Mortality rate of breast cancer in the remote regions of the country is higher due to lack of scalable and affordable breast screening systems for clinical prediction of abnormality. Despite the high mortality rate of breast cancer, there has been very limited research into integrating screening techniques with machine learning approaches to provide sufficient expertise and real time communication to remote regions for in time diagnosis. An affordable and portable breast screening system using infrared imaging facilitate breast health monitoring in the developing part of the country. Computer-assisted framework using thermography breast images would greatly benefit healthcare professionals for a precise and timely diagnosis of abnormality in the thermograms. Convolutional Neural Network (CNN) has been shown to be the most conformable approach for the recognition of abnormality in thermography images. Despite their high sensitivity and predictive power, their clinical translation is hampered due to a lack of interpretable insights about the prediction. Here, we present an interpretable computer assisted expert system that facilitates clinical inference by accompanying a prediction with visual identification of the region of focus on the thermography images. We propose an interpretable computer assisted decision support system that integrate CNN feature extractor with Attention Branch Network (ABN) for the binary classification of thermography images. Our system comprised of (1) Image acquisition and formulate new dataset named Amrita Breast Thermogram (ABT) dataset using a thermal camera and associated software (2) Modified DarkNet19 model as a feature extractor to extract discriminative features from breast RoI (3) In order to achieve visual interpretation as well as boost the performance of the DarkNet19 model, an attention branch is integrated with DarkNet19 network forming Attention Branch Network (ABN). The attention branch uses a convolution layer of DarkNet19 and Global Average Pooling (GAP) to model the context of the visual features and generate a heatmap to identify the region of interest. Finally, the perception branch is constituted using a fully connected layer to classify images. (5) Mitigated dataset imbalance in thermography images through the implementation of sample weighting techniques, thereby alleviating model bias favoring the dominant class (6) The accesses and transmits the breast thermograms, predicted results, and patient’s history to the healthcare professionals in the tertiary hospitals for further diagnosis. We trained and validated our model with a new dataset, Amrita Breast Thermogram (ABT) dataset. The model exhibited 96.7 % accuracy using ABT dataset, demonstrating comparable performance in binary classification of thermography images. Our results show that introducing Attention Branch Network boosted the performance of the baseline DarkNet19 CNN model by 8% on ABT dataset. We submit that our proposed model can be viewed as a generic solution which provides state- of-the-art performance on two thermography datasets. Our decision support system, which combines ABN with the modified DarkNet19 model, not only provides cutting-edge performance but also diagnostic interpretability, allowing doctors to identify the region of focus from which the model made its decision making. This can potentially lead to improved clinical translation of deep learning models for clinical decision support.
Publication Details
Patent Details
Proposed Future Work Details
Developing a scalable solution that can be used for clinical use
