Room: AAPM ePoster Library
Kermany et al. (2018) classified optical coherence tomography (OCT) images, which are often used for ophthalmology, into four classes using a classical convolution neural network (CNN), a commonly used deep learning network for computer-aided diagnosis. Classical CNN has often been criticized for suppressing positional relations in a pooling layer. Therefore, we attempted application of a capsule network to OCT images to overcome that shortcoming because capsule networks can learn positional information from images. This study is our attempt to improve classification accuracy by replacing CNN with a capsule network.
From an OCT dataset, we produced a training dataset of 83,484 images and a test dataset of 1,000 images. For training, the dataset comprises 37,205 images with choroidal neovascularization (CNV), 11,348 with diabetic macular edema (DME), 8,616 with drusen, and 26,315 normal images. The test dataset has 250 images from each category. The proposed model was constructed based on a capsule network for improving the classification accuracy and was trained using the training dataset. Subsequently, the test dataset was used to evaluate the trained model.
Classification of OCT images with our method achieved accuracy of 99.6%, which is 3.2 percentage points higher than that of methods described in the literature.
The proposed method achieved classification accuracy results equivalent to those reported for other methods for CNV, DME, drusen and normal images.
Funding Support, Disclosures, and Conflict of Interest: This work was partly supported by a Japan Society for the Promotion of Science (JSPS) KAKENHI Grant (No. 18K07646).