Method for identifying lesions in real time based on digestive endoscopy

Abstract

The invention relates to a method for identifying lesions in real time based on a digestive endoscope. The method comprises the following steps: processing by an artificial intelligence algorithm to obtain a preliminary lesion result in a digestive tract; removing potential false positive misinformation in the preliminary lesion result by an online filtering algorithm to obtain an effective lesionresult; calculating a lesion ratio value of an effective lesion result in a period of time by a sliding window mechanism, wherein the average lesion suspected degree is calculated by an exponential moving average algorithm through the lesion ratio value; carrying out alarming and screenshot display of a lesion area based on the average lesion suspected degree. According to the method, a lesion result is identified based on an artificial intelligence algorithm; a lesion ratio value and an average lesion suspected degree algorithm are calculated through an online filtering algorithm and a sliding window, the lesion result is post-processed, and the real lesion is highlighted; the purposes of improving the accuracy, reducing the false alarm rate, improving the working efficiency of doctorsand greatly improving the user experience are achieved; the real-time performance of the system is not affected due to the fact that the algorithm carries out calculation through a CPU.

Description

technical field [0001] The invention belongs to a method for identifying lesions, in particular to a method for identifying lesions in real time based on digestive endoscopy. Background technique [0002] Artificial intelligence is being widely used in various fields of medical and health care. Among them, in the application of artificial intelligence algorithms for real-time lesions (for example, polyps, ulcers, cancer) identification of gastrointestinal endoscopic surgery videos, the current algorithms are generally The pictures in the video stream are processed frame by frame, analyzed, and the results are displayed. The analysis results are affected by various factors such as the amount of data trained by its artificial intelligence algorithm, the accuracy of data annotation, algorithm selection and optimization, etc. [0003] In the clinical environment, due to the different shapes of lesions in the digestive tract, folds of the intestinal wall similar to lesions, possi...

AI Technical Summary

Problems solved by technology

[0003] In the clinical environment, due to the different shapes of lesions in the digestive tract, folds of the intestinal wall similar to lesions, possible foreign bodies (feces, blisters), image quality (motion blur, lens out of focus, light spots, reflections), etc., artificial The accuracy of intelligent algorithms will decrease

[0004] The usual practice in medical applications is to improve the sensitivity of the algorithm and reduce false negatives through tuning, but at the same time it will also lead to a decrease in the specificity of the algorithm, and more false positives and false positives will be generated in the actual clinical examination process. Judging the non-lesion area as the lesion area, these false positives and false positives require doctors to consume extra attention to eliminate them, thus increasing the work intensity and workload of doctors

Images