Endoscopic Bladder Tissue Classification Through Fusion of Handcrafted and Deep Features

The proper identification of tissue types during endoscopy may greatly affect both the patient’s care and the treatment options that are made, making the categorization of endoscopic bladder tissue a crucial responsibility in the field of urology. Combining handmade characteristics with deep features is the innovative strategy that is presented in this study to improve the accuracy of bladder tissue categorization. To extract handmade characteristics from endoscopic images, the proposed technique makes use of textural feature extraction and the Bag-of-Visual-Word (BoVW) paradigm. This research proposes a hybrid convolutional neural network (CNN) model as a means of further enhancing classification performance. This model takes handcrafted features in addition to CNN-based features as input. This hybrid model is intended to make use of the complementing qualities that both kinds of features possess to provide a depiction of bladder tissue that is more accurate and distinct. During the experimental assessment, the proposed hybrid CNN model is evaluated in comparison to fundamental machine learning approaches as well as alternative deep learning architectures. The findings show that the suggested model can achieve an amazing accuracy rate of 95%, which is higher than the performance of other techniques. This accomplishment demonstrates the ability to integrate handmade cues with deep features for endoscopic bladder tissue categorization. This work has major implications for increasing diagnosis accuracy and the outcomes for patients.