NIEHS intramural researchers and their collaborators have revealed the potential of AI and machine learning for diagnosing and classifying PCOS, which is a common hormonal disorder in women aged between 15 and 45. Their study systematically reviewed previous AI research and resulted with remarkable findings.
Dr. Janet Hall, an NIEHS senior investigator and co-author of the study, highlighted the significance of the research. He explained that their study aimed to see if the two technologies could assist in identifying individuals at risk for PCOS due to the problems of under- and misdiagnosed cases and the health issues that result. AI’s remarkable effectiveness in detecting PCOS surpassed their expectations.
PCOS, a condition linked to ovarian dysfunction and elevated testosterone, results in irregular periods, acne, excess hair growth and other health issues including an increased risk of type 2 diabetes. Dr. Skand Shekhar of NIEHS talked about the potential of AI and machine learning to improve PCOS diagnosis due to its diagnostic challenges.
PCOS diagnosis relies on criteria involving clinical features, lab indicators and radiological findings, but it often goes undetected due to symptom overlap with other disorders. AI and machine learning can help by efficiently processing diverse data such as electronic health records, to diagnose complex conditions like PCOS.
In a 25-year review (1997-2022) of AI and machine learning studies for PCOS identification, the researchers found 135 potentially eligible studies, with 31 included in the paper. These observational studies evaluated AI and machine learning for patient diagnosis, with half using ultrasound images. Study participants averaged 29 years old. Of the ten studies using standardized PCOS criteria, detection accuracy ranged from 80% to 90%.
Shekhar added that the key finding is the exceptional performance of AI and machine learning in detecting PCOS across various diagnostic and classification methods.
