Signals, in order.

EPFL researchers have created VirTues, a foundation model that learns from spatial proteomics data across 31 clinical cohorts, 12,000 images, and 5,000 patients. The model uses a new Transformer architecture to integrate diverse protein panels and can be applied to multiple analyses, from cell‑level mapping to patient outcome prediction. VirTues has been published in Nature and is designed to be flexible, allowing new tissue samples to be compared within the same framework.

Source: EPFL News

Published intelligence from the original source record.

Source: EPFL News

This research develops and externally validates the FORECAST-PC machine learning models to predict 90‑day functional outcomes for patients with acute ischemic stroke in the posterior circulation who are selected for endovascular thrombectomy (EVT). The study focuses on creating robust prediction scores that can guide clinical decision‑making and improve patient prognostication. It demonstrates the feasibility of applying advanced analytics to a specific stroke population in the United States.

Source: ClinicalTrials.gov

The AI-ACLD Study investigates an AI-based system that records variations in patients with advanced chronic liver disease using data collected from an Apple Watch. It aims to determine whether the system can detect early clinical improvement or deterioration and improve care management and quality of life. The study is conducted in the United States.

Source: ClinicalTrials.gov

The release of OpenTela version 0.0.6 introduces several improvements. It adds peer metadata such as hostname, version, labels, and status, corrects the reported build‑time version in Peer.Version, mirrors release binaries to CSCS ocf‑share via FirecREST, and resolves symlink targets to absolute paths. These changes enhance the software’s functionality and deployment reliability.

Source: GitHub Releases & Activity

This release updates the OpenTela project with new features and bug fixes, including ROCm stats printing, asynchronous CRDT synchronization, and a fix for orphan CID autoban issues. The update also marks the first contribution from user @xzyaoi. The release demonstrates ongoing development within the Swiss AI community, enhancing the project's capabilities and collaboration.

Source: GitHub Releases & Activity

This retrospective cohort study uses radiographic and CT imaging data from Schulthess Klinik Zürich to develop and validate AI-based tools for assessing trapeziometacarpal (TMC) joint osteoarthritis and monitoring implants. The project comprises four subprojects: a new radiographic classification system for TMC OA, deep‑learning automation of that classification, automated detection of implant migration, and 3D reconstruction of the TMC joint from biplanar radiographs. Data are drawn from patients with the Touch implant and from other wrist‑related conditions, and the resulting prototypes are…

Source: ClinicalTrials.gov

Published intelligence from the original source record.

Source: ClinicalTrials.gov

The United States research project develops clinical decision support tools that combine established diagnostic variables with machine learning models to rapidly diagnose life‑threatening cardiovascular conditions in emergency department patients presenting with chest pain or dyspnea. The goal is to improve diagnostic accuracy, accelerate patient management, and reduce medical errors.

Source: ClinicalTrials.gov

The project aims to develop a predictive model for clinically significant depressive symptoms (CSDS) in patients undergoing coronary artery bypass graft (CABG) surgery using pre‑ and perioperative data. CSDS occur in about 30 % of CABG patients, which is four times higher than in the general population, and are linked to poor quality of life and increased morbidity and mortality. The model seeks to identify patients at risk so clinicians can make informed decisions and implement preventive measures.

Source: ClinicalTrials.gov