Abstract
One of the main challenges for broad adoption of deep learning based models such as convolutional neural networks (CNN), is the lack of understanding of their decisions. In many applications, a simpler, less capable model that can be easily understood is favorable to a black-box model that has superior performance. In this paper, we present an approach for designing CNNs based on visualization of the internal activations of the model. We visualize the model's response through attentive response maps obtained using a fractional stride convolution technique and compare the results with known imaging landmarks from the medical literature. We show that sufficiently deep and capable models can be successfully trained to use the same medical landmarks a human expert would use. Our approach allows for communicating the model decision process well, but also offers insight towards detecting biases.
Original language | English |
---|---|
Title of host publication | 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018 |
Place of Publication | Piscataway |
Publisher | IEEE Computer Society |
Pages | 1130-1133 |
Number of pages | 4 |
ISBN (Electronic) | 978-1-5386-3636-7 |
DOIs | |
Publication status | Published - 23 May 2018 |
Event | 15th IEEE International Symposium on Biomedical Imaging (ISBI 2018) - Omni Shoreham Hotel, Washington, United States Duration: 4 Apr 2018 → 7 Apr 2018 Conference number: 15 https://biomedicalimaging.org/2018/ |
Conference
Conference | 15th IEEE International Symposium on Biomedical Imaging (ISBI 2018) |
---|---|
Abbreviated title | ISBI18 |
Country/Territory | United States |
City | Washington |
Period | 4/04/18 → 7/04/18 |
Internet address |
Keywords
- Anatomy
- CNN
- Deep Learning
- Visualization