Cross-species medical image classification method based on domain self-adaption

Abstract

The invention discloses a cross-species medical image classification method based on domain self-adaption, and the method comprises the steps of extracting domain invariant features and domain difference features between NAFLD image data of a rat and a human through a domain self-adaption algorithm; subjecting the uncalibrated human liver image data to category discrimination through a classifier trained by the calibrated rat liver image data by means of feature alignment and domain adversarial; building constraint based on distribution consistency of a source domain and a target domain and domain invariance constraint based on conditional adversarial learning, ensuring the mobility of a classifier by controlling the uncertainty of a classifier prediction result, and meanwhile, improving the discriminability of the classifier by calculating cross covariance between features and the classifier prediction result. Therefore, the pathological classifier suitable for the target field can be obtained on the medical image data with large sample distribution difference between the source field and the target field, and transfer learning of cross-species data is realized.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a domain-adaptive-based cross-species medical image classification method. Background technique [0002] Under the support and drive of national policies, the field of intelligence led by smart medical care has gradually attracted the attention of the society in recent years. Smart healthcare also utilizes advanced technology to realize the informatization interaction between patients and medical staff, medical institutions, and medical equipment. Among them, medical image intelligent analysis technology is an indispensable technology in intelligent medical treatment. It is widely used in medical image classification tasks, assists doctors in making decisions, further improves diagnostic efficiency, and reduces the pressure on patients to seek medical treatment. However, analysis techniques rely on large-scale standard calibration data, and these data are insufficient du...

AI Technical Summary

Problems solved by technology

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: first, the data of rats and humans are respectively input into the CNN-based multi-scale feature extraction network, thereby solving the problem that effective features cannot be extracted due to the different image sizes of rats and humans

Images