Method and system for automatically online learning and intelligently assisting in annotating medical images

Abstract

The invention provides a method for automatically online learning and intelligently assisting in annotating medical images, and relates to the technical field. The method comprises the steps: enablingan intelligent annotation system to be connected to a hospital PACS system, and carrying out the data screening, data cleaning and exporting; performing manual annotation, data division, model training, auxiliary recommendation annotation generation, model testing, index calculation, data set perfection and the like. According to the invention, the system is directly connected into a hospital system, and patient cases to be annotated can be automatically screened according to conditions; case images can be automatically cleaned, and the defect that in the prior art, needed data need to be manually collected and cleaned is overcome. According to the invention, the examination report and the pathological report can be automatically scanned to generate auxiliary labeling information. Meanwhile, recommended annotation can be intelligently prompted in the annotation process of a doctor, the training model is continuously optimized in the annotation process of the doctor, the online training of the model is achieved, and the method is different from the mode that in the prior art, model training is conducted after data set annotation is completed.

Description

technical field [0001] The invention relates to a method and system for automatic online learning and intelligently assisted labeling of medical images, belonging to the technical field of medical images. Background technique [0002] In 2002, China's population aged 65 and above accounted for more than 7%, marking China's entry into an aging society. In 2010, it entered a stage of deep aging. Since then, the dependency ratio of the elderly population has continued to increase, and in terms of the supply of medical resources, there is still a huge gap in the number of doctors per 10,000 people in my country compared with developed countries, which directly means the increase in medical and health needs. Therefore, the imbalance between supply and demand It is necessary to rely on technical means to improve the efficiency of doctors' diagnosis, which is why AI (artificial intelligence) medical care has been vigorously developed. [0003] The concept of artificial intelligence...

AI Technical Summary

Problems solved by technology

[0005] Artificial intelligence algorithms are usually supervised learning in the field of image recognition. The algorithm needs to train a large amount of artificially labeled data. However, compared with natural images, medical image data often exists in medical institutions and is difficult to obtain. Data labeling requires professional knowledge. Doctors, different doctors may have different views on the same medical image

For various reasons, although the AI ​​medical field has broad prospects, the development status is not optimistic. Only in the data field, the entire industry is facing the huge challenge of scarcity of medical data and difficult labeling

[0006] At present, for image annotation, the following related patents have been retrieved, "Interaction method and system for semi-automatic image annotation", "An artificial intelligence data annotation method and device", "An image intelligent annotation method based on YOLOv3 deep learning network", etc. , which all use the method of artificial intelligence to assist in data labeling, but none of them realize intelligence in data collection and production of data sets, and do not involve the problem of data set scale

Images