An Artificial Intelligence-based Multimodality-Integration for Precise Prediction of Prostate Cancer Lymph Node Metastasis: A Retrospective Two-center Study

Abstract BackgroundAccurate identification of pelvic lymph node metastasis (PLNM) in patients with prostate cancer (PCa) is crucial for determining appropriate treatment options. However, there is no clear consensus on the integration of clinicopathological and imaging findings available to predict PLNM. Therefore, we built a Prostate Cancer Risk (PRISK) tool using an artificial intelligence-based multimodality-integration to obtain a precisely informed decision about whether to perform extended pelvic lymph node dissection (ePLND). MethodsPRISK provides a novel precise risk assessment tool to reduce unnecessary ePLNDs while controlling PLNM missing rate. It was developed in 280 patients and verified in 71 patients internally and in 50 patients externally by integrating a set of radiologists’ interpretations, clinicopathological factors and newly refined imaging indicators from MR images with radiomics machine learning and deep learning algorithms. Its clinical applicability was compared with Briganti and Memorial Sloan Kettering Cancer Center (MSKCC) nomograms.ResultsPRISK yielded the best diagnostic performance with areas under the receiver operating characteristic curve (AUC) of 0.932 (95% CI, 0.895-0.958), 0.924 (95% CI, 0.837-0.974) and 0.758 (95% CI, 0.616-0.868) in the training/validation, internal test and external test cohorts. If the No. of ePLNDs missed for risk assessment is controlled at < 2%, PRISK can provide both higher No. of ePLNDs spared (PRISK 59.6% vs MSKCC 44.9% vs Briganti 37.7%) and lower No. of false-positives (PRISK 59.3% vs MSKCC 70.1% and Briganti 72.7%) as compared with MSKCC and Briganti score. In follow-up, patients stratified by PRISK showed significantly different biochemical recurrence rate after surgery.ConclusionsPRISK offers a noninvasive clinical biomarker to predict PLNM for patients with PCa. It shows improved accuracy of diagnosis support and reduced overtreatment burdens for patients with findings suggested of PCa.