Polyp Detection Using Deep Learning Technique

TL;DR: This study employs deep learning techniques, specifically folding networks and support vector machines, to improve polyp detection in colonoscopic images, achieving promising performance in accuracy, recall, and F1 score, with potential for real-time clinical applications.

Abstract: Colon cancer (CRC) is one of the most common causes of cancer-related mortality worldwide. Detection and removal of polyps, a pioneer of CRC, is extremely important for improving prediction. With advances in deep learning, significant advances have been made in the accuracy and efficiency of polyp detection. Especially in medical imaging. This work attempts to use folding networks (CNNS), in particular folding networks (CNNS), in particular, to improve detection of polyps in colonoscopic images. A Kvasir-SEG dataset containing colonoscopy images is used and preprocessing is performed, including size, normalization and enhancement to improve model output. This work uses both CNNs for end-to-end learning and support vector machines (SVMs) with features derived from the pre-formed VGG16 model. The model has been evaluated for accuracy, recall, F1 score, and ROC-AUC, and the results have promising performance in polyp detection. Optimization of the model by adjusting and verifying hyperparameters for invisible data suggests survival of this method for real-time recognition of polyps in clinical settings. Future research includes increasing data records, improving models, and investigating applications in a broader healthcare environment