Encoder-Decoder Contrast for Unsupervised Anomaly Detection in Medical Images
Unsupervised anomaly detection (UAD) aims to recognize anomalous images based on the training set that contains only normal images. In medical image analysis, UAD benefits from leveraging the easily obtained normal (healthy) images, avoiding the costly collecting and labeling of anomalous (unhealthy) images. Most advanced UAD methods rely on frozen encoder networks pre-trained using ImageNet for extracting feature representations. However, the features extracted from the frozen encoders that are borrowed from natural image domains coincide little with the features required in the target medical image domain. Moreover, optimizing encoders usually causes pattern collapse in UAD. In this paper, we propose a novel UAD method, namely Encoder-Decoder Contrast (EDC), which optimizes the entire network to reduce biases towards pre-trained image domain and orient the network in the target medical domain. We start from feature reconstruction approach that detects anomalies from reconstruction errors. Essentially, a contrastive learning paradigm is introduced to tackle the problem of pattern collapsing while optimizing the encoder and the reconstruction decoder simultaneously. In addition, to prevent instability and further improve performances, we propose to bring globality into the contrastive objective function. Extensive experiments are conducted across four medical image modalities including optical coherence tomography, color fundus image, brain MRI, and skin lesion image, where our method outperforms all current state-of-the-art UAD methods. Code is available at: https://github.com/guojiajeremy/EDC.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2024 |
---|---|
Erschienen: |
2024 |
Enthalten in: |
Zur Gesamtaufnahme - volume:43 |
---|---|
Enthalten in: |
IEEE transactions on medical imaging - 43(2024), 3 vom: 01. März, Seite 1102-1112 |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Guo, Jia [VerfasserIn] |
---|
Links: |
---|
Themen: |
---|
Anmerkungen: |
Date Completed 06.03.2024 Date Revised 06.03.2024 published: Print-Electronic Citation Status MEDLINE |
---|
doi: |
10.1109/TMI.2023.3327720 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM363753990 |
---|
LEADER | 01000caa a22002652 4500 | ||
---|---|---|---|
001 | NLM363753990 | ||
003 | DE-627 | ||
005 | 20240306232509.0 | ||
007 | cr uuu---uuuuu | ||
008 | 231226s2024 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1109/TMI.2023.3327720 |2 doi | |
028 | 5 | 2 | |a pubmed24n1318.xml |
035 | |a (DE-627)NLM363753990 | ||
035 | |a (NLM)37883280 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Guo, Jia |e verfasserin |4 aut | |
245 | 1 | 0 | |a Encoder-Decoder Contrast for Unsupervised Anomaly Detection in Medical Images |
264 | 1 | |c 2024 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ƒaComputermedien |b c |2 rdamedia | ||
338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
500 | |a Date Completed 06.03.2024 | ||
500 | |a Date Revised 06.03.2024 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status MEDLINE | ||
520 | |a Unsupervised anomaly detection (UAD) aims to recognize anomalous images based on the training set that contains only normal images. In medical image analysis, UAD benefits from leveraging the easily obtained normal (healthy) images, avoiding the costly collecting and labeling of anomalous (unhealthy) images. Most advanced UAD methods rely on frozen encoder networks pre-trained using ImageNet for extracting feature representations. However, the features extracted from the frozen encoders that are borrowed from natural image domains coincide little with the features required in the target medical image domain. Moreover, optimizing encoders usually causes pattern collapse in UAD. In this paper, we propose a novel UAD method, namely Encoder-Decoder Contrast (EDC), which optimizes the entire network to reduce biases towards pre-trained image domain and orient the network in the target medical domain. We start from feature reconstruction approach that detects anomalies from reconstruction errors. Essentially, a contrastive learning paradigm is introduced to tackle the problem of pattern collapsing while optimizing the encoder and the reconstruction decoder simultaneously. In addition, to prevent instability and further improve performances, we propose to bring globality into the contrastive objective function. Extensive experiments are conducted across four medical image modalities including optical coherence tomography, color fundus image, brain MRI, and skin lesion image, where our method outperforms all current state-of-the-art UAD methods. Code is available at: https://github.com/guojiajeremy/EDC | ||
650 | 4 | |a Journal Article | |
700 | 1 | |a Lu, Shuai |e verfasserin |4 aut | |
700 | 1 | |a Jia, Lize |e verfasserin |4 aut | |
700 | 1 | |a Zhang, Weihang |e verfasserin |4 aut | |
700 | 1 | |a Li, Huiqi |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t IEEE transactions on medical imaging |d 1982 |g 43(2024), 3 vom: 01. März, Seite 1102-1112 |w (DE-627)NLM082855269 |x 1558-254X |7 nnns |
773 | 1 | 8 | |g volume:43 |g year:2024 |g number:3 |g day:01 |g month:03 |g pages:1102-1112 |
856 | 4 | 0 | |u http://dx.doi.org/10.1109/TMI.2023.3327720 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 43 |j 2024 |e 3 |b 01 |c 03 |h 1102-1112 |