Analyzing Overfitting Under Class Imbalance in Neural Networks for Image Segmentation
Open accessThe DOI matches its Crossref record; 24 of the 24 references it cites were found in a registry or an index.
Abstract
Class imbalance poses a challenge for developing unbiased, accurate predictive models. In particular, in image segmentation neural networks may overfit to the foreground samples from small structures, which are often heavily under-represented in the training set, leading to poor generalization. In this study, we provide new insights on the problem of overfitting under class imbalance by inspecting the network behavior. We find empirically that when training with limited data and strong class imbalance, at test time the distribution of logit activations may shift across the decision boundary, while samples of the well-represented class seem unaffected. This bias leads to a systematic under-segmentation of small structures. This phenomenon is consistently observed for different databases, tasks and network architectures. To tackle this problem, we introduce new asymmetric variants of popular loss functions and regularization techniques including a large margin loss, focal loss, adversarial training, mixup and data augmentation, which are explicitly designed to counter logit shift of the under-represented classes. Extensive experiments are conducted on several challenging segmentation tasks. Our results demonstrate that the proposed modifications to the objective function can lead to significantly improved segmentation accuracy compared to baselines and alternative approaches.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
SMOTE synthetic minority over-sampling technique
Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002) · Journal of Artificial Intelligence Research · doi:10.1613/jair.953 · Cited by 32,865 · cited 1×
“2) Data Synthesis: Our work is related to data synthesis methods [6], [13] which generate synthetic samples of the minority classes based on intra-class relationship between samples to increase the variance of under-represented class.”
MethodFocal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.324 · Cited by 30,581
A survey on deep learning in medical image analysis
Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., van der Laak, J. A. W. M., van Ginneken, B., & Sánchez, C. I. (2017) · Medical Image Analysis · doi:10.1016/j.media.2017.07.005 · Cited by 15,332
V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
Milletari, F., Navab, N., & Ahmadi, S.-A. (2016) · 2016 Fourth International Conference on 3D Vision (3DV) · doi:10.1109/3dv.2016.79 · Cited by 11,931 · cited 3×
“alternative choice for the loss function which is widely used for medical image segmentation [33].”
BackgroundnnU-Net: a self-configuring method for deep learning-based biomedical image segmentation
Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J., & Maier-Hein, K. H. (2021) · Nature Methods · doi:10.1038/s41592-020-01008-z · Cited by 10,477
Explaining and Harnessing Adversarial Examples
Goodfellow, I. J., Shlens, J., & Szegedy, C. (2014) · arXiv · doi:10.48550/arxiv.1412.6572 · Cited by 10,381
Focal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2018.2858826 · Cited by 9,556
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
Deng, J., Guo, J., Xue, N., & Zafeiriou, S. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00482 · Cited by 7,781 · cited 1×
“Margin based loss functions were proposed to learn discriminative embeddings and widely adopted to metric-learning and face recognition [10], [31].”
BackgroundCutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features
Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., & Yoo, Y. (2019) · arXiv · doi:10.48550/arxiv.1905.04899 · Cited by 5,439
CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features
Yun, S., Han, D., Chun, S., Oh, S. J., Yoo, Y., & Choe, J. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00612 · Cited by 5,400
Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation
Kamnitsas, K., Ledig, C., Newcombe, V. F. J., Simpson, J. P., Kane, A. D., Menon, D. K., Rueckert, D., & Glocker, B. (2017) · Medical Image Analysis · doi:10.1016/j.media.2016.10.004 · Cited by 3,498 · cited 3×
“1) Brain Tumor Segmentation: We first evaluate the asymmetric techniques for the case of binary brain tumor core segmentation using the DeepMedic network architecture, a well performing method for this task [20].”
MethodSelf-citationBrain tumor segmentation with Deep Neural Networks
Havaei, M., Davy, A., Warde-Farley, D., Biard, A., Courville, A., Bengio, Y., Pal, C., Jodoin, P.-M., & Larochelle, H. (2017) · Medical Image Analysis · doi:10.1016/j.media.2016.05.004 · Cited by 3,278
Brain Tumor Segmentation with Deep Neural Networks
Havaei, M., Davy, A., Warde-Farley, D., Biard, A., Courville, A., Bengio, Y., Pal, C., Jodoin, P.-M., & Larochelle, H. (2015) · arXiv · doi:10.48550/arxiv.1505.03540 · Cited by 3,278
RandAugment: Practical automated data augmentation with a reduced search space
Cubuk, E. D., Zoph, B., Shlens, J., & Le, Q. V. (2019) · arXiv · doi:10.48550/arxiv.1909.13719 · Cited by 3,223
Class-Balanced Loss Based on Effective Number of Samples
Cui, Y., Jia, M., Lin, T.-Y., Song, Y., & Belongie, S. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00949 · Cited by 3,094 · cited 2×
“Similar to [9], the class imbalance ratio of one image can be defined as the ratio between the number of pixels of the background class (which is commonly the most frequent class) and the number of pixels of different object classes.”
BackgroundA systematic study of the class imbalance problem in convolutional neural networks
Buda, M., Maki, A., & Mazurowski, M. A. (2018) · Neural Networks · doi:10.1016/j.neunet.2018.07.011 · Cited by 3,088 · cited 1×
“Class imbalance, which has been the focus of previous works [4], [19], is a common issue in image classification and image segmentation.”
BackgroundBakas, S., Akbari, H., Sotiras, A., Bilello, M., Rozycki, M., Kirby, J. S., Freymann, J. B., Farahani, K., & Davatzikos, C. (2017) · Scientific Data · doi:10.1038/sdata.2017.117 · Cited by 3,049 · cited 2×
“We conduct experiments on challenging segmentation tasks using data from the Multimodal Brain Tumor Image Segmentation (BRATS) challenge [2], the Anatomical Tracings of Lesions After Stroke (ATLAS) dataset [28], small organ segmentation (data from [25]) and Kidney Tumor Segmentation (KiTS) [16].”
MethodSurvey on deep learning with class imbalance
Johnson, J. M., & Khoshgoftaar, T. M. (2019) · Journal of Big Data · doi:10.1186/s40537-019-0192-5 · Cited by 3,032
Setio, A. A. A., Ciompi, F., Litjens, G., Gerke, P., Jacobs, C., van Riel, S. J., Wille, M. M. W., Naqibullah, M., Sanchez, C. I., & van Ginneken, B. (2016) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2016.2536809 · Cited by 1,258
AutoAugment: Learning Augmentation Policies from Data
Cubuk, E. D., Zoph, B., Mane, D., Vasudevan, V., & Le, Q. V. (2018) · arXiv · doi:10.48550/arxiv.1805.09501 · Cited by 1,106
Cost-sensitive learning by cost-proportionate example weighting
Zadrozny, B., Langford, J., & Abe, N. (2004) · Third IEEE International Conference on Data Mining · doi:10.1109/icdm.2003.1250950 · Cited by 680 · cited 1×
“1) Re-weighting: A common approach to tackle class imbalance is class-level re-weighting, which assigns higher weights or higher sampling probability to the underrepresented classes based on sample frequency [41], [46] or advanced rules [9].”
BackgroundTraining deep neural networks on imbalanced data sets
Wang, S., Liu, W., Wu, J., Cao, L., Meng, Q., & Kennedy, P. J. (2016) · 2016 International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2016.7727770 · Cited by 491 · cited 1×
“or higher sampling probability to the under-represented classes based on sample frequency [41], [46] or advanced rules [9].”
BackgroundBrosch, T., Tang, L. Y. W., Yoo, Y., Li, D. K. B., Traboulsee, A., & Tam, R. (2016) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2016.2528821 · Cited by 460 · cited 1×
“foreground classes) are heavily under-represented in the training data [3], [30].”
BackgroundA large, open source dataset of stroke anatomical brain images and manual lesion segmentations
Liew, S.-L., Anglin, J. M., Banks, N. W., Sondag, M., Ito, K. L., Kim, H., Chan, J., Ito, J., Jung, C., Khoshab, N., et al. (2018) · Scientific Data · doi:10.1038/sdata.2018.11 · Cited by 283
Metadata
- Authors
- Li, Z., Kamnitsas, K., & Glocker, B.
- Venue
- IEEE Transactions on Medical Imaging
- Published
- 2021-03
- Volume
- 40
- Issue
- 3
- Pages
- 1065-1077
- ISSN
- 0278-0062 (ISSN-L)
- 1558-254X
- OpenAlex
- W3114304902
- Cited by
- 193
- References
- 24
- FWCI
- 6.62
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.