Medical image focus cross-domain detection method based on adversarial learning and adaptive analysis

Abstract

The invention discloses a medical image focus cross-domain detection method based on adversarial learning and adaptive analysis, and belongs to the field of medical image processing. Aiming at the practical problems that data with labels in medical images are few and difficult to obtain, sample distribution differences of different databases are large, generalization performance is poor during cross-database detection and the like, the method comprises the following steps: A, introducing adversarial learning into a deep learning focus detection framework to construct an unsupervised domain adaptive focus detection model; b, performing local adaptability analysis and feature selection; and C, carrying out global adaptability analysis and image selection. According to the invention, existing data with labels can be effectively utilized, the performance of cross-domain lesion detection can be effectively improved, the accuracy rate and the recall rate during cross-database testing can be improved, the generalization performance during medical image lesion detection through a deep learning model can be improved, the accuracy rate and the recall rate during cross-database testing can be improved, and the actual application requirement can be met.

Description

technical field [0001] The invention belongs to the field of medical image processing, and in particular relates to a method for cross-domain detection of medical image lesions based on adversarial learning and adaptive analysis. Background technique [0002] In recent years, thanks to the substantial improvement of computing power and the rapid development of machine learning, the use of computer-aided lesion detection in medical images has become more and more common. The use of computer intelligent assisted diagnosis system can greatly improve the diagnosis speed of lesions under the premise of ensuring a certain accuracy rate, relieve the pressure of doctors on diagnosis, and improve the detection efficiency of diseases. At present, the algorithm design of most intelligent auxiliary diagnosis systems is based on deep learning. Deep learning uses its powerful nonlinear fitting ability to abstract advanced semantic features of images to achieve specific tasks. [0003] Th...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the practical problems such as few and difficult to obtain tagged data in medical images, large differences in sample distribution in different databases, and poor generalization performance in cross-database de

Images