Publications

Understanding the decision-making process of black-box neural networks is crucial for safe use of AI in high-stakes medical tasks such as histopathology. We present Adaptive Example Selection (AES), a prototype-based explainable AI framework that improves interpretability of deep learning models for mitosis detection. AES retrieves a sparse set of supporting and contradicting real-world prototype images to locally approximate the model's confidence surface with high fidelity ([Formula: see text]). The framework is integrated with a robust Faster R-CNN detector that demonstrates strong cross-tumor performance, for example achieving an F1-score of 0.84 on the Canine Cutaneous Mast Cell Tumor dataset. AES generates concise, case-specific explanations that faithfully capture local decision boundaries while linking predictions to interpretable exemplars. This enables clinicians to visualize model reasoning, assess uncertainty, and conduct contrastive analyses. Unlike prior methods focused on discrete class predictions, AES shows how similarity to mitotic and non-mitotic prototypes shapes graded confidence, enhancing transparency, trust, and practical adoption of AI-assisted mitosis detection in cancer diagnostics.

IMPORTANCE: The initial treatment by emergency medical services (EMS) significantly affects the outcomes for severely injured patients. Effective control of hemorrhage, proper administration of blood products, and adherence to traumatic brain injury guidelines can reduce morbidity and mortality after trauma. Additionally, the experience of prehospital clinicians in high-acuity nontrauma conditions is associated with improved outcomes.

OBJECTIVE: To evaluate the association of annual trauma patient volume and outcomes at the individual EMS clinician level.

DESIGN, SETTING, AND PARTICIPANTS: A secondary subset analysis was performed of the Linking Investigations in Trauma and Emergency Services (LITES) Task Order 1 study, a prospective observational cohort from 2017 to 2021. It includes severely injured patients, identified by an Injury Severity Score of 9 or higher, who were transported to a trauma center by 1 air and 1 ground agency. Data were analyzed from February 2023 to June 2024.

EXPOSURES: EMS crew mean 3-year adult trauma volume and 6-hour mortality and several EMS industry quality metrics.

MAIN OUTCOMES AND MEASURES: Patient-level risk-adjusted regression models were constructed to determine the association between EMS crew mean 3-year adult trauma volume and 6-hour mortality and several EMS industry quality metrics. The association of airway success metrics and procedural intubation volume was also assessed.

RESULTS: A total of 6769 patient-clinician interactions involving 359 clinicians and 3649 patients (median [IQR] age, 54 [33-70] years; 2490 male [68.2%]) were included in this study. For every increase of 5 adult trauma patients annually per crew, there was a 10% decrease in 6-hour mortality (adjusted odds ratio [aOR], 0.899; 95% CI, 0.811-0.996) and a 2.6% decrease in in-hospital mortality (aOR, 0.974; 95% CI, 949-0.999). In subgroup analyses including traumatic brain injury (aOR, 0.974; 95% CI, 0.949-0.999) and prehospital shock (aOR, 0.974; 95% CI, 0.949-0.999), volume was associated with reduced 6-hour mortality. Highest trauma volume among treating EMS crew members, nontrauma volume, and years of experience were not significantly associated with differences in mortality. Among EMS industry quality metrics, decreasing scene time (regression coefficient, -0.134; 95% CI, -0.191 to -0.077) was significantly associated with higher clinician volume. Intubation procedural volume was associated with greater odds of success without hypotension or hypoxia (aOR, 1.110; 95% CI, 1.040-1.190).

CONCLUSIONS AND RELEVANCE: Results of this cohort study suggest that higher patient volumes per EMS clinician were associated with lower early mortality rates after trauma. Exploring this association further is essential to optimize staffing, education strategies, and performance benchmarks.

OBJECTIVE: The objective of this study was to examine patient experiences with virtual (telephone and video) encounters in primary care and make recommendations to inform the broader adoption of virtual care.

DESIGN: A descriptive qualitative study using semi-structured interviews for data collection.

SETTING: Ontario, Canada.

PARTICIPANTS: Fifty-five primary care patients across Ontario, Canada, who had experienced at least one virtual (telephone or video) encounter with a healthcare provider in primary care, participated in semi-structured individual interviews conducted between 15 January 2021 and 22 March 2021.

RESULTS: With respect to patients' experiences with virtual care appointments, we identified the following seven themes: (1) Enhancing access, (2) Importance of patient-provider relationship, (3) Active communication and attunement, (4) Assuring privacy and confidentiality, (5) Shorter appointments, (6) Asynchronous technologies being underutilised and (7) Strengthening the future of virtual care. Despite the rapid adoption of synchronous virtual care, participants generally reported positive experiences. Virtual care enhanced access to care and was overwhelmingly supported for continued use. While new patient-provider relationships faced challenges, pre-existing, positive relationships thrived. Concerns about the shortness of virtual care appointments were reported.

CONCLUSIONS: Virtual care offers a promising modality for patients to experience care. Moving forward, primary care practices should consider expanding options for asynchronous virtual care, consider the length of virtual care appointments and offer patients greater choice in the modality of their care appointments.

BACKGROUND: Traumatic intracranial hemorrhage (tICH) is a major driver responsible for traumatic brain injury (TBI)-related deaths. Studies have demonstrated an associated survival benefit with plasma administration among patients with tICH. The administration of whole blood (WB), from which plasma is derived, may provide similar or greater benefits. We hypothesize that WB, compared to plasma-based resuscitation, is associated with reduced 30-day mortality among patients presenting with tICH.

DESIGN: This cohort study using the American College of Surgeons Trauma Quality Improvement Program databank from January 1, 2020, and December 31, 2021, included adult trauma patients with tICH presenting to level I and level II US and Canadian civilian trauma centers. We compared WB resuscitation to plasma-based resuscitation within 4 hours of emergency department (ED) arrival. Primary outcome was mortality at 30 days.

RESULTS: Among 9175 patients analyzed, 1238 (14%) received whole blood and 7937 (86%) received plasma-based resuscitation. The overall 30-day mortality was 43%. WB was associated with reduced mortality at 30 days, demonstrating an unadjusted 30% lower risk of mortality (hazard ratio, 0.70; 95%CI, 0.63-0.77; P = <.001) and a 24% lower risk of 30-day mortality after adjusting for confounders (hazard ratio, 0.76; 95% CI, 0.58-0.98; P =.04).

CONCLUSION: In this cohort study, resuscitation with whole blood was associated with lower 30-day mortality compared to plasma-based resuscitation among patients presenting with tICH. These findings highlight WB as a promising therapeutic strategy for tICH, underscoring the need for future prospective studies to validate its clinical effectiveness.

UNLABELLED: Human platelets change over their 7-10 day lifespan, yet the molecular mechanisms underlying platelet aging remain poorly defined. Using two independent RNA sequencing datasets of fluorescence-activated cell sorted young and old human platelets, we developed a unified transcriptomic model to characterize RNA metabolism across the platelet lifespan, which we termed platelet molecular maturation. This was applied to RNA sequencing data from room-temperature stored platelets (up to 7 days) and cold-stored platelets (7, 14, or 21 days). We identified highly concordant aging signatures, including 6,015 shared expressed genes and 2,008 shared differentially expressed genes (DEGs) with strongly correlated fold changes, demonstrating a conserved platelet aging program. Nucleotide-level analyses revealed preferential 3'-directed degradation among downregulated transcripts during endogenous platelet aging and room-temperature storage, supporting an organized RNA decay process that was correlated with platelet function changes. Room-temperature storage recapitulated platelet molecular maturation, showing concordance with aging-related gene expression changes and enrichment of downregulated gene sets. In contrast, cold-storage significantly attenuated platelet molecular maturation and 3'-directed degradation. A total of 669 genes were consistently differentially expressed between room-temperature and cold-stored platelets, while no DEGs were detected during cold-storage, indicating transcriptional stability. Platelet transcript stability in cold-storage correlated with preserved platelet hemostatic function. These findings establish platelet molecular maturation as a conserved, functionally relevant model linking endogenous platelet aging to platelet storage lesions and providing mechanistic insight into preserved platelet hemostatic function in cold-storage. This atlas of platelet RNA metabolism supports biomarker discovery and strategies to improve storage.

KEY POINTS: By integrating multiple high-quality RNA sequencing datasets with novel analytic approaches tailored to the biology of anucleate platelets, we show that platelet aging is not a passive process of transcript decay, but follows a structured and reproducible molecular trajectory both endogenously and in storage, which we term platelet molecular maturation.Storage temperature emerged as a dominant modifier of this trajectory, with cold-storage markedly slowing RNA metabolic kinetics and preserving transcripts associated with younger, more hemostatically competent platelets.Together, these findings provide mechanistic insight into the platelet storage lesion and identify transcriptomic features that may serve as biomarkers or therapeutic targets to extend platelet shelf life.

Electrical stimulation at cellular resolution to restore the function of neural circuits is limited by the density of available electrode arrays. Although current steering with multi-electrode stimulation can be used to target cells between electrodes, it has not been proven for systematically targeting individual cells. We develop a framework for cellular-resolution current steering, leveraging the biophysics of electrically evoked spike generation, and test its efficacy in isolated macaque and human retina. Currents were passed through three electrodes simultaneously using large-scale high-density microelectrode arrays, directly evoking single spikes in retinal ganglion cells. The currents combined either linearly or nonlinearly to drive spiking, depending on the geometry of the electrodes relative to the cell. These findings were captured by a biophysical model and by a simpler parametric model in which spikes can initiate at several sites on the cell membrane and were leveraged to efficiently identify multi-electrode stimulation patterns that optimized cellular selectivity.

BACKGROUND: Traumatic brain injury (TBI) patients on antiplatelet medications lack definitive treatment for reversal of platelet inhibition and restoration of injury-induced platelet dysfunction. The use of platelet transfusions in this patient population remains controversial. Cold-stored platelets (CSPs) may be hemostatically superior to their room temperature platelet (RTP) counterparts for hemostatic resuscitation, but their impact on post-transfusion platelet function has yet to be assessed clinically. We aimed to evaluate the effect of CSP or RTP on post-transfusion platelet function in TBI patients on antiplatelet medications. We hypothesized that CSP would better restore platelet function based on the extensive in vitro data suggesting hemostatic superiority.

METHODS: We performed a post hoc analysis of a randomized controlled trial comparing CSP and RTP in TBI patients on antiplatelet medications. Platelet hemostatic function was determined pretransfusion and posttransfusion using VerifyNow or thromboelastography with platelet mapping (TEG-PM). Clinical outcomes included 30-day mortality, need for neurosurgical intervention, and follow-up Rotterdam scores.

RESULTS: Of the 94 patients with available data, 49 received CSP and 45 received RTP. Baseline characteristics and pre-transfusion assay measurements were similar between groups. Cold-stored platelet recipients had fewer neurosurgical procedures compared with RTP recipients (4.1% vs. 20.0%, p = 0.016). Room temperature platelet recipients showed a greater increase in TEG-PM kaolin maximum amplitude after transfusion compared with CSP recipients (2.4 mm vs. 0.6 mm, p = 0.004). No other differences were observed between RTP and CSP transfusions.

CONCLUSION: Despite a reduction in neurosurgical events, CSP did not significantly improve observed platelet function in TBI patients on antiplatelet medications. Our findings highlight the disconnect between platelet function assays and clinical results and suggest transfusion of CSP versus RTP has minimal effect on platelet hemostatic function. A definitive trial is needed to assess the efficacy of differentially stored products in the bleeding patient, with consideration placed on how platelet hemostatic function is assessed.

LEVEL OF EVIDENCE: Prognostic/epidemiologic; Level III.