Fig. 4: Workflows of human ophthalmologists and DeepIK in identifying the cause of keratitis. | npj Digital Medicine

Fig. 4: Workflows of human ophthalmologists and DeepIK in identifying the cause of keratitis.

From: Deep learning for multi-type infectious keratitis diagnosis: A nationwide, cross-sectional, multicenter study

Fig. 4

a The diagnostic pipeline of human ophthalmologists. When human ophthalmologists determine the cause of keratitis based on slit-lamp images, they often first distinguish whether keratitis is infectious or noninfectious. If infectious keratitis is identified, they will further analyze its specific type. b The analysis pipeline of DeepIK. The customized algorithm, DeepIK, which includes two classifiers, is employed to imitate the diagnostic process of human ophthalmologists. Classifier 1 is used to differentiate between infectious and noninfectious keratitis. Classifier 2 leveraged the prior knowledge learned from Classifier 1 to further classify keratitis into bacterial, fungal, viral, amebic, and noninfectious keratitis.

Back to article page