Segmentation method, device and storage medium of endoscopic fluorescence image

Abstract

The invention provides a partitioning method and device for an endoscope fluorescence image and a storage medium. The method comprises the steps that a convolutional neural network is acquired, wherein the convolutional neural network is obtained in the mode that a part of image is randomly selected from a collected fluorescence image to serve as a training set, a target image of interest is partitioned from the selected image, a tag is added to the target image, and an image bock with a pixel point exceeding a set threshold being the center in the training set and a tag of the image block areinput into a neural network for training; and the to-be-processed endoscope fluorescence image is acquired and input into the convolutional neural network, and an image partitioning result is obtained. According to the method, the convolutional neural network is utilized to convert a partitioning problem of the optical operation navigation endoscope fluorescence image into a dichotomy problem ofimage pixel points; and through continuous training and optimization of a whole system, inhomogeneity of the optical operation navigation endoscope fluorescence image is effectively overcome, and thetarget of interest in the image is accurately extracted.

Description

technical field [0001] The invention relates to the field of medical molecular imaging, in particular to an endoscope fluorescence image segmentation method, device and storage medium. Background technique [0002] Medical image processing and analysis is an interdisciplinary subject that has emerged in recent years, making medical images gradually develop from film to digital. The traditional medical imaging technology only obtains the image data of a certain section of the human body, and then the doctor makes a diagnosis through the film or observes it on the monitor. However, what doctors can observe is only two-dimensional images, and they can only observe the images in a fixed way, and the obtained diagnosis results are subject to empirical judgment, which largely depends on the doctor's clinical experience. Through graphic image technology, the image data can be arbitrarily enlarged, reduced, rotated, three-dimensional reconstruction and other processing, so that doc...

AI Technical Summary

Problems solved by technology

In addition, human anatomy is very complex and varies from person to person, which brings great difficulties to medical image segmentation

Images