Interpretable Automated Diagnosis of Retinal Disease Using Deep OCT Analysis

30 million Optical Coherence Tomography (OCT) imaging tests are issued every year to diagnose various retinal diseases, but accurate diagnosis of OCT scans requires trained ophthalmologists who are still prone to making errors. With better systems for diagnosis, many cases of vision loss caused by retinal disease could be entirely avoided. In this work, we develop a novel deep learning architecture for explainable, accurate classification of retinal disease which achieves state-of-the-art accuracy. Furthermore, we place an emphasis on producing both qualitative and quantitative explanations of the model's decisions. Our algorithm produces heatmaps indicating the exact regions in the OCT scan the model focused on when making its decision. In combination with an OCT segmentation model, this allows us to produce quantitative breakdowns of the specific retinal layers the model focused on for later review by an expert. Our work is the first to produce detailed quantitative explanations of the model's decisions in this way. Our combination of accuracy and interpretability can be clinically applied for better patient care.