Multi-head deep learning framework for pulmonary disease detection and severity scoring with modified progressive learning

Chest X-Rays (CXR) are the most commonly used imaging methodology to diagnose pulmonary diseases with close to 2 billion CXRs taken every year. The recent upsurge of COVID-19 accompanied by pneumonia and tuberculosis can be fatal and lives could be saved through an early detection and appropriate intervention. Thus CXRs can be used for an automated severity grading that can aid the radiologists in making better and informed diagnosis. In this article, we propose a single framework for disease classification and severity scoring produced by segmenting the lungs into six regions. We present a modified progressive learning technique which caps the amount of augmentations at each step. Our base network is first trained using modified progressive learning and can then be tweaked for new datasets. Furthermore, the segmentation task makes use of attention map generated by the network itself. This attention mechanism achieves segmentation results that are on par with networks having far greater parameters. We also propose severity score grading for 4 thoracic diseases that can provide a single digit score corresponding to the spread of opacity in different lung segments with the help of radiologists. The proposed framework is evaluated using the BRAX data set for segmentation and classification into six classes with severity grading for a subset of the classes. On BRAX, we achieve F1 scores of 0.924 and 0.939 without and with fine-tuning. A mean matching score of 80.8% is obtained for severity score grading while an average AUCROC of 0.88 is achieved for classification.