# Understanding the Mechanisms of Deep Transfer Learning for Medical Images

arXiv: https://arxiv.org/abs/1704.06040 · HTML: https://mlsanity.com/paper/1704.06040/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 20 Apr 2017.

## Built on by (most cited first, 10 of 10)

- [A Review of Deep Transfer Learning and Recent Advancements](https://mlsanity.com/paper/2201.09679/) (2022)
- [Parametric Scattering Networks](https://mlsanity.com/paper/2107.09539/) (2021)
- [A comparative study between vision transformers and CNNs in digital pathology](https://mlsanity.com/paper/2206.00389/) (2022)
- [ADAPT : Awesome Domain Adaptation Python Toolbox](https://mlsanity.com/paper/2107.03049/) (2021)
- [MRI Images, Brain Lesions and Deep Learning](https://mlsanity.com/paper/2101.05091/) (2021)
- [An overview of deep learning in medical imaging](https://mlsanity.com/paper/2202.08546/) (2022)
- [EXPANSE: A Deep Continual / Progressive Learning System for Deep Transfer Learning](https://mlsanity.com/paper/2205.10356/) (2022)
- [A Data Augmentation Method and the Embedding Mechanism for Detection and Classification of Pulmonary Nodules on Small Samples](https://mlsanity.com/paper/2303.12801/) (2023)
- [Transfer learning from a sparsely annotated dataset of 3D medical images](https://mlsanity.com/paper/2311.05032/) (2023)
- [Domain Knowledge Based Temporal-Spatial Graph Convolution Network for ECG Recognition](https://mlsanity.com/paper/2607.01282/) (2026)

## Abstract (arXiv)

The ability to automatically learn task specific feature representations has led to a huge success of deep learning methods. When large training data is scarce, such as in medical imaging problems, transfer learning has been very effective. In this paper, we systematically investigate the process of transferring a Convolutional Neural Network, trained on ImageNet images to perform image classification, to kidney detection problem in ultrasound images. We study how the detection performance depends on the extent of transfer. We show that a transferred and tuned CNN can outperform a state-of-the-art feature engineered pipeline and a hybridization of these two techniques achieves 20% higher performance. We also investigate how the evolution of intermediate response images from our network. Finally, we compare these responses to state-of-the-art image processing filters in order to gain greater insight into how transfer learning is able to effectively manage widely varying imaging regimes.