To address the challenges of high annotation costs, high inter-class similarity, and data imbalance in medical imaging, an unbiased semi-supervised classification model was proposed based on hybrid contrastive learning for medical image classification tasks. By introducing a semi-supervised training paradigm, the model reduces reliance on annotated data during training. The proposed model features two key components: a hybrid contrastive module and an adaptive debiasing module. The hybrid contrastive module performs class-level and instance-level contrastive learning on both labeled and unlabeled data, capturing rich implicit information and enhancing the model's feature extraction capabilities for medical images. The adaptive debiasing module decouples factors contributing to classification bias, enabling unbiased predictions and improving classification performance on imbalanced medical image datasets. Experimental results on two medical datasets with different modalities and imbalanced class distributions demonstrate that the proposed model outperforms existing state-of-the-art semi-supervised medical classification models across all evaluation metrics.
LeeDong-Hyun. Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks[C]∥IICML 2013 Workshop on Challenges in Representation Learning (WREPL), Atlanta, Georgia, USA, 2013,3: 896.
[2]
SamuliLaine, AilaTimo.Temporal ensembling for semi-supervised learning[J].arXiv,2016,1610.02242.
[3]
DavidBerthelot, CarliniNicholas, GoodfellowIan, et al. Mixmatch: a holistic approach to semi-supervised learning[C]∥33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada,2019.
[4]
ZahraSolatidehkordi, ZualkernanImran. Survey on recent trends in medical image classification using semi-supervised learning[J].Applied Sciences, 2022,12(23): 12094.
[5]
AmyarAmine, RomainModzelewski, LiHua, et al. Multi-task deep learning based CT imaging analysis for COVID-19 pneumonia: classification and segmentation[J].Computers in Biology and Medicine, 2020, 126: 104037.
[6]
HanKai, ShengVictor S, SongYu-qing, et al. Deep semi-supervised learning for medical image segmentation: a review[J]. Expert Systems with Applications, 2024, 245: 123052.
[7]
SohnK, BerthelotD, LiC L,et al.Fixmatch: simplifying semi-supervised learning with consistency and confidence[C]∥34th Conference on Neural Information Processing Systems (NeurIPS 2020), Vancouver, Canada,2020.
[8]
ZhangB, WangY, HouW,et al.FlexMatch: boosting semi-supervised learning with curriculum pseudo labeling[J]. Advances in Neural Information Processing Systems,2021, 34: 18408-18419.
[9]
WangY, ChenH, HengQ, et al.FreeMatch: self-adaptive thresholding for semi-supervised learning[J]. arXiv preprint arXiv: 2205. 07246.
[10]
TarvainenA, ValpolaH. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results[C]∥Proceedings of the 31st International Conference on Neural Information Processing Systems,Long Beach, California, USA,2017: 1195-1204.
[11]
LiuQ, YuL, LuoL,et al.Semi-supervised medical image classification with relation-driven self-ensembling model[J].IEEE Transactions on Medical Imaging, 2020, 39(11): 3429-3440.
[12]
JWuJ, FanH, ZhangX,et al.Semi-supervised semantic segmentation via entropy minimization[J].Advances in Neural Information Processing Systems, 2022, 35: 2803-2816.
[13]
CuiY, JiaM, LinT Y,et al.Class-Balanced Loss Based on Effective Number of Samples[C]∥In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,Long Beach, USA, 2019: 9268-9277.
[14]
FernandezA, GarciaS, ChawlaN V,et al.SMOTE for learning from imbalanced data: Progress and challenges, Marking the 15-year anniversary[J].Journal of Artificial Intelligence Research, 2018, 61: 863-905.
QuL, ZhuH S, ZhengR Q,et al. ImGAGN: imbalanced network embedding via generative adversarial graph networks[C]∥Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. Online:ACM,2021:1390-1398.
[17]
BareinboimE, PearlJ.Controlling selection bias in causal inference[J].Proceedings of the AAAI Conference on Artificial Intelligence, 2011, 25(1): 1754-1755.
[18]
ZhangD, ZhangH, TangJ, et al.Causal intervention for weakly-supervised semantic segmentation[J]. Advances in Neural Information Processing Systems, 2020, 33: 655-666.
[19]
RichiardiL, BelloccoR, ZugnaD, et al. Mediation analysis in epidemiology: methods, interpretation and bias[J].International Journal of Epidemiology, 2013, 42(5): 1511-1519.
[20]
CodellaN C F, GutmanD, CelebiE,et al.Skin lesion analysis toward melanoma detection: A challenge at the 2017 International symposium on biomedical imaging (ISBI), hosted by the international skin imaging collaboration (ISIC)[C]∥2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018), Washington, DC, USA, 2018, pp. 168-172,
[21]
WangX, PengY, LuL,et al.ChestX-ray8: hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases[J].In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2097-2106.
KrizhevskyA, HintonG. Learning multiple layers of features from tiny images[J].Handbook of Systemic Autoimmune Diseases, 2009, 1(4): 7.
[24]
CoatesA, NgA Y, LeeH. An analysis of single-layer networks in unsupervised feature learning[J].Journal of Machine Learning Research, 2011, 15: 215-223.
[25]
HuangG, LiuZ, LaurensV D M, et al.Densely Connected Convolutional Networks[C]∥2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017: 4700-4708.
[26]
HeK, ZhangX, RenS,et al. Deep residual learning for image recognition[C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016: 770-778.
[27]
DengJ, DongW, SocherR,et al.ImageNet: a large-scale hierarchical image database[C]∥2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, USA, 2009: 248-255.
[28]
LiuQ, YangH, DouQ,et al.. Federated semi-supervised medical image classification via inter-client relation matching[C]∥Medical Image Computing and Computer Assisted Intervention-MICCAI 2021,Strasbourg, France, 2021: 325-335.
[29]
ZhouZ, LuC, WangW,et al.Semi-supervised medical image classification based on attention and intrinsic features of samples[J].Applied Sciences, 2022, 12(13): 2076-3417.
[30]
PhamHieu, DaiZi-hang, XieQi-zhe,et al. Meta pseudo labels[C]∥In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,Nashville, TN, USA, 2021: 11557-11568.