons) show U-Net significantly outperforms DeepLab (p < 0.001, r = 0.638) and SAM ViT-B with oracle boundingbox (p = 0.005, r = 0.202); UNet++ does not significantly differ from DeepLab (p = 0.062). Connected-component postprocessing yields negligible benefit (mean {triangleup}IoU = +0.0003, 12 of 453 images improved). The extended dataset is available upon request

White matter lesion (WML) identification, assessment, and characterization using magnetic resonance imaging (MRI) are fundamental for diagnosis and monitoring of multiple sclerosis (MS). Portable ultra-low field (pULF) MRI at 64 millitesla (mT) has been shown to visualize WML with at least one dimension greater than 4 mm. An automated WML segmentation tool catered to pULF-MRI can provide standardized and accurate quantitative measurements of WML volume. In this study, we sought to investigate and compare the accuracy of machine-learning (ML) and deep-learning (DL) pULF MRI segmentation tools. Same-day paired pULF (64mT) and high-field (HF, 3T) MRI scans from 84 adults with MS or suspected-MS (mean age {+/-} SD: 48 {+/-} 13, 62 females) included T2-FLAIR and T1w images. Reference WML segmentations were manually annotated on pULF T2-FLAIR for all scans, with WML confirmed with registered HF T2-FLAIR. HF reference WML segmentations were created. Four automated segmentation methods were applied to pULF scans: Method for Inter-Modal Segmentation Analysis (MIMoSA), an ML algorithm trained on HF WML masks; WMH-SynthSeg, a convolutional neural network model with flexible segmentation capabilities across field strengths and resolution; nnU-Net, a DL algorithm trained on pULF reference WML masks; and Pseudo-Label Assisted nnU-Net (PLAn), a DL algorithm pre-trained on HF reference WML masks and refined with 64mT reference WML masks. Two models were trained with nnU-Net, one using T2-FLAIR images only (nnU-Net-FL) and one using T1w and T2-FLAIR images (nnU-Net-FL/T1). The same was done with PLAn, creating PLAn-FL and PLAn-FL/T1. The six automated WML segmentation outputs were compared to the manual segmentations to determine Dice Similarity Coefficient (DSC) scores. Associations of WML volume estimates with clinical measures were investigated. DSC scores with pULF reference WML masks from PLAn-FL (DSC mean {+/-} SD: 0.50 {+/-} 0.24) outperformed MIMoSA (0.24 {+/-} 0.20, p < 0.0001), WMH-SynthSeg (0.30 {+/-} 0.18, p < 0.0001), nnU-Net-FL (0.41 {+/-} 0.24, p < 0.0001), and nnU-Net-FL/T1 (0.41 {+/-} 0.26, p = 0.0004). Worse Expanded Disability Status Scale (EDSS) and Scripps Neurologic Rating Scale (SNRS) scores were correlated with higher WML volumes in the pULF and HF reference masks. They were also correlated with WML volumes derived from WHM-SynthSeg, nnU-Net-FL, nnU-Net-FL/T1, PLAn-FL, and PLAn-FL/T1, but not MIMoSA. After adjusting for age, WHM-SynthSeg, nnU-Net FL, nnU-Net-FL/T1, PLAn-FL, and PLAn-FL/T1 had significant associations with EDSS and SNRS scores. nnU-Net and PLAn performed best in segmenting WML on pULF-MRI at 64 mT, providing accurate quantitative estimates of WML burden. Moreover, WML volumes estimated by these algorithms were associated with clinical measures of disability, underscoring their utility for reflecting clinical and radiological disease severity. Given pULF-MRI's mobility and lower cost, these findings highlight its relevance in clinical trials, particularly in involving more participants who face logistical constraints and barriers.

This study examined the extent to which goodness of pronunciation (GoP) scores and phonological posterior probabilities capture perceptual ratings of speech severity in individuals with motor speech disorders (MSD). Speech recordings of the word catastrophe were obtained from 489 participants, including 333 neurologically typical controls and 156 individuals with MSD. GoP scores were derived using traditional acoustic features and self-supervised speech representations, including WavLM and XLS-R, across multiple modeling approaches, while phonological posterior probabilities were extracted using Phonet. Model performance was evaluated using Kendall's rank correlations, regression, and receiver operating characteristic analyses against speech-language pathologists' perceptual ratings of sound distortion and intelligibility. Both GoP and phonological posterior probabilities were significantly associated with perceptual ratings. Self-supervised speech representations substantially outperformed traditional acoustic features, with WavLM-based GoP using k-nearest neighbors achieving the strongest performance. Across correlation, regression, and classification analyses, GoP consistently outperformed phonological posterior probabilities for both sound distortion and intelligibility. Age and gender had minimal influence on model-derived measures or their relationships with perceptual ratings. These findings demonstrate the value of self-supervised GoP as an objective measure of speech impairment while highlighting the complementary role of phonological posterior probabilities in characterizing articulatory aspects of motor speech disorders.

Aphasia diagnosis in Indonesia remains challenging due to limited culturally and linguistically appropriate instruments. Widely used tools such as the Boston Diagnostic Aphasia Examination (BDAE) and Western Aphasia Battery (WAB) are not adapted to the Indonesian context, while Tes Afasia untuk Diagnosis, Informasi, dan Rehabilitasi (TADIR) provides screening but lacks diagnostic accuracy. To address this gap, we developed the Instrumen Diagnosis dan Evaluasi Afasia (IDEA) for native Indonesian speakers and evaluated its validity, reliability, and normative cutoff values in cognitively healthy Indonesian adults. Eighty-three cognitively normal adults (screened using MoCA-Ina) with no history of neurological disease were assessed using IDEA, which evaluates six language domains. Items were adapted from existing tools and reviewed by experts. Content validity, internal consistency (Cronbachs alpha), and construct validity (Exploratory Factor Analysis) were analyzed using SPSS v25. A total of 83 participants were included (median age = 55.81 years, 54% secondary education). IDEA demonstrated good feasibility, with an average completion time of 45-60 minutes depending on participant engagement. Content validity was established by unanimous expert consensus. Construct validity showed meritorious sampling adequacy (KMO = .872) and significant sphericity (Bartletts test {chi}^2 (15) = 278.523, p<.001), supporting factor analysis. Internal consistency showed good reliability across six domains (Cronbachs = 0.896). IDEA is a valid and reliable tool for assessing aphasia in Indonesian natives. It is a culturally appropriate assessment tool which offers structured, domain-based evaluation and supports differential diagnosis of both classical and progressive aphasia syndromes. Keywords: Aphasia, Language Assessment, Indonesian, IDEA, Validity

ABSTRACT Background: Metallic implants such as orthopedic screws, prostheses, and dental hardware produce beam-hardening, photon-starvation, and streak artifacts that degrade computed tomography (CT) image quality, and the metal artifact reduction (MAR) methods developed to mitigate them require objective, reproducible benchmarking. Purpose: Objective evaluation of MAR algorithms in CT is hindered by the absence of phantoms that simultaneously provide anatomically realistic backgrounds, embedded implants of known geometry, and controllable, ground-truth--referenced artifact intensity. We present a dual-filament, voxel-level three-dimensional (3D) printing method that fulfills these requirements and demonstrate its capabilities on a clinically representative cervical spine case with embedded orthopedic spinal screws. Methods: The proposed method extends the PixelPrint framework, a fused-deposition-modeling (FDM) workflow that converts clinical Digital Imaging and Communications in Medicine (DICOM) data directly into 3D-printer Geometric code (G-code) without intermediate segmentation or surface meshing, to interleaved, voxel-level deposition of two filaments: a calcium-doped polylactic acid (PLA) for soft tissue and bone, and a higher-attenuation metal-doped PLA for metallic implants. For demonstration, anonymized DICOM data of a healthy cervical spine were used to design and fabricate three matched phantoms, each with six embedded spinal screws at C4--C6: a 0% metal-infill ground-truth phantom, a 50% medium-metal-infill phantom, and an 85% high-metal-infill phantom. All phantoms were scanned on a clinical spectral CT system at 120 kVp and 1000 mAs, reconstructed at 0.67 mm slice thickness with virtual monoenergetic imaging (VMI) across 50--190 keV. Method performance was characterized by region of interest (ROI)-based Hounsfield Unit (HU) agreement with the source patient data and by the noise-independent Gumbel-distribution p-index metric. Results: The dual-filament method reproduced patient anatomy, soft-tissue contrast, and screw geometry with high fidelity. ROI HU values agreed with patient data within {+/-}25 HU for soft tissue and trabecular bone; cortical regions were underestimated owing to the current ceiling of the calcium-doped PLA used in this study. The tunable-artifact behavior was quantified as follows: the Gumbel location parameter scaled monotonically from 46.7 HU (no-metal background) to 57.1 HU (50% infill) to 90.5 HU (85% infill) for the VMI 70 keV with standard filter. High-keV VMI reconstructions substantially reduced streak and beam-hardening artifacts while preserving anatomic detail. Conclusions: The proposed dual-filament, voxel-level PixelPrint method enables the fabrication of patient-specific, multi-material CT phantoms with embedded metallic implants and controllable, ground-truth--referenced artifact intensity. Although demonstrated here in a single cervical-spine case, the workflow is anatomy- and implant-agnostic by construction and could in principle be adapted to other musculoskeletal sites (e.g., knee, hip, dental) and implant materials, providing a reproducible methodological foundation for benchmarking MAR algorithms, characterizing spectral CT performance, and validating emerging photon-counting detector systems. Keywords: 3D printing methodology; fused deposition modeling; voxel-level multi-material printing; spectral computed tomography; metal artifact reduction; phantom design; orthopedic implants; dual filament; PixelPrint.

This study presents ReCo (Research Cosmos), a self-configuring and self-extending agentic research framework for the biomedical domain. ReCo is orchestrated by a large language model that interacts with native computing tools, bundled Model Context Protocol (MCP) servers, structured skills, persistent project memory, and a desktop interface. Its bundled MCP servers provide biomedical analysis capabilities while serving as implementation paradigms for integrating new computational and AI frameworks. Structured skills encode procedures for environment configuration and framework ingestion, enabling ReCo to inspect repositories, manuscripts, or local codebases; identify dependencies and execution patterns; create isolated runtime environments; design and implement MCP interfaces. Self-extension was evaluated using five heterogeneous systems: the Merlin computed tomography foundation model, MAISI-v2 medical image synthesis framework, asari liquid chromatography-mass spectrometry workflow, DosimeTron agentic radiation-dosimetry platform, and Orthanc DICOM server. ReCo successfully operationalized all five systems and completed predefined functional evaluations. Re-hosted DosimeTron outputs demonstrated near-perfect agreement with the reference pipeline across 651 organ observations (Pearson correlation and Lin concordance correlation coefficient, 0.99999; mean absolute percentage difference, 0.37%). Notably, ReCo configured Orthanc as a PACS-like coordination layer, integrated it with DosimeTron, Merlin, and TotalSegmentator, and orchestrated data retrieval, analysis, and return of valid DICOM RTSTRUCT, RTDOSE, and Structured Report. ReCo provides a unified environment for configuring, documenting, and operationalizing heterogeneous biomedical frameworks, reducing technical barriers to the adoption and integration of emerging computational and AI methods. The official open-source ReCo GitHub repository is available at: https://github.com/eltzanis/ReCo

Motor speech disorders (MSDs) are early markers of neurological disease, but expert perceptual analysis is rarely available outside specialized centers. Automated speech analysis offers a scalable alternative, yet prior studies have not systematically compared modeling approaches or assessed clinically relevant metrics in independent datasets. This study compared static acoustic features, articulatory informed Phonet features, and self-supervised pretrained models for binary and multi label MSD classification. We trained and evaluated models on 583 speech samples using speaker level splits. Baseline models included logistic regression and Gated Recurrent Units (GRUs) trained on eGeMAPS and MFCCs. We extracted three types of Phonet derived features and evaluated pretrained HuBERT and SSAST models in frozen, partially fine-tuned, and fully fine-tuned configurations. Binary classification distinguished MSDs from controls, while multi label classification identified six MSD subtypes. Models were assessed using validation AUC, and cut points were tested on two independent datasets. Pretrained and Phonet based models substantially outperformed static acoustic features. In binary classification, HuBERT achieved the highest AUC (0.95), while compact Phonet derived GRUs achieved comparable performance (up to 0.94). These models generalized well to independent datasets, maintaining high sensitivity (0.94) and specificity (0.97). In multi label classification, Phonet models achieved the highest macro average AUC (0.86), but threshold-based subtype performance declined on unseen data. Automated MSD detection is feasible and clinically promising. Binary classification generalized well, whereas multi label classification showed limited threshold stability across datasets.

Introduction: Intracerebral hemorrhage (ICH) carries higher morbidity and mortality than ischemic stroke. Recent studies have demonstrated improved patient outcomes by applying ultra-early bundled interventions including blood pressure management, coagulopathy reversal, and osmotic therapy. Effective strategies to deliver these ultra-early treatment options are currently being explored. On December 19th, 2022, the Mayo Clinic Comprehensive Stroke Center (CSC) launched the "ICH Phases'' communication system to accelerate ICH patient care. Objective: To evaluate adherence to the AHA/ASA guidelines in acute ICH care following the implementation of our novel-tiered paging system. Methods: We retrospectively reviewed patients admitted with spontaneous ICH during 2024 and 2025. We excluded traumatic cases. We extracted clinical data such as time to imaging, documentation of ICH score, blood pressure control, reversal of anticoagulation, venous thrombo-embolism (VTE) prophylaxis and discharge disposition. Results: Among 67 patients, 68.7% underwent CT imaging within 25 minutes. We documented the ICH score within 6 hours in 82.9% of patients. Nearly 94.7% of patients with SBP>140 mm Hg received antihypertensive therapy, yet only 18% reached target BP within 60 minutes. We completed the reversal of anticoagulation within 120 minutes in 75% of patients. VTE prophylaxis was initiated within 24 hours in 91% of patients. Discussion: Our novel system demonstrated adherence to the AHA/ASA guidelines, and time sensitive benchmarks in neuroimaging, reversal of anticoagulation, and VTE prophylaxis. Early BP control remains a challenge, that highlights the discrepancy between guidelines and real-ground implementation. Conclusion: A novel tiered paging system is effective for enhancing early ICH care. Such a holistic system remains critical for sustained improvement in quality of care.

Background: Mechanical thrombectomy (MT) is a time-sensitive reperfusion treatment for acute ischemic stroke caused by large-vessel occlusion. Workflow time metrics for MT remain poorly characterized in Indonesia, where stroke burden is substantial. This study describes pre-hospital and in-hospital time-interval metrics for MT across two major tertiary hospitals in Jakarta and evaluates institutional trends over a nine-year period. Methods: We conducted a retrospective descriptive study of consecutive patients undergoing MT at dr. Cipto Mangunkusumo National General Hospital (RSCM) and Prof. Dr. dr. Mahar Mardjono National Brain Center Hospital (RSPON) from 2017 to 2025. Pre-hospital and in-hospital time-interval metrics were reported as median (interquartile range [IQR]) and stratified by institution. Results: Among 330 registered patients, 71 were excluded due to incomplete data, leaving 259 in the final cohort (RSCM n=38; RSPON n=221). The pooled cohort had a mean age of 58.12 {+/-} 11.09 years; 63.71% were male. Hypertension was the most prevalent vascular risk factor (53.67%). Median door-to-CT time was 9 minutes (IQR 18), door-to-decision 101 minutes (IQR 100), and door-to-groin puncture 272 minutes (IQR 152). Total ischemic time (onset-to-groin puncture) was 468 minutes (IQR 294). MT volume increased substantially over the study period, particularly after 2022 at RSPON, which also demonstrated progressive improvement in in-hospital workflow times. RSCM showed increasing delays in later years, consistent with institutional congestion at a general multispecialty center. Conclusions: Early brain imaging was achievable at both centers; however, post-imaging delays particularly in CT-to-groin intervals, represent the dominant in-hospital bottleneck. Future quality-improvement efforts should prioritize decision-making, team mobilization, and pre-hospital coordination to reduce total ischemic time and improve access to reperfusion therapy.

We conducted a scaling evaluation of unlabeled pretraining for electrocardiogram foundation model performance. One-dimensional vision transformer masked autoencoders were pretrained across increasing ECG volumes and fine-tuned for rhythm, morphology, diagnostic, and structural heart disease tasks. Models pretrained below 400,000 ECGs failed to consistently exceed controls without self-supervised pre-training, whereas 600,000 to 800,000 ECGs improved AUROC across tasks, suggesting a minimum threshold for effective ECG representation learning.

Autonomic dysfunction is a common sequela of mild traumatic brain injury (mTBI). Physical activity progression is an integral component of mTBI rehabilitation, particularly in addressing autonomic dysfunction. However, clinicians often rely on point-in-time evaluation of orthostatic and exercise intolerance to guide activity recommendations. Commercially available wearable devices (e.g., Fitbits) provide an opportunity to evaluate heart rate response to activity in a real-world setting. Previous work has used physiologic (heart rate) and activity (step count) data to identify subgroups of adults with stroke that may be used to guide activity recommendations. This method may be useful to subgroup youth post-mTBI to identify those who have abnormal physiologic responses to activity. We aimed to identify subgroups using heart rate and step count data in adolescents presenting for specialty care after diagnosed mTBI. Eighty participants aged 13-18 within six months of mTBI diagnosis were recruited to wear a Fitbit Sense 2. Data from seven days and two nights collected within fourteen days of enrollment were included. A group-based steps per minute (SPM) threshold (25th percentile; 10 SPM) and individualized heart rate threshold (20% heart rate reserve (HRR)) were used to classify each minute of active daytime data into one of four quadrants: SPM>10 & HRR>20% (QI), SPM<10 & HRR>20% (QII), SPM<10 & HRR<20% (QIII), and SPM>10 & HRR<20% (QIV). We used percentage of minutes in each quadrant, mean steps per day, percentage of minutes with zero steps, mean SPM in QI, and resting heart rate in a k-means clustering algorithm to identify subgroups. We evaluated subgroup differences by clustering variables using Kruskal-Wallis tests. Sixty-one participants were included. Three subgroups emerged: Sedentary (n=12), Active (n=23), and Atypically Elevated Heart Rate (AEHR; n=26). Subgroups varied significantly on all clustering variables (p<0.01). The Active subgroup took a high number of steps per day, had lower sedentary time, and had the highest activity intensity (mean SPM in QI). The Sedentary subgroup took fewer steps per day compared to the Active subgroup, had high sedentary time, and showed the highest resting heart rate. The AEHR subgroup took fewer steps per day compared to the Active subgroup and had high sedentary time. The AEHR subgroup also spent a higher percentage of time with an atypically high heart rate response to low levels of activity compared to the other subgroups. Our findings suggest that data from wearable devices can identify subgroups of adolescents with mTBI with distinct physiologic/physical activity profiles, which may ultimately be used to inform personalized activity prescriptions. Future work should aim to understand how the identified subgroups relate to longitudinal outcomes.

Automated interpretation of neuroimaging studies requires simultaneous assessment of multiple imaging evidence variables, each tied to distinct anatomical structures. Vision-language models (VLMs) offer a unified framework for multi-task analysis, but adapting pre-trained VLMs remains challenging. Full fine-tuning is computationally prohibitive, and joint multi-task training requires simultaneous access to all task data, which is often infeasible in clinical settings. Although model merging enables multi-task composition without joint re-training, existing methods focus on post-hoc algorithms with limited extension to VLMs and minimal application to neuroimaging. Here, we present GRadient-guided Adapter Merging (GRAM), a layer-selective low-rank adaptation (LoRA)-based fine-tuning and merging framework for multi-task neuroimaging visual question-answering (VQA). GRAM uses a gradient ratio that contrasts class-specific gradients to identify task-discriminative layers, and applies subspace-constrained projected gradient descent to restrict LoRA updates to directions consistent with the geometry of the pre-trained model. We leveraged a structured VQA benchmark, developed from the National Alzheimer's Coordinating Center (NACC) dataset, that pairs multi-sequence brain MRI studies with question-answer pairs across clinically relevant imaging evidence variables. Experiments on the VQA benchmark showed that GRAM outperformed or matched all-layer LoRA fine-tuning and a standard merging baseline while reducing inter-task interference during merging, and approached or surpassed the performance of joint multi-task training without joint re-training.

Background: Post-stroke dysphagia (PSD) contributes to aspiration, pneumonia, malnutrition, prolonged hospitalization and mortality. We evaluated the discrimination, validity and readiness of machine learning and data-driven prediction models for PSD-related outcomes. Methods: Following a prospectively registered protocol (PROSPERO CRD420261419259), we searched PubMed/MEDLINE, Embase, Web of Science Core Collection, CINAHL and CENTRAL from inception through June 7, 2026. Eligible studies developed or validated multivariable prediction models for PSD-related outcomes in adults with stroke. We used PROBAST and PROBAST+AI to assess risk of bias and applicability and TRIPOD+AI to evaluate reporting. Area under the curve (AUC) estimates were pooled on the logit scale with random-effects models. Results: Twenty-four studies were included and ten contributed to meta-analysis. Four studies predicting early or incident PSD yielded a pooled AUC of 0.94 (95% CI 0.60-0.99; I2 = 95.6%). Pooled AUCs were 0.84 (95% CI 0.71-0.92) for aspiration or penetration-aspiration and 0.89 (95% CI 0.24-1.00) for severe dysphagia. The exploratory analysis of all ten risk-prediction models produced an AUC of 0.90 (95% CI 0.80-0.95), but heterogeneity was substantial (I2 = 90.3%) and the prediction interval was 0.51-0.99. Every study had high risk of bias because of analysis-domain concerns; calibration and external validation were uncommon. Conclusions: Reported discrimination was often high, but the evidence does not establish reliable performance in care. Independent validation, calibration, complete model reporting and clinical-impact studies are needed before these models guide post-stroke swallowing care. Keywords: Post-stroke dysphagia; Stroke; Deglutition disorders; Machine learning; Clinical prediction model; Area under the curve; Meta-analysis

Background and Objectives: Stereoelectroencephalography (SEEG) involves the implantation of intracerebral electrodes to investigate drug-resistant epilepsy. SEEG requires millimetric accuracy to ensure safety and optimal mapping. Although studies have evaluated SEEG accuracy, there is substantial variability in reporting. Here we report on implantation accuracy in a large series using the most common accuracy metrics described in the literature and perform a detailed analysis of contributing factors. Methods: SEEG implantations between 2013 and 2025 were included. Application accuracy was computed for each implanted electrode. Specifically, Euclidean, radial, depth, and angle error were calculated at both target and entry points. Correlative and multivariable analyses were conducted between each variable and error metric. Trajectories were also grouped by atlas-derived lobar target. Results: No metrics met assumptions of normality and thus we report accuracy using median with interquartile range (IQR). In a series of 3176 trajectories, median Euclidean target and entry errors were lower for robot-assisted electrodes (n=2858) at 2.19 (IQR: 1.54-2.98) mm and 1.38 (IQR: 0.89-2.01) mm respectively, compared to frame-based (n=318, p<.001) at 2.76 (IQR:1.79-3.76) mm and 2.21 (IQR: 1.42-3.32) mm. Correlation and multivariable regression analysis showed target error was positively correlated with implantation angle, scalp thickness, skull thickness, and trajectory length. Target error was also higher in obese patients. On lobar analysis, parietal lobe trajectories were the most accurate and frontal lobe trajectories were the least accurate. On temporal lobe trajectory analysis, posterior hippocampus trajectories were the most accurate and temporal pole trajectories were the least accurate. Presence of mesial temporal sclerosis also impacted accuracy. Conclusions: We present a detailed description of SEEG implantation accuracy, demonstrating the superior accuracy and speed of robot-assisted to frame-based methods. Furthermore, we analyzed how accuracy varies with specific factors from a global to trajectory level, which can be accounted for when planning SEEG implantations.