Earlier detection of rop and rop sub-classification using TA-CHD-DNN and UD-TFCM from ultrasound digital B-Scan images

Abstract

The growth of the retinal blood vessels of the baby begins at 16 weeks and even after the birth of the baby, they don’t grow completely. For the effective analysis of Retinopathy of Prematurity (ROP), diverse methodologies are employed. Owing to the insufficient growth of tangled patterns, greater retinal layer thickness, and the presence of hyperreflective material, the majority of the traditional methodologies were unsuccessful in classifying the Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and DRUSEN stages of ROP. Thus, for the classification of different stages of the ROP, an effective Twin Active- Cosine-Hausdorff Distance-Deep Neural Network (TA-CHD-DNN) is proposed in this paper. Chiefly, the image obtained from the retinal image dataset is pre-processed and fed into TA- CHD-DNN. After the extraction of critical features, the segmentation of the retinal layer, choroid layer, and vascular layer occurs in TA-CHD- DNN. The classifier is trained to make decisions for the classification of diverse stages of ROP centered on the extracted features. The normal or abnormal stages of CNV, DME, and DRUSEN are classified by the classifier during testing. By utilizing the Uniformly Distributed-Trapezoidal Fuzzy C Means (UD-TFCM) technique, the decision regarding the severity stage of CNV, DME, and DRUSEN is made in the abnormal stage. Thus, centered on the experimental outcomes, the proposed system performance will be analogized to the conventional techniques to confirm the proposed system’s efficiency.