Embedded mobile computational framework for multidimensional diabetic retinopathy extraction and detection technique using recursive neural network approach for unstructured tomography datasets

Abstract

Diabetic Retinopathy (DR) is considered to be the leading cause for preventive blindness in humans, the DR is sighted with a diabetic stage of progression and hence the patient is required to undergo regular health checkups on DR formation and detection. In this paper, the objective is to extract and detect the patterns of DR with respect to the propagation stages using Recursive Neural Network (RNN). In this work, we have developed and validated a novel Inter-Correlated Attribute Coordination (ICAC) Technique for attribute based feature mapping and feature inter-dependent cluster generation. The ICAC technique generates a series of standard dataset attributes [Formula: see text] for process alignment towards the generation of feature set (f). The proposed technique has validated the categorization of DR into grade 1 and grade 0 patients for an unambiguous decision making. The technique’s trained datasets provide a self-learning RNN for multidimensional tomography dataset processing. The ICAC technique has developed a detection rate of 97.3% for the 276 feature set clusters.