• U<superscript>2</supe...

*Result*: U2AD: Uncertainty-based unsupervised anomaly detection framework for detecting T2 hyperintensity in MRI spinal cord.

Title:
U2AD: Uncertainty-based unsupervised anomaly detection framework for detecting T2 hyperintensity in MRI spinal cord.
Authors:
Zhang Q; School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China. Electronic address: zhi-bai-shou-hei@sjtu.edu.cn., Chen X; Department of Spine Surgery, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, 200127, China. Electronic address: chenxiuyuan@renji.com., He Z; Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, 200025, China. Electronic address: hzy12950@rjh.com.cn., Wang K; Department of Spine Surgery, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, 200127, China. Electronic address: wk52693158@126.com., Wu L; Department of Radiology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, 200127, China. Electronic address: wlmssmu@126.com., Shen H; Department of Spine Surgery, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, 200127, China. Electronic address: shenhongxing@renji.com., Sun J; School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China; National Engineering Research Center of Advanced Magnetic Resonance Technologies for Diagnosis and Therapy (NERC-AMRT), Shanghai, China; Med-X Research Institute, Shanghai Jiao Tong University, Shanghai, 200030, China. Electronic address: milesun@sjtu.edu.cn.
Source:
Medical image analysis [Med Image Anal] 2026 Mar; Vol. 109, pp. 103939. Date of Electronic Publication: 2026 Jan 17.
Publication Type:
Journal Article
Language:
English
Journal Info:
Publisher: Elsevier Country of Publication: Netherlands NLM ID: 9713490 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1361-8423 (Electronic) Linking ISSN: 13618415 NLM ISO Abbreviation: Med Image Anal Subsets: MEDLINE
Imprint Name(s):
Publication: Amsterdam : Elsevier
Original Publication: London : Oxford University Press, [1996-
MeSH Terms:
Magnetic Resonance Imaging*/methods , Spinal Cord*/diagnostic imaging , Image Interpretation, Computer-Assisted*/methods , Unsupervised Machine Learning*, Humans ; Uncertainty ; Deep Learning
Contributed Indexing:
Keywords: Masked image modeling; Spinal cord lesion; T2 hyperintensity; Uncertainty estimation; Unsupervised anomaly detection
Entry Date(s):
Date Created: 20260120 Date Completed: 20260207 Latest Revision: 20260207
Update Code:
20260208
DOI:
10.1016/j.media.2026.103939
PMID:
41558246
Database:
MEDLINE

*Further Information*

*T2 hyperintensities in spinal cord MR images are crucial biomarkers for conditions such as degenerative cervical myelopathy (DCM). However, current clinical diagnoses primarily rely on manual evaluation. Deep learning methods have shown promise in lesion detection, but most supervised approaches are heavily dependent on large, annotated datasets. Unsupervised anomaly detection (UAD) offers a compelling alternative by eliminating the need for abnormal data annotations. However, existing UAD methods face challenges of domain shifts and task conflict. We propose an Uncertainty-based Unsupervised Anomaly Detection framework, termed U<sup>2</sup>AD, to address these limitations. Unlike traditional methods, U<sup>2</sup>AD is designed to be trained and tested within the same clinical dataset, following a "mask-and-reconstruction" paradigm built on a Vision Transformer-based architecture. We introduce an uncertainty-guided masking strategy to resolve task conflicts between normal reconstruction and anomaly detection to achieve an optimal balance. Specifically, we employ a Monte-Carlo inference technique to estimate reconstruction uncertainty mappings during training. By iteratively optimizing reconstruction training under the guidance of both epistemic and aleatoric uncertainty, U<sup>2</sup>AD improves the normal representation learning while maintaining the sensitivity to anomalies. Experimental results demonstrate that U<sup>2</sup>AD outperforms existing UAD methods in patient-level identification and segment-level localization of spinal cord T2 hyperintensities. This framework establishes a new benchmark for incorporating uncertainty guidance into UAD. Our code is available at: https://github.com/zhibaishouheilab/U2AD.
(Copyright © 2026 Elsevier B.V. All rights reserved.)*

*Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.*