一种基于双层超声导管的超声图像处理方法及装置、介质
By acquiring image data through a double-layer ultrasound catheter, and combining variational autoencoder and Gaussian mixture model to generate a synthetic image dataset, a multiplicative speckle noise denoising model is trained, which solves the problem of insufficient ultrasound image denoising accuracy in existing technologies and achieves more efficient noise suppression and image quality improvement.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing ultrasound image denoising models struggle to effectively remove multiplicative noise in the absence of clean image labels, resulting in low denoising accuracy and impacting image quality and the accuracy of computer-aided diagnosis.
Image data is acquired using a double-layer ultrasound catheter. A synthetic image dataset is generated by combining a variational autoencoder and a Gaussian mixture model. A multiplicative speckle noise denoising model is trained, and the model is optimized using self-supervised and perceptual loss functions to gradually improve the denoising accuracy.
It effectively eliminates the distribution deviation between synthetic noise and real noise, provides a clear optimization direction and gradient backpropagation path, improves the denoising accuracy of ultrasound images and the adaptability of the model, and enhances image quality and diagnostic accuracy.
Smart Images
Figure CN122415380A_ABST