Improve the performance of transfer learning without fine-tuning using dissimilarity-based multi-view learning for breast cancer histology images
This work addresses breast cancer diagnosis for medical imaging, but it is incremental as it applies existing methods to a specific dataset.
The paper tackled breast cancer histology image classification by comparing handcrafted and deep learning feature extractors, finding deep learning pretrained on ImageNet outperformed handcrafted features with up to 79.30% average accuracy. Using a random forest dissimilarity-based integration method improved accuracy to 87.10% when combining all features.
Breast cancer is one of the most common types of cancer and leading cancer-related death causes for women. In the context of ICIAR 2018 Grand Challenge on Breast Cancer Histology Images, we compare one handcrafted feature extractor and five transfer learning feature extractors based on deep learning. We find out that the deep learning networks pretrained on ImageNet have better performance than the popular handcrafted features used for breast cancer histology images. The best feature extractor achieves an average accuracy of 79.30%. To improve the classification performance, a random forest dissimilarity based integration method is used to combine different feature groups together. When the five deep learning feature groups are combined, the average accuracy is improved to 82.90% (best accuracy 85.00%). When handcrafted features are combined with the five deep learning feature groups, the average accuracy is improved to 87.10% (best accuracy 93.00%).