Healthy Vessel Wall Detection Using U-Net in Optical Coherence Tomography
Intravascular optical coherence tomography can be applied for high-resolution imaging in the coronary arteries with ischemic heart disease. The differentiation of the healthy and diseased vessel wall can be used to assess the extent and severity of coronary artery disease, and to guide therapeutic interventions. The aim of this study is to develop a recognition framework that can be potentially used for real-time intraoperative application. Structures in an image were labeled into five categories: diseased, healthy, luminal, guide-wire, and others. A U-net was implemented to directly take Cartesian images as input without any additional processing steps. A sigmoid activation and binary cross-entropy loss were applied to perform multi-labeling segmentation. Three transformations were specifically proposed in the polar domain for data augmentation. For evaluation of the proposed framework, 200 images from 20 patients were used and a triple-leave-2-out cross validation was carried out. Performance was evaluated using the average loss in the validation dataset, and the Dice scores were reported as well. Results showed that the proposed framework can perform the segmentation generally with an average performance of 0.88 ± 0.02 in Dice scores. These preliminary results suggest that the proposed framework can be potentially applied for assisting diagnosis in real-time. In the future, we intend to include more data, also take into consideration artifacts such as bad flushing, more deformed lumen, and side-branches.
KeywordsOCT Tissue recognition Coronary artery disease
- 1.Tearney, G.J., et al.: Consensus standards for acquisition, measurement, and reporting of intravascular optical coherence tomography studies: a report from the international working group for intravascular optical coherence tomography standardization and validation. J. Am. Coll. Cardiol. 59, 1058–1072 (2012)CrossRefGoogle Scholar
- 4.Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedical image segmentation. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9351, pp. 234–241. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-24574-4_28CrossRefGoogle Scholar
- 5.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 (ICML 2013) (2013)Google Scholar