J Shanghai Jiaotong Univ Sci

Journal of Shanghai Jiaotong University(Science) >

2026 , Vol. 31 >Issue 3: 631 - 645

DOI: https://doi.org/10.1007/s12204-026-2910-4

Celebrating 30 Years

Unsupervised Anomaly Detection in 3D Medical Imaging via Normal Sample Learning: A Survey and Benchmark

Expand
  • 1. School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Shanghai 200240, China; 2. Department of Radiology, The Second Af?liated Hospital of Fujian Medical University, Quanzhou 362000, Fujian, China; 3. Department of Radiology, Gansu Provincial Maternity and Child-Care Hospital, Lanzhou 730050, China

Received date: 2025-12-24

  Revised date: 2026-01-12

  Accepted date: 2026-02-02

  Online published: 2026-03-24

Fold

Abstract

How computers can acquire the capability of radiologists to identify various anomalies within 3D medical images remains a central and challenging problem in intelligent imaging diagnosis. Supervised learning methods relying on lesion-level annotations are restricted to detecting a limited set of lesion types similar to those in the training data. Furthermore, obtaining voxel-level annotations is notoriously labor-intensive and time-consuming. To address these challenges, researchers have proposed unsupervised anomaly detection (UAD) methods for 3D medical images that learn from normal (healthy) samples. These approaches aim to identify any lesion that deviates from the learned normal anatomical distribution. Centered on this principle, three distinct UAD paradigms have emerged: self-supervised learning-based methods that train supervised models by constructing synthetic anomalies; deep feature embedding-based methods that measure the distance between test features and normal feature distribution; and reconstruction-based methods that transform abnormal images into pseudo-healthy images based on learned normal patterns. This paper provides a comprehensive review, analysis, and comparison of key studies within these three categories, highlighting their respective strengths, limitations, and potential research directions. Moreover, existing UAD methods are often validated using inconsistent datasets or preprocessing pipelines, making fair performance comparison challenging. To address this issue, we establish a unified benchmark for evaluating UAD methods on 3D medical images, comprising nine brain MRI datasets (4365 cases) and six liver CT datasets (376 cases). Using this benchmark, we evaluate the performance of 21 representative state-of-the-art algorithms on the 3D voxel-level anomaly localization task, objectively revealing the advantages and limitations of these methods.

Cite this article

Yang Tao, Chen Ningxin, Yao Liguo, Wang Lisheng . Unsupervised Anomaly Detection in 3D Medical Imaging via Normal Sample Learning: A Survey and Benchmark[J]. Journal of Shanghai Jiaotong University(Science), 2026 , 31(3) : 631 -645 . DOI: 10.1007/s12204-026-2910-4

References

[1] Xu W W, Xu L F, Liu N H, et al. Histological image diagnosis of breast cancer based on multi-attention convolution neural network [J]. Journal of Shanghai Jiao Tong University (Science), 2025, 30(1): 91-106.
[2] Hsu W C, Meuschke M, Frangi A F, et al. A survey of intracranial aneurysm detection and segmentation [J]. Medical Image Analysis, 2025, 101: 103493.
[3] Miao J, Chang Y R, Chen C, et al. Ground-glass lung nodules recognition based on CatBoost feature selection and stacking ensemble learning [J]. Journal of Shanghai Jiao Tong University (Science), 2025, 30(4): 790-799.
[4] Aiello M, Cavaliere C, D’Albore A, et al. The challenges of diagnostic imaging in the era of big data [J]. Journal of Clinical Medicine, 2019, 8(3): 316.
[5] Menze B H, Jakab A, Bauer S, et al. The multimodal brain tumor image segmentation benchmark (BRATS) [J]. IEEE Transactions on Medical Imaging, 2015, 34(10): 1993-2024.
[6] Kuang H L, Wang Y H, Liu J, et al. Hybrid CNN-transformer network with circular feature interaction for acute ischemic stroke lesion segmentation on non-contrast CT scans [J]. IEEE Transactions on Medical Imaging, 2024, 43(6): 2303-2316.
[7] Wang D, Zhang Y, Zhang K X, et al. FocalMix: Semi-supervised learning for 3D medical image detection [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 3950-3959.
[8] Dong H Y, Zhang Y F, Gu H X, et al. SWSSL: Sliding window-based self-supervised learning for anomaly detection in high-resolution images [J]. IEEE Transactions on Medical Imaging, 2023, 42(12): 3860-3870.
[9] Luo G T, Xie W, Gao R H, et al. Unsupervised anomaly detection in brain MRI: Learning abstract distribution from massive healthy brains [J]. Computers in Biology and Medicine, 2023, 154: 106610.
[10] Xu S C, Li W, Li Z Y, et al. Facing differences of similarity: Intra- and inter-correlation unsupervised learning for chest X-ray anomaly detection [J]. IEEE Transactions on Medical Imaging, 2025, 44(2): 801-814.
[11] Kascenas A, Sanchez P, Schrempf P, et al. The role of noise in denoising models for anomaly detection in medical images [J]. Medical Image Analysis, 2023, 90: 102963.
[12] Cai Y, Zhang W W, Chen H, et al. MedIAnomaly: A comparative study of anomaly detection in medical images [J]. Medical Image Analysis, 2025, 102: 103500.
[13] Li C L, Sohn K, Yoon J, et al. CutPaste: Self-supervised learning for anomaly detection and localization [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 9659-9669.
[14] Tan J, Hou B, Batten J, et al. Detecting outliers with foreign patch interpolation [J]. Machine Learning for Biomedical Imaging, 2022, 1: 1-27.
[15] Tan J, Hou B, Day T, et al. Detecting outliers with Poisson image interpolation [M]//Medical image computing and computer assisted intervention – MICCAI 2021. Cham: Springer, 2021: 581-591.
[16] Deng H Q, Li X Y. Self-supervised anomaly detection with random-shape pseudo-outliers [C]//2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society. Glasgow: IEEE, 2022: 4768-4772.
[17] Zavrtanik V, Kristan M, Skočaj D. DRÆM–A discriminatively trained reconstruction embedding for surface anomaly detection [C]//2021 IEEE/CVF International Conference on Computer Vision. Montreal: IEEE, 2021: 8310-8319.
[18] Schlüter H M, Tan J, Hou B, et al. Natural synthetic anomalies for self-supervised anomaly detection and localization [M]//Computer vision – ECCV 2022. Cham: Springer, 2022: 474-489.
[19] Sato J, Suzuki Y, Wataya T, et al. Anatomy-aware self-supervised learning for anomaly detection in chest radiographs [J]. iScience, 2023, 26(7): 107086.
[20] Li W, Liu G H, Fan H Y, et al. Self-supervised multi-scale cropping and simple masked attentive predicting for lung CT-scan anomaly detection [J]. IEEE Transactions on Medical Imaging, 2024, 43(1): 594-607.
[21] Defard T, Setkov A, Loesch A, et al. PaDiM: A patch distribution modeling framework for anomaly detection and localization [M]//Pattern recognition. ICPR international workshops and challenges. Cham: Springer, 2021: 475-489.
[22] Roth K, Pemula L, Zepeda J, et al. Towards total recall in industrial anomaly detection [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 14298-14308.
[23] Lee S, Lee S, Song B C. CFA: Coupled-hypersphere-based feature adaptation for target-oriented anomaly localization [J]. IEEE Access, 2022, 10: 78446-78454.
[24] Salehi M, Sadjadi N, Baselizadeh S, et al. Multiresolution knowledge distillation for anomaly detection [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 14897-14907.
[25] Wang G D, Han S M, Ding E R, et al. Student-teacher feature pyramid matching for anomaly detection [C]//Proceedings of the British Machine Vision Conference 2021. Online: British Machine Vision Association, 2021: 349.
[26] Deng H Q, Li X Y. Anomaly detection via reverse distillation from one-class embedding [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 9727-9736.
[27] Batzner K, Heckler L, König R. EfficientAD: Accurate visual anomaly detection at millisecond-level latencies [C]//2024 IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa: IEEE, 2024: 127-137.
[28] Gudovskiy D, Ishizaka S, Kozuka K. CFLOW-AD: Real-time unsupervised anomaly detection with localization via conditional normalizing flows [C]//2022 IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa: IEEE, 2022: 1819-1828.
[29] Silva-Rodríguez J, Naranjo V, Dolz J. Constrained unsupervised anomaly segmentation [J]. Medical Image Analysis, 2022, 80: 102526.
[30] Chen L Y, You Z Y, Zhang N, et al. UTRAD: Anomaly detection and localization with U-Transformer [J]. Neural Networks, 2022, 147: 53-62.
[31] Baur C, Wiestler B, Albarqouni S, et al. Deep autoencoding models for unsupervised anomaly segmentation in brain MR images [M]//Brainlesion: Glioma, multiple sclerosis, stroke and traumatic brain injuries. Cham: Springer, 2019: 161-169.
[32] Gong D, Liu L Q, Le V, et al. Memorizing normality to detect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection [C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2020: 1705-1714.
[33] Zavrtanik V, Kristan M, Skočaj D. Reconstruction by inpainting for visual anomaly detection [J]. Pattern Recognition, 2021, 112: 107706.
[34] Zhou K, Li J, Luo W X, et al. Proxy-bridged image reconstruction network for anomaly detection in medical images [J]. IEEE Transactions on Medical Imaging, 2022, 41(3): 582-594.
[35] Kingma D P, Welling M. Auto-encoding variational Bayes [PP/OL]. V11. arXiv (2022-12-10) [2025-12-21]. https://doi.org/10.48550/arXiv.1312.6114.
[36] Goodfellow I J, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets [C]// 28th Conference on Neural Information Processing Systems. Montreal: NIPS, 2014: 2672-2680.
[37] Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models [C]// 34th Conference on Neural Information Processing Systems. Vancouver: NIPS, 2020: 6840-6851.
[38] Zimmerer D, Isensee F, Petersen J, et al. Unsupervised anomaly localization using variational auto-encoders [M]//Medical image computing and computer assisted intervention – MICCAI 2019. Cham: Springer, 2019: 289-297.
[39] Bercea C I, Wiestler B, Rueckert D, et al. Reversing the abnormal: Pseudo-healthy generative networks for anomaly detection [M]//Medical image computing and computer assisted intervention – MICCAI 2023. Cham: Springer, 2023: 293-303.
[40] Zeng Y H, Fu J L, Chao H Y, et al. Aggregated contextual transformations for high-resolution image inpainting [J]. IEEE Transactions on Visualization and Computer Graphics, 2023, 29(7): 3266-3280.
[41] Schlegl T, Seeböck P, Waldstein S M, et al. F-AnoGAN: Fast unsupervised anomaly detection with generative adversarial networks [J]. Medical Image Analysis, 2019, 54: 30-44. 
[42] Wyatt J, Leach A, Schmon S M, et al. AnoDDPM: Anomaly detection with denoising diffusion probabilistic models using simplex noise [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. New Orleans: IEEE, 2022: 649-655. 
[43] Bercea C I, Neumayr M, Rueckert D, et al. Mask, stitch, and re-sample: Enhancing robustness and generalizability in anomaly detection through automatic diffusion models [PP/OL]. ArXiv (2023-05-31) [2025-12-21]. https://doi.org/10.48550/arXiv.2305.19643.
[44] Baid U, Ghodasara S, Mohan S, et al. The RSNA-ASNR-MICCAI BraTS 2021 benchmark on brain tumor segmentation and radiogenomic classification [PP/OL]. V2. arXiv (2021-09-12) [2025-12-21]. https://doi.org/10.48550/arXiv.2107.02314. 
[45] Bakas S, Sako C, Akbari H, et al. The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: Advanced MRI, clinical, genomics, & radiomics [J]. Scientific Data, 2022, 9: 453. 
[46] Calabrese E, Villanueva-Meyer J E, Rudie J D, et al. The university of California San Francisco preoperative diffuse glioma MRI dataset [J]. Radiology Artificial Intelligence, 2022, 4(6): e220058. 
[47] Adewole M, Rudie J D, Gbadamosi A, et al. The brain tumor segmentation (BraTS) challenge 2023: Glioma segmentation in sub-Saharan Africa patient population (BraTS-Africa) [PP/OL]. arXiv (2023-05-30) [2025-12-21]. https://doi.org/10.48550/arXiv.2305.19369. 
[48] Kazerooni A F, Khalili N, Liu X Y, et al. The brain tumor segmentation (BraTS) challenge 2023: Focus on pediatrics (CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs) [PP/OL]. V7. arXiv (2024-05-23) [2025-12-21]. https://doi.org/10.48550/arXiv.2305.17033. 
[49] LaBella D, Khanna O, McBurney-Lin S, et al. A multi-institutional meningioma MRI dataset for automated multi-sequence image segmentation [J]. Scientific Data, 2024, 11: 496. 
[50] Moawad A W, Janas A, Baid U, et al. The brain tumor segmentation (BraTS-METS) challenge 2023: Brain metastasis segmentation on pre-treatment MRI [PP/OL]. V3. arXiv (2024-12-09) [2025-12-21]. https://doi.org/10.48550/arXiv.2306.00838. 
[51] Lesjak Ž, Galimzianova A, Koren A, et al. A novel public MR image dataset of multiple sclerosis patients with lesion segmentations based on multi-rater consensus [J]. Neuroinformatics, 2018, 16(1): 51-63. 
[52] Roth H R, Lu L, Farag A, et al. DeepOrgan: Multi-level deep convolutional networks for automated pancreas segmentation [M]//Medical image computing and computer-assisted intervention – MICCAI 2015. Cham: Springer, 2015: 556-564. 
[53] Kavur A E, Gezer N S, Barış M, et al. CHAOS Challenge - combined (CT-MR) healthy abdominal organ segmentation [J]. Medical Image Analysis, 2021, 69: 101950.
[54] Antonelli M, Reinke A, Bakas S, et al. The medical segmentation decathlon [J]. Nature Communications, 2022, 13: 4128.
[55] Bilic P, Christ P, Li H B, et al. The liver tumor segmentation benchmark (LiTS) [J]. Medical Image Analysis, 2023, 84: 102680.
[56] Moawad A W, Morshid A, Khalaf A M, et al. Multimodality annotated hepatocellular carcinoma data set including pre- and post-TACE with imaging segmentation [J]. Scientific Data, 2023, 10: 33.
[57] Bao J N, Sun H S, Deng H Q, et al. BMAD: Benchmarks for medical anomaly detection [C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Seattle: IEEE, 2024: 4042-4053.
[58] Davatzikos C, Rathore S, Bakas S, et al. Cancer imaging phenomics toolkit: Quantitative imaging analytics for precision diagnostics and predictive modeling of clinical outcome [J]. Journal of Medical Imaging, 2018, 5(1): 011018.
[59] Thakur S, Doshi J, Pati S, et al. Brain extraction on MRI scans in presence of diffuse glioma: Multi-institutional performance evaluation of deep learning methods and robust modality-agnostic training [J]. NeuroImage, 2020, 220: 117081.
[60] Isensee F, Jaeger P F, Kohl S A A, et al. nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation [J]. Nature Methods, 2021, 18(2): 203-211.
[61] Ji Y, Bai H, Ge C, et al. AMOS: A large-scale abdominal multi-organ benchmark for versatile medical image segmentation [C]// 36th Conference on Neural Information Processing Systems. New Orleans: NIPS, 2022: 36722-36732.
[62] Zuo L R, Dewey B E, Liu Y H, et al. Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory [J]. NeuroImage, 2021, 243: 118569.
[63] Yang T, Wang X Y, Liu H, et al. Unsupervised anomaly detection in brain MRI via disentangled anatomy learning [J]. Medical Image Analysis, 2026, 109: 103922.
[64] Zimmerer D, Full P M, Isensee F, et al. MOOD 2020: A public benchmark for out-of-distribution detection and localization on medical images [J]. IEEE Transactions on Medical Imaging, 2022, 41(10): 2728-2738.
[65] Baur C, Denner S, Wiestler B, et al. Autoencoders for unsupervised anomaly segmentation in brain MR images: A comparative study [J]. Medical Image Analysis, 2021, 69: 101952.
[66] Song J, Meng C, Ermon S. Denoising diffusion implicit models [PP/OL]. V4. arXiv (2022-10-05) [2025-12-21]. https://doi.org/10.48550/arXiv.2010.02502.
[67] You Z Y, Cui L, Shen Y, et al. A unified model for multi-class anomaly detection [C]// 36th International Conference on Neural Information Processing Systems. New Orleans: NIPS, 2022: 4571-4584.
[68] Kirillov A, Mintun E, Ravi N, et al. Segment anything [C]//2023 IEEE/CVF International Conference on Computer Vision. Paris: IEEE, 2024: 3992-4003.

Options
Outlines

/

〈 〉