Construction Of A Model To Predict The Risk Of Recurrence Of Non-muscle Invasive Bladder Cancer Using Pathological Images Combined With Deep Learning

ObjectiveWe aimed to build a deep learning-based pathological model to predict the early recurrence of non-muscle-infiltrating bladder cancer(NMIBC)in this work.MethodsTwo cohorts of patients were collected,including 147 cases from Xuzhou Central Hospital as a training cohort and 63 cases from Suqian Affiliated Hospital of Xuzhou Medical University as a test cohort.Based on two consecutive phases of patch level prediction and WSI level prediction,we built a patronymics model,with the initial model developed in the training cohort and subjected to transfer learning,and then the test cohort was validated for generalization.The features extracted from the visualization model were used for model interpretation.ResultsAfter migration learning,the area under the receiver operating characteristic curve(AUC)for the deep learning-based patronymics model in the test cohort was 0.860(95% CI: 0.752-0.969),with good agreement between the migration training cohort and the test cohort in predicting recurrence,and the predicted values matched well with the observed values,with P-values of 0.667766 and 0.140233 for the Hosmer-Lemeshow test,respectively.The good clinical application was observed using a decision curve analysis method.ConclusionWe developed a deep learning-based pathomics model that demonstrated good sensitivity and accuracy,in which ten pathological features that play an important role are visualized and can be used to facilitate personalized management of patients with non-muscle invasive bladder cancer,avoiding ineffective or unnecessary treatment for the benefit of patients.