Calculation of ophthalmic diagnostic parameters on a single eye image based on deep neural network

It is necessary to manually measure many parameters of eyes when an ophthalmologist diagnoses, which is time consuming, unsanitary, subjective and unrepeatable. Those manually achieved parameters often risk clinical trials in challenges on objectivity, resulting in unreliable clinical conclusions. We designed a two-phase algorithm to automatically measure these parameters instead of manual measurement to facilitate the diagnosis procedure of ptosis, eyelid retraction and other eye pathologies. Firstly, cornea, sclera, and internal and external canthus were identified using a multi-task convolutional neural network (CNN). Then we used the identification results to calculate a series of eye parameters needed in clinic. Totally 9 widely accepted parameters were calculated, including the height of ocular fissure, the longitudinal and transverse diameters of cornea, etc. The experimental results showed that most of parameters had achieved good accuracy, and most errors were less than 2 mm. We found that our approach can improve the measurements accuracy of eyelid, conjunctiva, cornea, sclera, and internal and external canthus, and promote efficiency and efficacy of relative clinical researches.