Every story in this issue with a short summary, so you can decide what to read before leaving the site. Summaries are written by AI (Claude) only from each publisher's own description and checked by Jev for anything not in that description. They are not full translations. Follow "Read the original" for the full story.
A meta-analysis of 18 studies finds that deep learning applied to dynamic contrast-enhanced breast MRI reaches 90 percent sensitivity and 83 percent specificity. It was used to distinguish benign from malignant lesions.
Read the original →The ECG-PH clearance adds pulmonary hypertension to Tempus' FDA-cleared AI products. Its existing cleared products cover atrial fibrillation and low ejection fraction.
Read the original →Three deep learning architectures trained on dermoscopic images reached the same band of accuracy. Their measured energy use differed by more than fourfold.
Read the original →Researchers have built a temporal-aware retrieval-augmented generation framework. It helps large language models generate safer, evidence-based treatment plans for idiopathic pulmonary fibrosis.
Read the original →A prospective study of 412 children found that GPT-5.2 Instant missed nearly half of elbow fractures on radiographs. Its sensitivity was 56.1 percent, with only fair agreement with expert diagnosis.
Read the original →ModMed, a provider of software and services to specialty medical practices, announced the opening of MOMENTUM 2026, its largest annual user conference to date. Nearly 2,000 attendees have gathered at the Gaylord Palms Resort & Convention Center in Orlando. From October 16–18, attendees will see live demonstrations of The AI-Powered Practice and take part in hands-on training.
Read the original →The Indian Institute of Management (IIM), Jammu, has signed a memorandum of understanding with the University of Alabama at Birmingham (UAB) in the US, an official said on Saturday. The agreement is aimed at strengthening international academic collaboration and fostering innovation in healthcare.
Read the original →A University of Manchester team has built an AI workflow that combines stakeholder weights, Bayesian deep learning and reinforcement learning. It is used to sequence UK public building retrofits.
Read the original →Tautomerism involves the relocation of a single proton and a corresponding rearrangement of bonds. The description presents it as conceptually simple.
Read the original →A new review argues that AI-driven antiviral discovery must be benchmarked against respiratory RNA virus polymerases under virologically realistic conditions. It says this should be done rather than relying on docking scores.
Read the original →Healthcare delivery and biomedical informatics are undergoing an architectural paradigm shift. Conversational software in healthcare, dominated by rule-based, scripted decision trees and stateless conversational wrappers, is being superseded by agentic AI. Earlier healthcare chatbots functioned primarily as rigid digital directories constrained by hardcoded deterministic pathways and static symptom-checking logic.
Read the original →The global digital slide scanner market is expected to grow steadily toward 2035. Wider adoption of remote pathology diagnosis and AI integration into pathology workflows are expected to support the growth.
Read the original →The South Korean government will share its AI healthcare strategy with WHO Western Pacific member states. It will also discuss regional cooperation to supply medicines and vaccines stably during health crises. The 77th WHO Western Pacific meeting is held in Manila, the Philippines, from the 19th to the 22nd.
Read the original →