Method and system for evaluating the severity of ulcerative colitis based on deep learning

Abstract

The invention provides an ulcerative colitis severity assessment method and system based on deep learning. The method comprises the following steps: marking a Mayo endoscopic score and an ulcerative colitis endoscopic activity index score on a white light colonoscope image to form a sample set; constructing an ulcerative colitis severity evaluation model, and performing training by using the marked white light colonoscope image sample in the sample set, wherein the ulcerative colitis severity evaluation model is an SPPNet network with a characteristic pyramid pooling function, and the output of the ulcerative colitis severity evaluation model comprises four softmax functions; receiving a white light colonoscope image in real time; and outputting score prediction results of Mayo endoscopicscores, vascular typing, spontaneous hemorrhage and erosion ulcer characteristics by utilizing the ulcerative colitis severity degree evaluation model, and accumulating the score prediction results ofthe vascular typing, spontaneous hemorrhage and erosion ulcer characteristics to obtain an ulcerative colitis endoscopic activity index score. The method can be used for automatically grading inflammation, can meet the grading requirements of two clinical most common UC severity degrees at the same time, and is high in result accuracy and good in repeatability.

Description

Technical field [0001] The invention belongs to the field of ulcerative colitis image processing, and in particular relates to a method and system for evaluating the severity of ulcerative colitis based on deep learning. Background technique [0002] The statements in this section merely provide background information related to the present invention, and do not necessarily constitute prior art. [0003] Ulcerative colitis (UC) is a chronic non-specific inflammation with recurrent abdominal pain, diarrhea, mucus, pus, blood, and stool as the main clinical manifestations. The lesions are often distributed continuously and can involve the rectum, sigmoid colon, and multiple intestinal segments of the entire colon. , Critically ill patients may have symptoms of systemic infection and poisoning, and even life-threatening. Typical UC endoscopy shows continuous, diffusely distributed erosion ulcers, mucosal hyperemia and brittle texture, often accompanied by spontaneous bleeding and att...

AI Technical Summary

Problems solved by technology

Due to the lack of a unified and objective scoring standard, the existing artificial UC endoscopic severity score has the following inherent defects: First, there is a significant difference in the scoring accuracy between experienced doctors and inexperienced doctors

Existing scoring is mainly based on the subjective judgment of the operating physician on the severity of lesion characteristics, and it is often difficult for inexperienced junior physicians to make accurate judgments, especially for mild to moderate lesions with unclear boundaries, resulting in inaccurate scoring

Secondly, it is difficult to guarantee the consistency of the same rater

The entire colonoscopy observation process lasts for 6-7 minutes. UC lesions usually involve multiple intestinal segments and are diffusely distributed. It is often difficult for the operator to clearly recall the characteristics of the most severe inflammation and make an accurate evaluation; Many, manual scoring is time-consuming and laborious, and it also greatly hinders the work efficiency of clinical colonoscopy

In addition, the heterogeneity among different raters is large and the reproducibility is poor

It is often difficult to obtain consistent scoring results among multiple evaluating physicians, and the previous scoring is often difficult to trace, which is not conducive to the long-term management of UC

Images