A method and system for distinguishing different scenes inside and outside a body cavity

Abstract

The invention relates to a method and a system for distinguishing different scenes inside and outside a body cavity, and relates to the technical field of artificial intelligence, comprising: S1, constructing a database for training a classification model for identifying the visual field inside and outside the body cavity, and constructing a classification model for identifying the visual field inside and outside the body cavity; S2, Access and analyze different surgical scenes in the surgical video pictures through the visual field recognition model inside and outside the body cavity, obtain the classification data of different types of surgical scenes, and output the classification data results; S3. Count the state of the surgical lens inside and outside the body cavity, and the lens placement during the operation Or the number of body cavity removals, the time point of the classification change, the duration and the proportion of the total operation time occupied. The invention distinguishes the state outside the body cavity and its duration from the invalid duration, which helps to distinguish the non-technical factors affecting the surgical efficiency in the surgical quality control; further improves the accuracy of distinguishing different scenes of the surgical video, so that it can be clinically usable level.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a method and system for distinguishing different scenes inside and outside a body cavity. Background technique [0002] The research of artificial intelligence in endoscopic minimally invasive surgery is increasing day by day. At present, there are related researches from instrument identification to stage, anatomical area, surgical action, evaluation of anatomical treatment and preliminary prediction of surgical duration; There are few people involved in scene detection. At the same time, in the traditional sense, different scenes inside and outside the body cavity in endoscopic minimally invasive surgery and video clips without anatomy, organs and surgical actions belong to the range of invalid time; The action clips are different, and the scenes outside the body cavity are not only obviously different from those in the body cavity, but also different from the n...

AI Technical Summary

Problems solved by technology

[0002] The research on artificial intelligence in minimally invasive laparoscopic surgery is increasing, from instrument identification to assessment of stage, anatomical region, surgical action, anatomical treatment, and preliminary prediction of operation time. Few people are involved in scene detection. At the same time, in the traditional sense, different scenes inside and outside the body cavity in laparoscopic minimally invasive surgery and video clips without anatomy, organs and surgical actions are also in the range of invalid duration; however, with no The action sequences are different, and the scenes outside the body cavity are not only obviously different from those inside the body cavity, but also in clinical practice. Unreasonable and poor anti-fog effect of the surgical lens, not necessarily the lack of technical proficiency of the surgical team

[0003] For different scenes inside and outside the body cavity in minimally invasive laparoscopic surgery, since the state and frequency of scene switching between inside and outside the body cavity during the operation are one of the criteria for evaluating the smoothness of the operation, the proficiency of skills, the distribution and effectiveness of surgical segments, And to a certain extent, it is related to the difficulty of the operation itself. The detection of different states inside and outside the body cavity provides a guarantee for the automatic implementation of the above functions. At the same time, using traditional computer vision algorithms to compare the scenes inside and outside the body cavity Simultaneous learning in the invalid stage will not only consume a large amount of data sets, but also have limited accuracy improvement, which cannot meet the needs of clinical work

Images