Recognition method and system based on dermatopathology images

Abstract

The embodiment of the invention discloses a recognition method and system based on dermatopathology images. A multi-task learning deep learning technology is adopted, three tasks of tissue area segment, pathological feature extraction and disease sorting recognition are sequentially completed, three deep learning models of tissue area segment model, pathological feature extraction model and disease sorting recognition model are correspondingly established, systematic modeling is conducted following a clinical structural analysis method of a doctor, and controllable key points and indicators ofartificial intelligent in pathological diagnosis are achieved; the tissue area segment and pathological feature extraction models are characterized in that training and learning are conducted takingpixel as the smallest unit, the scale of the needed sample number is controllable, the restriction of a large number of samples is broken through, the ultimate learning result is used for processing the pathological recognition of various kinds of inflammatory dermatosis, the dermatosis diagnosis and recognition efficiency in China is greatly improved, the automated technology level is high, and alarge amount of labor force and material resource are saved.

Description

technical field [0001] The invention relates to the technical field of medical image recognition, in particular to a recognition method and system based on skin pathology images. Background technique [0002] Skin pathological examination is the “gold standard” for doctors to judge skin characteristics in clinical practice. Eosin (H-E) staining is not directly implemented on living human or animal bodies), and the microscopic observation of skin tissue and cells under a microscope is carried out to confirm the type of disease. At present, the number of patients with inflammatory skin diseases occupies a major part of clinical dermatology in my country, and the number of pathological examinations is increasing year by year. Since inflammatory skin diseases are all infiltrating cells in the lesion tissue are inflammatory cells, which contain a wide variety, including not only the global tissue structure information, but also local differential changes, the pattern is complex ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the embodiments of the present invention is to provide a recognition method and system based on skin pathological images, which can be applied to pathological images of inflammatory skin diseases with complex pathological features (not directly implemented on living human or animal bodies), and overcome the current Low level of automation and low efficiency in the identification of inflammatory skin characteristics can assist doctors to judge and improve the screening rate and accuracy based on skin pathology images

Images