Scientist Advises Caution When Using AI in Mammography

AI making mistakes

Examining breast-cancer tumors with artificial intelligence can enhance healthcare efficiency and results. However, specialists must proceed carefully, considering that identical technological jumps previously caused greater false-positive tests rates and over-treatment.

That is according to a new article in JAMA Health Forum co-written by Joann G. Elmore, MD, MPH, a scientist at the UCLA Jonsson Comprehensive Cancer Center, the Rosalinde and Arthur Gilbert Foundation Endowed Chair in Healthcare Delivery, and professor of medicine at the David Geffen School of Medicine at UCLA.

The JAMA Health Forum editorial stated that without a more robust strategy for the assessment and application of AI, provided the unabated acceptance of emerging systems in clinical practice. The sector is still struggling to learn from previous errors in mammography. The item, published online, was co-written with Christoph I. Lee, MD, MS, MBA, a professor of radiology at the University of Washington School of Medicine.

Overconfidence in technology

According to the authors, among those previous errors in mammography were adjunct computer-aided detection (CAD) devices, which proliferated in popularity in the field of breast cancer testing beginning more than twenty years earlier. The FDA validated CAD in 1998, and by 2016 beyond 92% of United States imaging centers were using the systems to interpret mammograms and fish for tumors. However, the proof revealed CAD did not enhance mammography precision. The authors wrote that CAD tools are associated with a rise in false-positive rates, resulting in over-diagnosis of ductal carcinoma in situ and unneeded analysis screening. Medicare stopped covering CAD in 2018, however at that time, the tools had reached greater than $400 million a year in unnecessary health expenses.

Elmore and Lee published that the early adoption of CAD is a premonitory sign of the wholehearted embrace of emerging systems before thoroughly comprehending their impact on patient results.

The specialists recommend numerous safeguards to prevent repeating previous errors, including tying Medicare compensation to enhance patient results and not just enhance technical performance in artificial settings.

Originally published by: medicalxpress.com