Tuberculosis-based chest radiograph lesion detection method and system

By using anatomical partition constraints and tone mapping guided by diagnostic fingerprints, the problem of image distribution shift caused by differences in equipment models and scanning protocols during cross-central domain adaptation was solved, achieving more efficient tuberculosis lesion detection and meeting the needs of public health and disease control for cross-regional tuberculosis epidemic monitoring and large-scale screening.

CN122199530APending Publication Date: 2026-06-12NANJING UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANJING UNIV OF SCI & TECH
Filing Date
2026-05-12
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

In cross-central domain adaptation, image distribution shifts caused by differences in equipment models and scanning protocols may lead to the rewriting of fine-grained features related to tuberculosis diagnosis by generative adversarial networks. This can result in model performance degradation, artifacts, or the erasure of subtle lesion features, failing to meet the accuracy requirements of cross-regional tuberculosis epidemic monitoring and large-scale screening in public health and disease control.

⚗Method used

By using anatomical partition constraints and tone mapping guided by diagnostic fingerprints, the relative grayscale contrast and local grayscale ranking between tuberculous lesions and adjacent normal tissues are maintained. Combined with the updating of the model by characterization constraints, the cross-center adaptation capability is improved, and the risk of artifact generation and missed detection of subtle lesions is reduced.

🎯Benefits of technology

It effectively maintained the relative grayscale contrast and local grayscale ranking between tuberculous lesions and adjacent normal tissues, reduced the unexpected rewriting of key diagnostic information during cross-center adaptation, improved the model's ability to adapt to differences in cross-center image distribution, and enhanced the consistency and accuracy of screening results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122199530A_ABST
    Figure CN122199530A_ABST
Patent Text Reader

Abstract

The application discloses a chest radiograph lesion detection method and system based on tuberculosis, relates to the technical field of medical and health information science, and comprises the following steps: collecting chest images with tuberculosis annotation to obtain a source set, and collecting chest images without tuberculosis annotation to obtain a target set; the source set is cut to obtain an anatomical partition table, and the gray noise is counted to obtain a style codebook; the target set is smoothed to obtain a base tone image, the original image is subtracted from the base tone image to obtain a detail image, the gray energy is extracted to obtain a diagnosis fingerprint image; the base tone image is counted according to the anatomical partition table and compared with the style codebook to obtain a matching identifier; the base tone image is segmented, monotonically mapped and subjected to amplitude constraint according to the matching identifier to obtain a converted base tone image, and the converted base tone image is superimposed with the detail image to obtain a converted set; a screening model is trained by using the source set, a representation vector pair is obtained by inputting the converted set and the target set, and a cross-center tuberculosis screening model is updated according to a representation constraint quantity; the model is updated according to the representation constraint quantity, the adaptation capability of the model to cross-center image distribution differences is improved, and the risk of false image generation and missed detection of slight tuberculosis lesions is reduced.
Need to check novelty before this filing date? Find Prior Art