A method and system for automatic segmentation of cataract lens opacity regions
By constructing an elliptic curve model and a U-Net network, highlight interference is eliminated, low-turbidity regions are screened, and a lens center point is introduced, thus solving the problem of center positioning error caused by reflection and turbidity in cataract images and achieving accurate segmentation of the turbid lens region.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING TONGREN HOSPITAL AFFILIATED TO CAPITAL MEDICAL UNIV
- Filing Date
- 2026-03-19
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, automatic cataract image segmentation schemes suffer from overexposure of the center and occlusion of the edges due to reflective areas, which reduces the accuracy of the geometric center positioning of the lens. Furthermore, internal turbidity of the lens disrupts the uniformity and symmetry of grayscale distribution, leading to an aggravation of the center positioning error.
By constructing multiple elliptic curve models to eliminate interference from high-brightness areas, and combining Canny edge detection and RANSAC algorithm to fit elliptic curves, sector regions are divided, low-turbidity target regions are selected, and a U-Net network is constructed using the lens center point for segmentation. An attention mechanism is introduced to optimize feature extraction.
It effectively alleviates the interference of reflection and cloudiness on central positioning, improves the accuracy of lens geometric center positioning, reduces errors, and achieves precise segmentation of the clouded area of the lens in cataracts.
Smart Images
Figure CN122415641A_ABST