Latent Diffusion Autoencoders: Toward Efficient and Meaningful Unsupervised Representation Learning in Medical Imaging

Name: Latent Diffusion Autoencoders: Toward Efficient and Meaningful Unsupervised Representation Learning in Medical Imaging
Uploaded: 2026-10-02T10:33:47.840285+00:00
Duration: 604 s
Description: LDAE introduces a novel diffusion-based framework for learning efficient representations of 3D brain MR images for Alzheimer's disease (AD). It uses a compressed latent space and pre-trained diffusion models to generate high-fidelity reconstructions and extract clinically meaningful semantic codes, enabling attribute manipulation of scans.

LDAE introduces a novel diffusion-based framework for learning efficient representations of 3D brain MR images for Alzheimer's disease (AD). It uses a compressed latent space and pre-trained diffusion models to generate high-fidelity reconstructions and extract clinically meaningful semantic codes, enabling attribute manipulation of scans.

Open the full episode