A method and system for detecting privacy masking of lesioned parts in clinical skin images

Through a semi-supervised learning method, a model that can accurately detect private parts and lesion sites in clinical skin images is trained, solving the problem of difficulty in extracting accurate information in the prior art, and achieving efficient privacy protection and lesion detection.

CN118379259BActive Publication Date: 2025-03-25XIANGYA HOSPITAL CENT SOUTH UNIV +1
2 Cites 0 Cited by

Patent Information

Application Number
CN202410493036.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-23
Publication Date
2025-03-25
Estimated Expiration
2044-04-23

AI Technical Summary

Technical Problem

Existing models are difficult to accurately extract information about private parts and lesions in complex clinical skin images, and it is difficult to avoid private parts covering the lesion, resulting in missed diagnosis.

Method used

A semi-supervised learning method is adopted to learn pre-trained detection models through momentum comparison, combining supervised and unsupervised data, a model that can accurately detect lesions and private parts is trained, and a lesion-first mask generation step is designed to ensure that lesion information is not obscured by private parts.

Benefits of technology

It realizes accurate extraction of private parts and lesion information in complex clinical skin images, avoids errors in the private parts covering the lesion, and improves the accuracy and efficiency of diagnosis.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention provides a method and system for detecting privacy masking of lesion parts in clinical skin images. The method comprises the steps of preprocessing, grouping and labeling lesion parts and private parts, pretraining a detection model, training a lesion and private part detection model based on semi-supervised learning, testing a lesion and private part detection model based on semi-supervised learning, inputting a clinical skin image acquired in real time to acquire a privacy and lesion part mask, synthesizing a segmentation mask of the private part with the clinical skin image to obtain a privacy-removed clinical skin image, and the like. The present invention trains an automated clinical skin image privacy removal model using clinical skin images, and can automatically realize the covering of private parts and the extraction of lesion areas.
Need to check novelty before this filing date? Find Prior Art