Efficient medical image marking and learning system

Abstract

According to the method in the field of medical image processing, an efficient medical image marking and learning system is realized. According to the system, an unsupervised comparative learning pre-training module introduces an unsupervised comparative learning method in a pre-training process based on an MoCo algorithm; the training module based on active learning and comparative learning obtains a pre-trained model and an actual target task data set; and further adjusts the model to adapt to the actual demand by discovering the most valuable sample in the unlabeled samples to obtain the label and providing the difficult case mining of higher weight for the sample classified by mistake and introducing the supervised contrast learning method of labeled data; and a manual annotation and model training interaction control module provides an annotation interaction interface and a model training control module. Through the system of the architecture, a universal training framework is realized.

Description

technical field [0001] The invention relates to the field of medical image processing, in particular to an efficient medical image labeling and learning system. Background technique [0002] At present, various medical images are widely used clinically, such as X-rays, CT images, and MRIs for human imaging, as well as pathological images as a means of analyzing diseased tissues. [0003] Due to the characteristics of fast, simple and popular CT images, the imaging performance based on lung CT is taken as an important basis in the diagnosis of new coronary pneumonia and other lung diseases. CT images play an important role in the diagnosis of various lung diseases. The value is self-evident; due to its characteristic of relying on hydrogen atom resonance imaging, MRI can clearly image the water-rich soft tissue represented by the brain. At the same time, due to the occlusion of the skull, it is difficult for CT to image the brain. MRI examination is the most common method; w...

AI Technical Summary

Problems solved by technology

[0006] The main defect of the current deep learning medical imaging diagnosis method is that whether it starts training directly for the target application scenario from a randomly initialized model, or further fine-tunes and adapts the pre-trained model that has been trained for other related application scenarios. Matching to the current target application scenario requires a large amount of finely labeled data, and the acquisition of data labeling is very expensive, time-consuming and labor-intensive; in addition, the pre-training mentioned above - fine-tuned migration learning And in the training process of training-test-practical application, there may be category imbalance problems, as well as the differences between pre-training data sets, target task training data sets, and actual application scenarios, which will affect the final use effect

Images