Manifold-Prior Diverse Distillation for
Medical Anomaly Detection

Xijun Lu · Hongying Liu · Fanhua Shang · Yanming Hui · Liang Wan
Tianjin University · Medical School & College of Intelligence and Computing

CVPR 2026

Code | Paper | Project Page

PDD🛍️🛒is a novel framework for medical image anomaly detection, designed to tackle subtle, heterogeneous anomalies in complex anatomical structures. Addressing the limitations of traditional Grad-CAM methods on medical data, PDD innovatively unifies dual-teacher priors— frozen VMamba-Tiny (global context) and ResNet50 (local structure)—into a shared high-dimensional manifold.

Table of Contents

1. TODO
2. Citation

News:

• [2026/02/21] PDD is accepted to CVPR 2026 🔥. The code will be released before June.

TODO

• Release the code of PDD

Citation

If you find our code or paper useful, please cite

@inproceedings{lu2026pdd,
  title={PDD: Manifold-Prior Diverse Distillation for Medical Anomaly Detection},
  author={Lu, Xijun and Liu, Hongying and Shang, Fanhua and Hui, Yanming and Wan, Liang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={28534--28544},
  year={2026}
}