Investigations of CNN for Medical Image Analysis for Illness Prediction

Table 10

A comprehensive view of classification and preprocessing in a DCNN with the proposed GMPR-PReLU along with the efficacies of GLCM, AHE, and CLAHERD in the proposed DCNN.

GLCM with GMPR-PReLU	AHE with GMPR-PReLU	CLAHERD with GMPR-PReLU
Image is converted to gray-scale image	Image is converted to HSV array	Image is not disturbed and its contrast values of RGB values for relevant colors are extracted
Light colors of pixels are confused with other symptoms	Value of pixel does not signify the exact contrast of the required pixels	Required objects of the images are extracted exactly, since colors codes are applied
Ambiguity of identifying exudates	Color loss due to high intensity of value	Color is intact, and objects are selected
Two-value histogram is drawn and does not signify the existence of exudate symptoms	Full-color histogram is drawn, difficult to distinguish the objects with exudates	As only objects with exudate are developed, histogram signifies the intensity of exudates
Not possible to distinguish objects	Possible distinction of objects with much aberration	Objects are distinguished with very slight aberration

Total population: 89; samples: 35; average samples size: 40.