SMU Data Science Review


In this paper we present a lightweight solution to help iden- tify a pathological condition called pleural effusion using chest x-rays (CXR). Patients with pleural effusion have been found to have increased mortality rates, and if left undiagnosed effusion has been found to con- tribute to congestive heart failure, malignancy, pulmonary embolism, and tuberculosis [15] [13]. Using convolutional neural network architectures we developed a model to assist in the successful diagnosis of pleural ef- fusion. The effectiveness of our model was evaluated against 200 studies manually labeled by consensus from 3 board certified radiologist. We demonstrate that our model is able to reproduce current baseline perfor- mance for this task with a model that is 10x smaller and 30x faster. This lighter architecture allows for more flexibility in deployment including the ability to deploy directly on an edge node. We present this model as a tool for the radiologists to diagnose the presence of pleural effusion from a diagnostic imaging study.

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License