Multimodal Physiologic State Intelligence Study

Overview

The OMNIPHYS study will investigate whether multimodal ultrasound and other routinely collected clinical and physiologic data can be combined with artificial intelligence to characterize changes in a person's physiologic state over time. The study will examine cardiac, vascular, pulmonary, and systemic physiologic measurements and develop longitudinal models that describe baseline physiology, physiologic perturbation, compensation, deterioration, treatment response, and recovery. The study is observational and will not assign experimental treatments. The goal is to determine whether changes in multimodal physiologic patterns can be identified and characterized earlier and more reliably than conventional single-time-point assessment.

OMNIPHYS: A Prospective Multimodal Longitudinal Study of AI-Enabled Physiologic State Modeling and Early Detection of Clinical Deterioration The OMNIPHYS study is a prospective, longitudinal observational investigation designed to develop and evaluate a multimodal framework for characterizing human physiologic state and physiologic trajectory over time. The study is centered on the hypothesis that clinically meaningful deterioration may be preceded by measurable changes across multiple physiologic domains, including cardiac function, vascular flow, pulmonary findings, systemic venous congestion, and other routinely available clinical measurements. The study will collect and analyze multimodal data obtained during clinically appropriate assessments. Ultrasound-derived data may include two-dimensional and three-dimensional imaging, color Doppler, spectral Doppler, cardiac motion and functional measurements, vascular flow characteristics, venous findings, and pulmonary ultrasound observations. Additional clinical data may include vital signs, electrocardiographic measurements, oxygen saturation, laboratory results, medication and intervention records, diagnoses, encounters, and other relevant longitudinal clinical observations. The study will develop a Physiologic State Vector (PSV) representing multidimensional observations of an individual at defined points in time. Rather than treating disease status as a binary outcome, OMNIPHYS will investigate physiologic states and transitions between states. The prespecified conceptual state framework includes: S0 - Baseline: stable observed physiologic phenotype. S1 - Physiologic Perturbation: measurable deviation from an individual's established baseline. S2 - Compensation: persistent physiologic abnormality without defined acute clinical deterioration. S3 - Pre-Decompensation: a longitudinal trajectory associated with increasing probability of clinically significant deterioration. S4 - Acute Decompensation: clinically meaningful physiologic or clinical deterioration. S5 - Intervention Response: measurable physiologic change following clinical intervention. S6 - Recovery: movement toward the participant's prior or expected physiologic state. S7 - Persistent Dysfunction: sustained deviation from baseline following an acute event or intervention. The principal research construct is therefore: Physiologic State → Longitudinal Trajectory → State Transition → Clinical Outcome The primary investigational performance measure will be the Physiologic Transition Detection Time (PTDT), defined as the interval between a prospectively defined algorithmic detection of a clinically meaningful physiologic transition and the corresponding predefined clinical reference event. PTDT will be evaluated as a research endpoint and will not independently direct clinical care unless separately authorized under an applicable clinical protocol. Secondary analyses will evaluate physiologic-state classification, trajectory prediction, treatment-response characterization, recovery prediction, longitudinal model calibration, uncertainty estimation, human-AI concordance, cross-site performance, cross-device robustness, and model stability over time. The OMNIPHYS framework is intended to support multiple pathology-specific cohorts while maintaining a common physiologic modeling architecture. Potential cohorts may include heart failure, cardiomyopathy, pulmonary hypertension, valvular disease, myocarditis, pericardial disease, pulmonary edema, acute respiratory distress syndrome, pulmonary embolic disease, venous thrombosis, sepsis-associated cardiac dysfunction, systemic venous congestion, acute kidney injury, and other clinically appropriate conditions. A designated research component will investigate federated learning, in which appropriately governed model-development processes may be evaluated across participating institutions without requiring centralized transfer of patient-level datasets. The study will assess whether multimodal physiologic models maintain performance across institutions, ultrasound platforms, acquisition environments, operators, and heterogeneous clinical populations. The study may additionally investigate a research construct termed the TRUWAY Physiologic Digital Twin, defined as a computational longitudinal representation of observed physiologic measurements and their temporal relationships. This construct is intended for research into trajectory modeling and prediction and does not represent a clinical diagnosis or autonomous clinical decision-making system. Data interoperability and research infrastructure may incorporate recognized clinical data representations, including DICOM for medical imaging and FHIR-compatible clinical data structures where available and appropriate. Data provenance, version control, quality assurance, model-version tracking, and longitudinal auditability will be incorporated into the research architecture. The anticipated study horizon is approximately 20 years, from September 2026 through September 2046, permitting investigation of short-term physiologic transitions as well as long-term disease trajectories, recurrence, recovery, persistent dysfunction, technological evolution, and longitudinal model performance. OMNIPHYS is designed as an observational research protocol. It does not by itself prescribe medical treatment, alter clinical management, or establish that an investigational algorithm is safe or effective for clinical use. Research findings will be evaluated according to the approved protocol, applicable institutional requirements, human-subject protections, data-governance requirements, and prospective statistical analysis plans.