The application of deep learning framework in quantifying retinal structures on ophthalmic image in research eye-PACS

摘要

The rise of deep learning (DL) framework and its application in object recognition could benefitimage-based medical diagnosis. Since eye is believed to be a window into human health, theapplication of DL on differentiating abnormal ophthalmic photography (OP) will greatly empowerophthalmologists to relieve their workload for disease screening. In our previous work, we employedResNet-50 to construct classification model for diabetic retinopathy(DR) within the PACS. In thisstudy, we implemented latest DL object detection and semantic segmentation framework to empowerthe eye-PACS. Mask R-CNN framework was selected for object detection and instance segmentationof the optic disc (OD) and the macula. Furthermore, Unet framework was utilized for semanticsegmentation of retinal vessel pixels from OP. The performance of the segmented results by twoframeworks achieved state-of-art efficiency and the segmented results were transmitted to PACS asgrayscale softcopy presentation state (GSPS) file. We also developed a prototype for OP quantitativeanalysis. It’s believed that the implementation of DL framework into the object recognition andanalysis on Ops is meaningful and worth further investigation.