An LSTM-Based Time-Series Framework for Early Detection of Prostatitis Using Longitudinal Clinical Indicators

Abstract

Prostatitis, a prevalent urological condition among men, often remains undetected in its early stages due to subtle or overlapping symptoms. This research suggests a method based on deep learning for early detection, utilizing clinically generated synthetic data. A dataset comprising 120 patients with prostate-specific antigen (PSA) levels and other haematological parameters was pre-processed and modeled using a Long Short-Term Memory (LSTM) neural network. The model demonstrated outstanding classification performance, achieving an accuracy of 98.5%, precision of 97.6%, recall of 98.2%, F1 score of 97.9%, and an AUC of 0.992, confirming its robustness and high discriminative capability. Evaluation through The model’s reliability in differentiating between prostatitis and non-prostatitis cases was further confirmed by the confusion matrix and ROC curve. These results affirm the potential of LSTM-based models in supporting clinical diagnosis, particularly where access to real patient data is limited. The study contributes a scalable and ethical diagnostic framework adaptable to similar medical prediction tasks and recommends future validation using real clinical datasets such as electronic health records (EHRs) to enhance generalizability and real-world applicability.