Loading...

Real-time Automatic Detection and Classification of Colorectal Polyps during Colonoscopy using Interpretable Artificial Intelligence

Pourmand, Amir | 2023

62 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56187 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza
  7. Abstract:
  8. Cancer is the leading cause of death worldwide, and colorectal cancer is the second leading cause of death in women and the third in men. On the other hand, colon polyps can cause colorectal cancer. Therefore, early detection of polyps is of great importance. In recent years, many methods have been proposed for polyp detection using deep learning with high accuracy, but most of them have problems with speed, accuracy, or interpretability. Speed is important because colonoscopy should be performed as quickly and promptly as possible, and in many cases, it is not possible to repeat the colonoscopy. In addition, many of them only address the issue of polyp detection, while from a medical point of view, polyp classification and determining whether they are cancerous or not, is also important. Interpretability is another aspect, and until the writing of this thesis, interpretability has not been considered in the field of object detection algorithms. In fact, this research is the first study to propose a real-time interpretable model for object detection. In this research, we will present a pipeline for real-time classification and detection of polyps. Our proposed pipeline consists of an object detection algorithm (YOLOv5) for real-time polyp detection, followed by an interpretability module for interpreting the output. We will also use a sampler and a semi-supervised method to improve the overall performance of the algorithm. Finally, we propose two methods for making the YOLO algorithm interpretable and test both methods on the dataset for evaluation. With this, it will be possible to use a large number of derivative-based interpretability algorithms for YOLO. Overall, our proposed method in this research achieves an mAP of 90 percent and a speed of 100 frames per second (or 10 milliseconds delay) for the detection process and interprets the results at the same time
  9. Keywords:
  10. Colon Cancer ; Interpretability ; You Only Look Once (YOLOv7) ; Colonoscopy ; Real-Time Detection ; Polyp Classification ; Polyp Detection

 Digital Object List

 Bookmark