Semantic Segmentation of Optic Disc and Optic Cup using Deep Learning

Most serious eye disease that causes visual loss is glaucoma. The individual will benefit from getting better therapy and illness management if they are diagnosed with glaucoma early. In this context,a neural network aided medical image analysis system can serve as a decision-support tool for the ophthalmologist, lowering the potential for subjective interpretations while boosting accuracy. Consequently, the identification of glaucoma heavily relies on the optic cup and disc segmentation from fundus images. As a result, a modified U-Net model is proposed in the current study with an emphasis on the combined semantic segmentation of the optic disc and optic cup using fundus pictures. This approach makes use of a modified Atrous Spatial Pyramid Pooling (ASPP) module that captures more extensive spatial data for precise segmentation of these regions of interest. Additionally, an experiment is conducted to assess the impact of different loss functions on the effectiveness of the model. Both qualitative and quantitative analysis of the proposed method is carried out on public data set which suggests promising results.