Adaptive isomap feature extractive gradient deep belief network classifier for diabetic retinopathy identification

Singh, A; Kumar, R; Gandomi, AH

Adaptive isomap feature extractive gradient deep belief network classifier for diabetic retinopathy identification

Singh, A Kumar, R Gandomi, AH

Permalink

Publisher:: Springer Nature
Publication Type:: Journal Article
Citation:: Multimedia Tools and Applications, 2024, pp. 1-22
Issue Date:: 2024-01-01

Closed Access

	Filename	Description	Size
	s11042-024-19216-6.pdf	Published version	2.05 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Singh, A
dc.contributor.author	Kumar, R
dc.contributor.author	Gandomi, AH
dc.date.accessioned	2024-11-24T23:24:22Z
dc.date.available	2024-11-24T23:24:22Z
dc.date.issued	2024-01-01
dc.identifier.citation	Multimedia Tools and Applications, 2024, pp. 1-22
dc.identifier.issn	1380-7501
dc.identifier.issn	1573-7721
dc.identifier.uri	http://hdl.handle.net/10453/182092
dc.description.abstract	Diabetic retinopathy (DR) is an infection that bases eternal visualization loss in patients with diabetes mellitus. With DR, the glucose level in the blood increases, as well as its viscosity, this results in fluid leakage into surrounding tissues in the retina. In other words, DR represents the pathology of capillaries and venules in the retina with leakage effects, being the main acute retinal disorder caused by diabetes. Many DR detection methods have been previously discussed by different researchers; however, accurate DR detection with a reduced execution time has not been achieved by existing methods. The proposed method, the Shape Adaptive box linear filtering-based Gradient Deep Belief network classifier (SAGDEB) Model, is performed to enhance the accuracy of DR detection. The objective of the SAGDEB Model is to perform an efficient DR identification with a higher accuracy and lower execution time. This model comprises three phases: pre-processing, feature extraction, and classification. The shape adaptive box linear filtering image pre-processing is carried out to reduce the image noise without removing significant parts of image content. Then, an isomap geometric feature extraction is performed to compute features of different natures, like shape, texture, and color, from the pre-processed images. After that, the Adaptive gradient Tversky Deep belief network classifier is to perform classification. The deep belief network is probabilistic and generative graphical model that consists of multiple layers such as one input unit, three hidden units, and one output unit. The extracted image featuresare considered in the input layer and these images are sent to hidden layers. Tversky similarity index is applied in hidden layer 1 to analyze the extracted features with testing features. Regarding the similarity value, the sigmoid activation function is determined in hidden layer 2 so different levels of DR can be identified. Finally, the adaptive gradient method is applied in hidden layer 3 to minimize the error. Finally, the classification results are obtained at the output layer. The results were achieved by using a retinal image dataset and the following metrics were analyzed: peak signal-to-noise ratio (PSNR), DR detection accuracy, error rate, and DR detection time. The quantitative analysis results show that the proposed SAGDEB model achieves a higher 6% PSNR, 5% DR detection accuracy, and a lesser 46% error rate, and 13% DR detection time as compared to the two existing methods.
dc.language	en
dc.publisher	Springer Nature
dc.relation.ispartof	Multimedia Tools and Applications
dc.relation.isbasedon	10.1007/s11042-024-19216-6
dc.rights	info:eu-repo/semantics/closedAccess
dc.subject	0801 Artificial Intelligence and Image Processing, 0803 Computer Software, 0805 Distributed Computing, 0806 Information Systems
dc.subject.classification	Artificial Intelligence & Image Processing
dc.subject.classification	Software Engineering
dc.subject.classification	4009 Electronics, sensors and digital hardware
dc.subject.classification	4603 Computer vision and multimedia computation
dc.subject.classification	4605 Data management and data science
dc.subject.classification	4606 Distributed computing and systems software
dc.title	Adaptive isomap feature extractive gradient deep belief network classifier for diabetic retinopathy identification
dc.type	Journal Article
utslib.for	0801 Artificial Intelligence and Image Processing
utslib.for	0803 Computer Software
utslib.for	0805 Distributed Computing
utslib.for	0806 Information Systems
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/UTS Groups
pubs.organisational-group	University of Technology Sydney/UTS Groups/Data Science Institute (DSI)
pubs.organisational-group	University of Technology Sydney/Strength - DSI - Data Science Institute
utslib.copyright.status	closed_access	*
dc.date.updated	2024-11-24T23:24:21Z
pubs.publication-status	Published

Abstract:

Diabetic retinopathy (DR) is an infection that bases eternal visualization loss in patients with diabetes mellitus. With DR, the glucose level in the blood increases, as well as its viscosity, this results in fluid leakage into surrounding tissues in the retina. In other words, DR represents the pathology of capillaries and venules in the retina with leakage effects, being the main acute retinal disorder caused by diabetes. Many DR detection methods have been previously discussed by different researchers; however, accurate DR detection with a reduced execution time has not been achieved by existing methods. The proposed method, the Shape Adaptive box linear filtering-based Gradient Deep Belief network classifier (SAGDEB) Model, is performed to enhance the accuracy of DR detection. The objective of the SAGDEB Model is to perform an efficient DR identification with a higher accuracy and lower execution time. This model comprises three phases: pre-processing, feature extraction, and classification. The shape adaptive box linear filtering image pre-processing is carried out to reduce the image noise without removing significant parts of image content. Then, an isomap geometric feature extraction is performed to compute features of different natures, like shape, texture, and color, from the pre-processed images. After that, the Adaptive gradient Tversky Deep belief network classifier is to perform classification. The deep belief network is probabilistic and generative graphical model that consists of multiple layers such as one input unit, three hidden units, and one output unit. The extracted image featuresare considered in the input layer and these images are sent to hidden layers. Tversky similarity index is applied in hidden layer 1 to analyze the extracted features with testing features. Regarding the similarity value, the sigmoid activation function is determined in hidden layer 2 so different levels of DR can be identified. Finally, the adaptive gradient method is applied in hidden layer 3 to minimize the error. Finally, the classification results are obtained at the output layer. The results were achieved by using a retinal image dataset and the following metrics were analyzed: peak signal-to-noise ratio (PSNR), DR detection accuracy, error rate, and DR detection time. The quantitative analysis results show that the proposed SAGDEB model achieves a higher 6% PSNR, 5% DR detection accuracy, and a lesser 46% error rate, and 13% DR detection time as compared to the two existing methods.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/182092