An AI healthcare company claims its software can identify the extent of prostate cancer with greater accuracy than human doctors.
Avenda Health published a study last month that involved ten physicians who each evaluated 50 prostate cancer cases. The company’s Unfold AI software demonstrated an accuracy rate of 84.7% in detecting cancer, whereas the doctors assessing the cases manually had accuracy rates ranging from 67.2% to 75.9%.
Conducted in collaboration with UCLA Health and featured in the Journal of Urology, the study also revealed that AI-assisted cancer contouring predictions of cancer size were 45 times more accurate and reliable compared to traditional methods.
Shyam Natarajan, an assistant adjunct professor of urology, surgery, and bioengineering at UCLA and the study’s senior author, noted that the integration of AI assistance not only enhanced the precision of doctors’ assessments but also increased consistency in their evaluations.
Doctors typically rely on MRIs to determine tumor size, yet some tumors are described as “MRI-invisible,” explained Dr. Wayne Brisbane, an assistant professor of urology at UCLA’s David Geffen School of Medicine. AI technology can fill the gaps where MRIs fall short.
“Overall, the use of AI in cancer treatment could lead to more effective and personalized care for patients, providing treatments that are better tailored to their individual requirements and more successful in combating the disease,” Brisbane stated, adding that AI can extend beyond human capacity.
Avenda Health’s CEO, Dr. Shyam Natarajan, expressed that it is encouraging for doctors to witness such innovations being validated through research and acknowledged by the American Medical Association (AMA).
According to the American Cancer Society, approximately 1 in 8 men in the U.S. will be diagnosed with prostate cancer in their lifetime, and 1 in 44 will succumb to the disease. It is projected that there will be 299,010 new prostate cancer cases in the U.S. this year, with 35,250 fatalities attributed to the condition.