Application of Stacked Sparse Autoencoder in Automated Detection of Glaucoma in Fundus Images

In this contribution, intelligent identification of glaucoma from digital fundus images using stacked sparse autoencoder (SSAE) is proposed. The fundus images are initially converted to gray-scale and normalized w.r.t., background illuminance while maintaining contrast constancy across the dataset. Unfolded feature vectors from the pre-processed with proper rescaling and grays-scale converted fundus images are fed to SSAE for learning efficient feature representation and classification thereof using a softmax layer. A comparative evaluation highlighting the superiority of SSAE method with existing state-of the art techniques is presented to validate its efficacy in glaucoma detection. The proposed framework can be used as a clinical decision support system assisting ophthalmologists in confirming their diagnosis with high reliability & accuracy.