Description
The present invention discloses a medical image segmentation method based on wavelet-conditional latent diffusion, which relates to the technical field of medical image segmentation. The method comprises the following steps: acquiring an input medical image, performing a structure-preserving color enhancement operation during the training phase, and using the original input image during the inference phase; extracting conditional embeddings from the preprocessed image using a multi-scale wavelet encoder, wherein the encoder implements this through multi-layer discrete wavelet transform, high-frequency residual enhancement, inverse discrete wavelet transform, and downsampling concatenation; encoding a segmentation target corresponding to the input image using a pre-trained and parameter-frozen autoencoder to obtain a latent mask; inputting semantic prior conditions and the latent mask into a diffusion UNet, embedding a frequency-aware gated block in the skip connections of the diffusion UNet, and performing latent space denoising via a regularized iterative algorithm to obtain an optimized latent mask; and decoding the optimized latent mask using the decoder function of the autoencoder to output the final medical image segmentation result.| Period | 7 Nov 2025 |
|---|---|
| Held at | Monash Suzhou Research Institute |