Automated Pneumothorax Diagnosis Using Deep Neural Networks
Thoracic ultrasound can provide information leading to rapid diagnosis of pneumothorax with improved accuracy over the standard physical examination and with higher sensitivity than anteroposterior chest radiography. However, the clinical interpretation of a patient medical image is highly operator dependent. Furthermore, remote environments, such as the battlefield or deep-space exploration, may lack expertise for diagnosing certain pathologies. We have developed an automated image interpretation pipeline for the analysis of thoracic ultrasound data and the classification of pneumothorax events to provide decision support in such situations. Our pipeline consists of image preprocessing, data augmentation, and deep learning architectures for medical diagnosis. In this work, we demonstrate that robust, accurate interpretation of chest images and video can be achieved using deep neural networks. A number of novel image processing techniques were employed to achieve this result. Affine transformations were applied for data augmentation. Hyperparameters were optimized for learning rate, dropout regularization, batch size, and epoch iteration by a sequential model-based Bayesian approach. In addition, we utilized pretrained architectures, applying transfer learning and fine-tuning techniques to fully connected layers. Our pipeline yielded binary classification validation accuracies of \(98.3\%\) for M-mode images and \(99.8\%\) with B-mode video frames.
KeywordsDeep learning Pneumothorax classification Ultrasound Transfer learning Bayesian optimization
We acknowledge Jose Salinas, PhD, chief of the Clinical Decision Support and Automation research at the US Army Institute of Surgical Research, which provided laboratory and biomedical device equipment support. We also acknowledge Maria Serio-Melvin, MSN, and Army Col. Shawn Nessen, DO, FACS, for their invaluable trauma patient clinical insights.
- 2.Authors, T.G.: GPyOpt: a Bayesian optimization framework in python (2016). http://github.com/SheffieldML/GPyOpt
- 3.Bergstra, J., Yamins, D., Cox, D.D.: Making a science of model search: hyperparameter optimization in hundreds of dimensions for vision architectures. In: Proceedings of the 30th International Conference on Machine Learning, vol. 28, no. 1, pp. 115–123, June 2013Google Scholar
- 5.Oquab, M., Bottou, L., Laptev, I., Sivic, J.: Learning and transferring mid-level image representations using convolutional neural networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1717–1724 (2014)Google Scholar