Determining the Necessity of Surgical Operation in Patients with Lower Urinary Tract Symptoms in the Gray Zone Using a Deep Learning-Based Model

Le résumé fourni par la source

Pressure-flow studies (PFS), considered the gold standard in the diagnosis of Lower Urinary Tract Symptoms (LUTS), often fail to provide a definitive diagnosis for "Gray Zone" patients where the Bladder Outlet Obstruction (BOO) index ranges between 20 and 40. In current clinical practice, this uncertainty is typically resolved through cystoscopy, an invasive procedure. This study aims to develop a novel, spectrogram-based deep learning approach capable of predicting the necessity of surgical operation in gray zone patients with high accuracy, eliminating the need for cystoscopy. The study utilized Detrusor Pressure (Pdet) and Urine Flow Rate (Qura) signals obtained from Fırat University Hospital, belonging to a total of 366 patients (183 requiring surgery, 183 not requiring surgery) within the gray zone range. These raw signals of varying temporal lengths were converted into spectrogram images using the Short-Time Fourier Transform (STFT) to preserve time-frequency components. For classification, a novel SR-Net (Stack Residual Network) architecture consisting of 104 layers based on the "Stack Residual" strategy was designed, and the model was trained on this balanced dataset. The developed SR-Net model achieved an overall accuracy of 77.04% in the testing phase. When performance metrics were analyzed, the model's sensitivity in detecting surgical necessity was found to be 0.761, and its specificity in distinguishing cases not requiring surgery was 0.761. This balanced performance across classes demonstrates that the model successfully learned both clinical conditions without bias. The results indicate that analyzing urodynamic signals via deep networks through spectrogram representations presents a strong and objective alternative to invasive methods in resolving diagnostic uncertainty in gray zone patients. The proposed system holds the potential to reduce unnecessary cystoscopy procedures by providing clinicians with robust decision support.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.