AI could significantly improve the value of patient recalls following mammography, but so far radiologists seem reluctant to rely on computer-aided readings. Radiologists tend to trust their own judgment – and that of their colleagues – in mammogram readings far more than AI-based diagnostics, even when the AI is much more accurate, a prospective trial analysis published in Radiology by Karolinska Institutet researchers found.

Researchers from the University of Cincinnati filed for protection of an electrochemical aptamer-based biosensor technology with improved sensitivity and longevity, which has the potential for monitoring several biomarkers over sustained periods.

The EU’s Medical Devices Coordination Group (MDCG) issued another revision of its guidance for risk classification for in vitro diagnostics — the fourth such rewrite of a guidance that came out in 2020.

Ignite Biomedical Inc. and Liquid Biosciences have announced the discovery of biomarkers to diagnose substance use disorder (SUD). The set of biomarkers includes some that have not previously been associated with addiction and spans all 6 SUDs (opioid, cocaine, methamphetamine, alcohol, nicotine and cannabis use disorders).

Researchers from George Washington University filed for protection of a system and method that leverages generative models, specifically Variational Autoencoders, to conduct functional connectivity analysis from functional magnetic resonance imaging scans.

Diopsys Inc. agreed to pay $14.25 million to settle allegations it violated the False Claims Act (FCA) in encouraging the unnecessary use of the company’s Nova device for retinal testing.

The financial collapse of direct-to-consumer genetic testing pioneer 23andme Holding Co. is already one of the more notable developments of 2025, but the company’s customers have less to worry about with regard to their data.

The U.S. FDA famously lost a lawsuit over its final rule for regulation of lab-developed tests (LDTs), but Jeff Shapiro of King & Spalding told BioWorld that the implications of this decision go beyond FDA regulation of clinical lab operations.

When every minute matters, quickly determining which patients in the emergency department need urgent care for myocardial infarction can save lives. Researchers at the University Hospital Münster in Münster, Germany, developed a deep learning model that can detect features on electrocardiograms that more accurately identifies which patients require urgent revascularization than clinicians and provides results faster than high-sensitivity troponin lab tests.

The use of an AI tool could speed up the time it takes to do pregnancy scans by almost half and it is still able to identify any abnormalities in the fetus, researchers found. Results from a study showed the AI tool used to assist with 20-week pregnancy scans reduced the scan length by more than 40% and still maintained the same accuracy and reliability of diagnoses.