mlsanityData to 7 Oct 2026

What Makes Transfer Learning Work For Medical Images: Feature Reuse & Other Factors

arXiv 2203.01825 ↗First posted 2 Mar 2022cs.LG cs.CV eess.IVcited by 61

Not on the map. Why?

Abstract (arXiv)

Transfer learning is a standard technique to transfer knowledge from one domain to another. For applications in medical imaging, transfer from ImageNet has become the de-facto approach, despite differences in the tasks and image characteristics between the domains. However, it is unclear what factors determine whether - and to what extent - transfer learning to the medical domain is useful. The long-standing assumption that features from the source domain get reused has recently been called into question. Through a series of experiments on several medical image benchmark datasets, we explore the relationship between transfer learning, data size, the capacity and inductive bias of the model, as well as the distance between the source and target domain. Our findings suggest that transfer learning is beneficial in most cases, and we characterize the important role feature reuse plays in its success.

Key facts. Cited by 61 papers in the map (data through 7 Oct 2026). Builds on 32 of its 47 references that are in the map. First posted on arXiv 2 Mar 2022.

Builds on

32 of 47 references.

References · 32

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  4. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,082
  5. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  6. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  7. 2020Training data-efficient image transformers & distillation through attentioncited by 4,560
  8. 2015Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classificationcited by 3,590
  9. 2013Rich feature hierarchies for accurate object detection and semantic segmentationcited by 2,160
  10. 2019Similarity of Neural Network Representations Revisitedcited by 1,695
  11. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  12. 2019CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparisoncited by 1,410
20 more
  1. 2014How transferable are features in deep neural networks?cited by 1,216
  2. 2021Do Vision Transformers See Like Convolutional Neural Networks?cited by 671
  3. 2018Do Better ImageNet Models Transfer Better?cited by 560
  4. 2017Skin Lesion Analysis Toward Melanoma Detection: A Challenge at the 2017 International Symposium on Biomedical Imaging (ISBI), Hosted by the International Skin Imaging Collaboration (ISIC)cited by 549
  5. 2018Rotation Equivariant CNNs for Digital Pathologycited by 420
  6. 2018Rethinking ImageNet Pre-trainingcited by 376
  7. 2019Transfusion: Understanding Transfer Learning for Medical Imagingcited by 366
  8. 2020What is being transferred in transfer learning?cited by 365
  9. 2019Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAMLcited by 359
  10. 2017Are GANs Created Equal? A Large-Scale Studycited by 267
  11. 2020Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depthcited by 241
  12. 2018Pros and Cons of GAN Evaluation Measurescited by 224
  13. 2019BCN20000: Dermoscopic Lesions in the Wildcited by 224
  14. 2017Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning?cited by 220
  15. 2021Revisiting ResNets: Improved Training and Scaling Strategiescited by 158
  16. 2019Are All Layers Created Equal?cited by 94
  17. 2021Is it Time to Replace CNNs with Transformers for Medical Images?cited by 60
  18. 2020A scoping review of transfer learning research on medical image analysis using ImageNetcited by 54
  19. 2021Supervised Transfer Learning at Scale for Medical Imagingcited by 38
  20. 2018Domain Adaptive Transfer Learning with Specialist Modelscited by 31

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 6; 2023: 22; 2024: 16; 2025: 12; 2026: 5.

61 papers.

Sort:

By area:

  1. 2022Benchmarking Self-Supervised Learning on Diverse Pathology Datasetscited by 148
  2. 2023A Foundation Language-Image Model of the Retina (FLAIR): Encoding Expert Knowledge in Text Supervisioncited by 78
  3. 2023Are Natural Domain Foundation Models Useful for Medical Image Classification?cited by 44
  4. 2022PatchDropout: Economizing Vision Transformers Using Patch Dropoutcited by 28
  5. 2023Promise:Prompt-driven 3D Medical Image Segmentation Using Pretrained Image Foundation Modelscited by 18
  6. 2023Rotation-Agnostic Image Representation Learning for Digital Pathologycited by 15
  7. 2023Pretrained ViTs Yield Versatile Representations For Medical Imagescited by 13
  8. 2025Understanding the Learning Dynamics of LoRA: A Gradient Flow Perspective on Low-Rank Adaptation in Matrix Factorizationcited by 13
  9. 2025Towards Scalable Language-Image Pre-training for 3D Medical Imagingcited by 13
  10. 2024Source Matters: Source Dataset Impact on Model Robustness in Medical Imagingcited by 10
  11. 2022Metadata-guided Consistency Learning for High Content Imagescited by 8
  12. 2023DPA-P2PNet: Deformable Proposal-aware P2PNet for Accurate Point-based Cell Detectioncited by 8
38 more
  1. 2023Generalizing Across Domains in Diabetic Retinopathy via Variational Autoencoderscited by 8
  2. 2024RadCLIP: Enhancing Radiologic Image Analysis through Contrastive Language-Image Pre-trainingcited by 8
  3. 2023Generalizable Local Feature Pre-training for Deformable Shape Analysiscited by 6
  4. 2023The Performance of Transferability Metrics does not Translate to Medical Taskscited by 6
  5. 2025Self-Supervision Enhances Instance-based Multiple Instance Learning Methods in Digital Pathology: A Benchmark Studycited by 6
  6. 2022Generating and Weighting Semantically Consistent Sample Pairs for Ultrasound Contrastive Learningcited by 5
  7. 2023DeDA: Deep Directed Accumulatorcited by 5
  8. 2023Assessing Test-time Variability for Interactive 3D Medical Image Segmentation with Diverse Point Promptscited by 5
  9. 2023Shifting to Machine Supervision: Annotation-Efficient Semi and Self-Supervised Learning for Automatic Medical Image Segmentation and Classificationcited by 5
  10. 2024Part-aware Prompted Segment Anything Model for Adaptive Segmentationcited by 5
  11. 2024MedMerge: Merging Models for Effective Transfer Learning to Medical Imaging Taskscited by 5
  12. 2024Scaling Laws for Galaxy Imagescited by 5
  13. 2022Can we Adopt Self-supervised Pretraining for Chest X-Rays?cited by 4
  14. 2023Embrace Limited and Imperfect Training Datasets: Opportunities and Challenges in Plant Disease Recognition Using Deep Learningcited by 4
  15. 2023Training Transitive and Commutative Multimodal Transformers with LoReTTacited by 4
  16. 2022Rethinking Two Consensuses of the Transferability in Deep Learningcited by 3
  17. 2023AViT: Adapting Vision Transformers for Small Skin Lesion Segmentation Datasetscited by 3
  18. 2024FissionFusion: Fast Geometric Generation and Hierarchical Souping for Medical Image Analysiscited by 3
  19. 2026GloCTM: Cross-Lingual Topic Modeling via a Global Context Spacecited by 3
  20. 2023Enhancing Medical Image Segmentation: Optimizing Cross-Entropy Weights and Post-Processing with Autoencoderscited by 2
  21. 2023How Can We Tame the Long-Tail of Chest X-ray Datasets?cited by 2
  22. 2023Bridging Generalization Gaps in High Content Imaging Through Online Self-Supervised Domain Adaptationcited by 2
  23. 2024B-Cos Aligned Transformers Learn Human-Interpretable Featurescited by 2
  24. 2023Generalized Cross-domain Multi-label Few-shot Learning for Chest X-rayscited by 1
  25. 2024Pre-Trained Model Recommendation for Downstream Fine-tuningcited by 1
  26. 2024Multi-Scale Texture Loss for CT denoising with GANscited by 1
  27. 2024To Supervise or Not to Supervise: Understanding and Addressing the Key Challenges of Point Cloud Transfer Learningcited by 1
  28. 2024Self-supervised learning improves robustness of deep learning lung tumor segmentation to CT imaging differencescited by 1
  29. 2024Machine learning approach to brain tumor detection and classificationcited by 1
  30. 2025APLA: A Simple Adaptation Method for Vision Transformerscited by 1
  31. 2025k-NN as a Simple and Effective Estimator of Transferabilitycited by 1
  32. 2025Advancements in Medical Image Classification through Fine-Tuning Natural Domain Foundation Modelscited by 1
  33. 2023DAGrid: Directed Accumulator Gridcited by 0
  34. 2023Progressive Feature Adjustment for Semi-supervised Learning from Pretrained Modelscited by 0
  35. 2023TransMed: Large Language Models Enhance Vision Transformer for Biomedical Image Classificationcited by 0
  36. 2024BarlowTwins-CXR : Enhancing Chest X-Ray abnormality localization in heterogeneous data with cross-domain self-supervised learningcited by 0
  37. 2024Analysis of the Two-Step Heterogeneous Transfer Learning for Laryngeal Blood Vessel Classification: Issue and Improvementcited by 0
  38. 2024Domain Shift Analysis in Chest Radiographs Classification in a Veterans Healthcare Administration Populationcited by 0

50 most cited of 61 All 61 (JSON)