Diffusion models with implicit guidance for medical anomaly detection.
In: (27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024, 06-10 October 2024, Marrakesh, Morocco). Berlin [u.a.]: Springer, 2024. 211-220 (Lect. Notes Comput. Sc. ; 15011)
DOI
as soon as is submitted to ZB.
Diffusion models have advanced unsupervised anomaly detection by improving the transformation of pathological images into pseudo-healthy equivalents. Nonetheless, standard approaches may compromise critical information during pathology removal, leading to restorations that do not align with unaffected regions in the original scans. Such discrepancies can inadvertently increase false positive rates and reduce specificity, complicating radiological evaluations. This paper introduces Temporal Harmonization for Optimal Restoration (THOR), which refines the reverse diffusion process by integrating implicit guidance through intermediate masks. THOR aims to preserve the integrity of healthy tissue details in reconstructed images, ensuring fidelity to the original scan in areas unaffected by pathology. Comparative evaluations reveal that THOR surpasses existing diffusion-based methods in retaining detail and precision in image restoration and detecting and segmenting anomalies in brain MRIs and wrist X-rays. Code: https://github.com/compai-lab/2024-miccai-bercea-thor.git.
Altmetric
Additional Metrics?
Publication type
Article: Conference contribution
Document type
Thesis type
Editors
Corresponding Author
Keywords
Generative AI; OoD; Brain MRI; Wrist X-ray
Keywords plus
ISSN (print) / ISBN
0302-9743
e-ISSN
1611-3349
ISBN
Book Volume Title
Conference Title
27th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2024
Conference Date
06-10 October 2024
Conference Location
Marrakesh, Morocco
Proceedings Title
Quellenangaben
Volume: 15011,
Issue: ,
Pages: 211-220
Article Number: ,
Supplement: ,
Series
Publisher
Springer
Publishing Place
Berlin [u.a.]
University
University place
Faculty
Publication date
0000-00-00
Application date
0000-00-00
Patent owner
Further owners
Application country
Patent priority
Reviewing status
Grants
Helmholtz Association under the joint research school 'Munich School for Data Science'
EVUK program ("Next-generation Al for Integrated Diagnostics") of the Free State of Bavaria