Esophageal endoscope video frame sequence quality classification algorithm using space-time information of adjacent frames

Abstract

The invention belongs to the technical field of medical image processing, and particularly relates to an esophageal endoscope video frame sequence quality classification algorithm using adjacent frame space-time information. The algorithm comprises the following steps: constructing a convolutional neural network model for a video frame sequence prediction algorithm, including a content feature extraction sub-network and a motion feature extraction sub-network, referring to information of two features, and finally giving a video quality score of an intermediate frame through a full-connection sub-network; data collection and model training are carried out, and when a trained objective function is reduced to a certain acceptable threshold value, it can be considered that the network is converged; and finally, inputting three continuous frames of images into the trained network model to obtain quality classification of a middle frame. Experimental results show that the accuracy of quality classification of the algorithm exceeds 85%, and the algorithm has very high application value for diagnosis and quality control of clinical esophageal endoscopes.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to an esophageal endoscope video frame sequence quality classification algorithm. Background technique [0002] With the development of computer science, intelligent medical care has become a major technological innovation to improve the level of modern medical care. As the combination of artificial intelligence and new medical care, its advantages in various aspects are getting more and more recognition and attention. [0003] In the application scenario of endoscopic image analysis, when doctors perform endoscopic examination, they are often prone to encounter situations with low endoscopic image quality, such as the lens is pushed and pulled too fast, too close to the wall of the digestive tract, and blocked by blood foam. Low-quality pictures interfere with doctors' observation and diagnosis. Therefore, prompting the doctor to classify the quality ...

AI Technical Summary

Problems solved by technology

[0003] In the application scenario of endoscopic image analysis, when doctors perform endoscopic examination, they are often prone to encounter situations with low endoscopic image quality, such as the lens is pushed and pulled too fast, too close to the wall of the digestive tract, and blocked by blood foam. Low-quality pictures interfere with doctors' observation and diagnosis

Images