participant recruitment

Current AI-driven healthcare models predominantly rely on electronic health records (EHRs), overlooking real-time, multimodal patient–clinician interactions—such as audio, video, and text—which constitute a core clinical information source and risk reinforcing a narrow biomedical perspective. Method: We developed and validated the first reproducible, longitudinal, multimodal system that systematically integrates 360-degree audiovisual recordings, structured patient-reported outcome (PRO) questionnaires, and EHR data to construct a standardized dataset dynamically reflecting authentic clinical encounters. Our methodology encompasses ethically compliant, cross-modal data acquisition, linkage, and governance workflows. Contribution/Results: As of August 2025, the system achieved 97% clinician and 75% patient participation rates; 76% of visits were fully multimodally recorded, and 96% of follow-up PRO questionnaires were returned—demonstrating high feasibility and implementation efficacy. This infrastructure overcomes the limitations of unimodal EHRs and establishes a new paradigm and foundational resource for interaction-centered AI in healthcare.

To address challenges in clinical trials—including manual protocol execution, poor real-time responsiveness, and delayed adherence monitoring—this paper proposes a personalized participant agent system integrating finite-state machines (FSMs) with large language models (LLMs). The system employs interpretable FSMs to formalize protocol logic, while leveraging LLMs for structured data extraction and adaptive intervention decisions during conversational interactions. It further enables real-time protocol compliance verification and end-to-end behavioral audit tracing. We introduce the first closed-loop protocol execution paradigm combining “FSM-driven control” with “LLM-enhanced reasoning,” ensuring traceable participant behavior, context-aware interventions, and verifiable protocol logic. Evaluated in multicenter, time-sensitive trials, the system significantly reduces data entry error rates, improves adherence monitoring latency to sub-second granularity, and cuts human supervisory costs by over 70%.

Clinical workflow fragmentation severely impedes efficiency: heterogeneous scripting, ad-hoc model ensembles, and lack of data-driven modality identification and standardized outputs result in high deployment overhead, costly monitoring, and poor interoperability. To address this, we propose a healthcare-first vision-language unified framework that pioneers the use of a single vision-language model (VLM) for two-tier clinical decision-making—first, an auditable, three-stage routing mechanism matches inputs to expert-defined model cards; second, domain-specific multi-task joint inference (with early-exit capability and candidate arbitration) adheres to clinical risk constraints. Leveraging phased prompting, a candidate answer selector, and specialty-specific fine-tuning, our framework unifies modality identification, abnormality classification, model selection, and multi-task reasoning. Evaluated across gastroenterology, hematology, ophthalmology, and pathology, our single-model solution achieves performance on par with specialized models while substantially reducing deployment complexity, operational overhead, and integration effort.

Clinical trial eligibility criteria design faces challenges including a vast exploration space, weak interactivity, and difficulty integrating fine-grained electronic health record (EHR) features. To address these, we propose a knowledge-driven and outcome-driven dual-path visual analytics system. The system enables clinicians to iteratively explore eligibility criteria over multidimensional patient clinical indicators—such as vital signs, laboratory test results, and temporal medication histories—and incorporates history tracking and data provenance to support coupled analysis of inclusion/exclusion criteria, patient characteristics, and clinical outcomes. Key innovations include dual-path coupled modeling, dynamic criteria–outcome mapping visualization, and interpretable decision logging. Evaluated on real-world EHR data from septic shock and sepsis-associated acute kidney injury cohorts, the system significantly improves the efficiency, scientific rigor, and transparency of eligibility criteria design.

This study addresses the multi-objective optimization challenge faced by blood collection centers in efficiently allocating multi-site donation invitations while satisfying blood type demands, ensuring donor safety and convenience, and preventing donor fatigue. The authors propose the first unified optimization framework that integrates donor eligibility, geographic accessibility, blood type requirements, safety constraints, and forward-looking demand forecasting. Combining binary integer linear programming (BILP) with an efficient greedy heuristic, the framework dynamically generates scalable donor invitation schedules. Empirical results from the Lisbon region demonstrate that the approach achieves an 86.1% demand fulfillment rate; the greedy algorithm accelerates computation by 115× and reduces memory usage by 188× compared to BILP, at a modest performance cost of only 3.9 percentage points, thereby confirming its feasibility for large-scale deployment and highlighting the critical role of reactivating dormant donors in closing supply-demand gaps.

This work addresses the common lack of continuous evaluation and governance mechanisms in deployed clinical AI systems, which hinders dynamic performance optimization. The authors propose the first end-to-end continuous governance framework tailored for clinical AI, integrating standards-driven validation, A/B testing for controlled version updates, real-time performance monitoring, fault tolerance, and deep integration with electronic health records (EHRs) to establish a closed-loop synergy between engineering iteration and clinical feedback. Applied to Hyperscribe—a speech-to-structured-clinical-note system—the framework achieved substantial improvements over seven iterative cycles: median clinician rating increased from 84% to 95%, negative user feedback decreased from 79% to 30%, median audio processing latency was 8.1 seconds, and task completion rate reached 99.6%, collectively enhancing system reliability and user satisfaction.

This study addresses the opacity and accountability challenges in AI-powered hiring systems, which stem from their complex supply chains that obscure the origins of algorithmic bias. Through regulatory analysis, system dependency modeling, and a multi-stakeholder perspective—complemented by case studies and an examination of implementation ambiguities—the work demonstrates for the first time that bias arises primarily from interactions among system components rather than from isolated modules. It further identifies a structural contradiction: deploying organizations bear legal responsibility yet lack technical visibility into upstream components. The research pinpoints two core barriers to effective bias assessment and accountability and proposes a holistic, supply-chain-wide governance framework featuring system-level audits, vendor guidelines, continuous monitoring, and cross-component documentation.

This study addresses the lack of standardized and automated case planning processes in medical social work, which currently relies heavily on individual practitioner experience and suffers from inefficiency. The authors propose a model-agnostic, open-source large language model (LLM) workflow that systematically integrates established social work practice frameworks into LLM prompt design for the first time. The approach decomposes case planning into six sequential stages—assessment, problem analysis, goal setting, intervention planning, risk anticipation, and outcome evaluation—and combines structured client profiling with staged prompt engineering to generate professional, reviewable draft assessment forms and service plans. Designed to be compatible across multiple LLM platforms, the framework ensures cross-model reproducibility, and its code has been publicly released to provide a standardized tool for advancing intelligent support in medical social work.

Clinical research is often hindered by cumbersome workflows, reliance on programming expertise, and restricted access to sensitive data, limiting its accessibility to non-technical investigators. This work proposes CARIS (Clinical AI Research System), the first framework integrating large language models with the Model Context Protocol (MCP) to enable end-to-end, code-free, and privacy-preserving automated clinical research—from study design and analysis to report generation—while supporting human-in-the-loop iterative refinement. CARIS incorporates Vibe machine learning, literature retrieval, and the TRIPOD+AI reporting framework. Evaluated across three heterogeneous clinical datasets, the system generated study protocols and ethics documentation within 3–4 iterative rounds, achieving 96% coverage in LLM-based report assessments and 82% in manual evaluations, substantially lowering technical barriers while safeguarding data privacy.