Method, device, equipment and storage medium for enhancing image features around target area
Feature enhancement of CBCT images through deep learning network model solves the problem of low contrast and high noise in the area around the target area, and significantly improves image quality and the accuracy of treatment plan.
Patent Information
- Application Number
- CN202411660368.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-04-01
AI Technical Summary
CBCT images have problems with low contrast, high noise and fuzzy details in the surrounding areas of the target area, which affects the accurate outline of tumors and the formulation of radiotherapy plans.
By acquiring the planned CT images and CBCT images, the surrounding images of the target area are extracted after registration, and the high-frequency information is removed using the CNN deep learning network model based on the pyramid structure, the predicted theoretical low-frequency information is output, and the training is repeated by adjusting the model parameters until the requirements are met. Finally, the model is used in the radiotherapy system to enhance the surrounding images of the target area.
The quality of CBCT images is significantly improved, especially the signal-to-noise ratio in the area around the target area, avoiding additional image features or artifacts, providing a more accurate radiotherapy plan, and improving the treatment effect and diagnostic accuracy.