Determining accuracy of diagnosis and management of common presenting semen analyses using artificial intelligence...

Over the past few years, artificial intelligence (AI) platforms have rapidly gained popularity within medicine. While AI has been applied in various subspecialties of urology, its role in evaluating male factor infertility has not been explored. The objective of this study was to evaluate the diagnostic accuracy of two commonly used AI programs, Google's "Bard" and Bing. This study aimed to assess each program's accuracy in correctly diagnosing a sample patient's semen analysis results and recommending appropriate next steps following diagnosis.

Each respective AI program was given a set of data which included semen volume, pH, concentration, and sperm motility as a percentage along with a command to list the three most likely diagnoses and the next steps the patient should take. The data sets ranged from entirely normal to abnormal with clearly obstructive and non-obstructive azoospermia, teratozoospermia, oligospermia, or asthenospermia. Study personnel determined the clinical diagnostic accuracy of both Bard's and Bing's semen analysis interpretations. No patient data was utilized for this study.

Bing resulted in only 29% accuracy of interpretation while 57% of results provided partially correct responses. First, second, and third, diagnoses provided resulted in 43%, 29% and 43% accuracy, respectively. Each analysis was 100% accurate in the next steps the patient should take and recommended discussing results with a physician 100% of the time. Bard was slightly more accurate regarding semen analysis with 50% accuracy. First, second, and third diagnoses provided resulted in 75%, 25%, and 25% accuracy, respectively. Bard had 75% accuracy regarding next steps but also had a 100% accuracy rating for recommending discussing results with a physician.

Overall, Bard was more accurate in providing correct analytical information regarding semen analysis (50% vs. 29%). Bard generated consistently accurate first diagnosis (75% vs. 43%). Bing resulted in increased accuracy regarding next steps (100% vs. 75%). Both programs recommended discussing semen analysis results with a physician. Overall, Bing and Bard are not capable of consistently providing patients with accurate analysis, diagnosis, or next steps when given a sample semen analysis. Specific training sets must be developed to provide with accurate interpretation of their urological results in a user-friendly format that can be further addressed with their physician.