番茄社区

Welcome to MedAI Roundup, highlighting the latest news and research in healthcare-related artificial intelligence each month.

An AI model can distinguish between people with and without type 2 diabetes by analyzing 10-second voice clips recorded on smartphones, according to a study published in .

A simulation study published in the that used data from 130,000 critical care admissions showed predictive AI models may lose accuracy over time due to issues with re-training that can create negative feedback loops.

Large language models appear to be perpetuating . At the same time, humans are to incorporating AI-generated biased information into their own clinical decision-making. (Nature)

The FDA added more than 150 devices to of Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices, and expects to see a greater than 30% year-over-year increase in devices this year. The majority of devices (79%) authorized this year are in radiology, FDA said.

Yet only 10 of the 521 authorized devices on the FDA's AI/ML list as of 2022 were likely capable of informing critical decision-making on critically ill patients, according to .

Meanwhile, the WHO has outlined for regulating AI in healthcare, including transparency and documentation; risk management for cybersecurity threats; a commitment to data quality; and protection of patients' privacy and personal health data.

North Carolina-based Atrium Health said almost 85% of its physicians reported an with Nuance's DAX Copilot program.

Microsoft, which owns Nuance, unveiled new generative AI products for healthcare organizations at the HLTH conference in Las Vegas. Among these are new capabilities for its cloud service Azure, which will allow clinicians to more easily from a patient's electronic health record. (Healthcare Dive)

Google also shared its latest updates at HLTH, including new features for its Vertex AI Search, which is designed to give healthcare organizations high-quality, medically-focused search capabilities.

Comment