Improving deep learning-based polyp detection using feature extraction and data augmentation

In recent years, Colorectal Cancer (CRC) has been common reasons of lethal disease and cancer. However, colonoscopy can examine this disease, and the location of polyps and tumors can be detected. However, the early symptoms of CRC are not evident and specific, which is easy to be ignored by patients and doctors. As a result, the opportunity for early diagnosis and treatment was missed. This study aims to provide auxiliary detection to obtain accurate polyp diagnosis and assist clinicians in more precise detection. This paper proposes a novel polyp detection method through deep learning, which uses a fusion module combining feature extraction and data augmentation to enhance images. The Discrete Wavelet Transform (DWT) is applied to extract the texture features of polyps and strengthen the texture features that are not obvious in the polyp image. Then style-based GAN2 is used to enhance the image data, increase the image training data of YOLOv4, and let YOLOv4 learn more features of polyps. According to the experimental results, our method is better than state-of-the-art methods in polyp detection efficiency. In addition, because we have enhanced the image, the detection rate of small polyps is significantly improved.