Contrastive learning of global and local features for medical image\n segmentation with limited annotations

2020/06/18 by Krishna Chaitanya, Chaitanya, Krishna, Ertunç Erdil +5 · 16 citations
Computer Science · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #MRI in cancer diagnosis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Radiomics and Machine Learning in Medical Imaging #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2006.10511

openalex publication_date 2020/06/18 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28

Abstract

A key requirement for the success of supervised deep learning is a large\nlabeled dataset - a condition that is difficult to meet in medical image\nanalysis. Self-supervised learning (SSL) can help in this regard by providing a\nstrategy to pre-train a neural network with unlabeled data, followed by\nfine-tuning for a downstream task with limited annotations. Contrastive\nlearning, a particular variant of SSL, is a powerful technique for learning\nimage-level representations. In this work, we propose strategies for extending\nthe contrastive learning framework for segmentation of volumetric medical\nimages in the semi-supervised setting with limited annotations, by leveraging\ndomain-specific and problem-specific cues. Specifically, we propose (1) novel\ncontrasting strategies that leverage structural similarity across volumetric\nmedical images (domain-specific cue) and (2) a local version of the contrastive\nloss to learn distinctive representations of local regions that are useful for\nper-pixel segmentation (problem-specific cue). We carry out an extensive\nevaluation on three Magnetic Resonance Imaging (MRI) datasets. In the limited\nannotation setting, the proposed method yields substantial improvements\ncompared to other self-supervision and semi-supervised learning techniques.\nWhen combined with a simple data augmentation technique, the proposed method\nreaches within 8% of benchmark performance using only two labeled MRI volumes\nfor training, corresponding to only 4% (for ACDC) of the training data used to\ntrain the benchmark. The code is made public at\nhttps://github.com/krishnabits001/domain_specific_cl.\n

Cited by

• Beyond CLIP: Knowledge-Enhanced Multimodal Transformers for Cross-Modal Alignment in Diabetic Retinopathy Diagnosis
• MRI-Based Brain Age Estimation with Supervised Contrastive Learning of Continuous Representation
• Coordinative Learning with Ordinal and Relational Priors for Volumetric Medical Image Segmentation
• Cyclic Self-Supervised Diffusion for Ultra Low-field to High-field MRI Synthesis
• SSVIF: Self-Supervised Segmentation-Oriented Visible and Infrared Image Fusion
• Domain and Task-Focused Example Selection for Data-Efficient Contrastive Medical Image Segmentation
• DiSSECT: Structuring Transfer-Ready Medical Image Representations through Discrete Self-Supervision
• Aggregated Mutual Learning between CNN and Transformer for semi-supervised medical image segmentation
• MultiMAE for Brain MRIs: Robustness to Missing Inputs Using Multi-Modal Masked Autoencoder
• Barlow-Swin: Toward a novel siamese-based segmentation architecture using Swin-Transformers
• Self-Validated Learning for Particle Separation: A Correctness-Based Self-Training Framework Without Human Labels
• Multi-Level CLS Token Fusion for Contrastive Learning in Endoscopy Image Classification
• VasoMIM: Vascular Anatomy-Aware Masked Image Modeling for Vessel Segmentation
• Improve Retinal Artery/Vein Classification via Channel Couplin
• AGA: An adaptive group alignment framework for structured medical cross-modal representation learning
• Is Exchangeability better than I.I.D to handle Data Distribution Shifts while Pooling Data for Data-scarce Medical image segmentation?

Related

• Localized Region Contrast for Enhancing Self-Supervised Learning in Medical Image Segmentation
• Overcoming Dimensional Collapse in Self-supervised Contrastive Learning for Medical Image Segmentation
• Semi-supervised Contrastive Learning for Label-efficient Medical Image Segmentation
• Domain and Task-Focused Example Selection for Data-Efficient Contrastive Medical Image Segmentation
• Self-Ensembling Contrastive Learning for Semi-Supervised Medical Image Segmentation