Figure 2 | Scientific Reports

Figure 2

From: Distinguishing retinal angiomatous proliferation from polypoidal choroidal vasculopathy with a deep neural network based on optical coherence tomography

Figure 2

Confusion matrixes of the eight ophthalmologists and the proposed model. (a) Normal retina vs. neovascular AMD. Confusion matrixes showing how the eight ophthalmologists and the model classified the test set. Five of the eight ophthalmologists have an accuracy of more than 99%, whereas the model shows 99.1% accuracy. (b) Confusion matrixes of distinguishing between PCV and RAP cases. The average accuracy of the eight ophthalmologists is 83.0%, and our proposed model has an accuracy score of 89.1%. R1, R2, and R3 denote ophthalmology residents with 1 year, 3 years, and 4 years of experience, respectively. F1, F2, and F3 denote retina fellows with 2 years of experience. RE1 and RE2 refer to retina experts with more than 10 years of experience at a retina clinic.

Back to article page