AI-driven prediction of prostate cancer risk: A comparative analysis with C the Signs in the Mayo Clinic data platform.

Abstract

371 Background: Prostate cancer is the most common malignancy in men globally and remains among the top five leading causes of cancer-related deaths. Emerging evidence suggests that early identification of symptoms can detect localized disease thus enabling opportunities for curative intent treatments. This study evaluates the use of the AI-powered prediction model, C the Signs, to passively screen for prostate cancer by leveraging data from electronic medical records (EMRs), offering a novel pathway for identifying high-risk individuals. Methods: A retrospective analysis was conducted using the Mayo Data Platform, encompassing 418,477 male patient records, of which 16,835 were diagnosed with prostate cancer. C the Signs identified patients at risk based solely on their EMR data, utilizing AI-driven pattern recognition. Sensitivity and specificity analyses were performed to assess the prediction model’s accuracy. Additionally, we examined the time-to-diagnosis advantage for patients flagged by the model compared to traditional physician clinical diagnoses. Results: C the Signs demonstrated a sensitivity of 83.3% and a specificity of 52.5% for identifying patients at risk of prostate cancer. Notably, 31.8% of prostate cancer cases were identified at risk up to five years earlier by the model compared to traditional physician clinical diagnosis. Conclusions: The integration of AI-driven prediction models like C the Signs into prostate cancer screening pathways provides an opportunity to enhance early detection, particularly in symptomatic individuals. Compared to prostate-specific antigen (PSA) testing, the model achieved equivalent sensitivity and 38% higher specificity, positioning it as a valuable companion tool for improving patient outcomes.

Ähnliche Arbeiten