Multimodal Artificial Intelligence for Non-Invasive Colorectal Cancer Screening Based on Holographic Biology

The purpose of this study is to propose a non-invasive screening method by integrating holographic biological theory and artificial intelligence technology. A colorectal cancer risk assessment model will be constructed by analyzing multimodal data including facial features, tongue image characteristics and exhaled gas. The hypothesis of this study is that the model can attain both sensitivity and specificity of 80%.This study will enroll patients aged 18-80 years scheduled to undergo colonoscopy. All participants shall provide informed consent and sign the informed consent form. Patients will be excluded if they have severe cardiac, cerebral, pulmonary or renal dysfunction, or psychiatric disorders precluding colonoscopy, have a history of gastrointestinal surgery, or have taken bismuth agents or other staining medications. Based on estimates from existing literature, the training set will require tongue images, facial images, exhaled gas analysis results and colonoscopic diagnoses from 5000 patients, and the test set will require such data from 6000 patients.