Comparison of the Diagnostic Performance of Different Artificial Intelligence Assisted Endocytoscopy for Colorectal Lesions

Comparison of the Diagnostic Performance of Different Artificial Intelligence Assisted Endocytoscopy for Colorectal Lesions

Colorectal cancer (colorectal cancer, CRC) is the third most common malignant tumor globally and the second leading cause of cancer-related deaths. Colonoscopy is considered the preferred method for screening colorectal cancer; early detection and removal of colorectal neoplasms can significantly reduce the incidence and mortality of colorectal cancer. To improve the diagnostic accuracy of endoscopy in colorectal lesions, many endoscopic techniques have been applied clinically, such as image-enhanced endoscopy, including narrow band imaging (narrow-band imaging, NBI), magnifying endoscopy, chromoendoscopy, confocal laser endoscopy, and endocytoscopy (EC). However, with the increasing number of endoscopic resections, the costs associated with the pathological diagnosis of resected specimens have risen year by year. In clinical practice, some non-neoplastic colorectal lesions may not require resection, so it is important to differentiate the nature of lesions during colonoscopy.

Endocytoscopy is an ultra-high magnification endoscope that, when combined with chemical staining and narrowband imaging techniques, allows endoscopists to observe the nuclear morphology of colorectal lesions, the shape of glands, and the morphology of microvessels with the naked eye, thus avoiding pathological examination and achieving the goal of real-time biopsy in vivo. However, the accuracy of endocytoscopy images requires extensive experience accumulation to improve judgment, and there is a certain degree of subjectivity and error in the process of endoscopists making judgments. Therefore, to address this issue, clinical applications have proposed using artificial intelligence (AI) for computer-aided diagnosis. Currently, Japan has developed an endoscopic cytology auxiliary diagnostic system-EndoBRAIN, based on the Japanese population, which uses support vector machines to build model. The investigator's center has developed a deep learning-based endoscopic cytology AI auxiliary diagnostic system for Chinese populations to assist in determining the nature of colorectal lesions. There is currently a lack of comparative studies on the diagnostic performance of these two systems, so the investigator aim to conduct a clinical study to compare and analyze the differences between the two AI auxiliary diagnostic systems.

Inclusion Criteria:

• colorectal lesions

Exclusion Criteria:

• lesions lacking high-quality images;
• Inflammatory bowel disease, familial adenomatous polyposis and other special diseases;
• submucosal tumors;
• Pathological diagnosis of Peutz-Jeghers polyps, juvenile polyps, lymphoma and other pathological types.