A Survey on AI-Powered Solutions for Gastrointestinal Cancer: Endoscopy Image Segmentation and Classification

Gastrointestinal cancer is not like a lung cancer, breast cancer, skin cancer and so on. It refers to a collection of cancers that impact the entire human digestive system. It is a significant cause of cancer worldwide. Gastrointestinal (GI) cancer can be examined in two ways: endoscopy and colonoscopy. Several challenges occur during pre-processing, segmentation, and classification of gastrointestinal cancer endoscopy images. The challenges that occur when doing segmentation in endoscopic images can be small early lesions are missed due to poor image quality like blur, specular reflection, lighting and color visibility, low resolution image as well. Feature extraction and segmentation process are depends on the outcome of pre-processing images. Therefore, this study aims to enhance the GI endoscopy images using advanced preprocessing techniques such as color space transformation (HIS/YCbCr). The outcome of pre-processed image are used for segmentation. Therefore, segmentation problem may be addresses by improving data collection, leveraging domain adaptation techniques and the challenges in multi-classification are to classify wheather the cancer is early stage, local spread and metastasis. Finally, the multi-classification of gastrointestinal cancer may be addressed using various machine learning technique (ResNet, U-Net). This study offers a promising approach to enhancing the GI cancer patients and achieving better long term outcomes.