Comparative analysis of neural model and statistical model for abnormal retinal image segmentation

Abstract

Artificial Neural Networks (ANN) are gaining significant importance in themedical field, especially in the area of ophthalmology. Though the performance of ANN is theoretically stated, the practical applications of ANN are not fully explored. In this work, the suitability of Back Propagation Neural Network (BPN) for ophthalmologic applications is highlighted in the context of retinal blood vessel segmentation. The neural technique is tested with Diabetic Retinopathy (DR) images. The performance of the BPN is compared with the k-Nearest Neighbor (k-NN) classifier which is a statistical classifier. Experimental results verify the superior nature of the BPN over the k-NN approach