Raven Roosting Optimized based Deep Belief Network for Esophageal Adenocarcinoma Prediction

  • G. Vani, Dr.A. Hema
Keywords: Deep Learning Model, Adenocarcinoma Esophagus, Deep Belief Network, Raven Roosting Optimization, Restricted Boltzmann Machine, fine tuning

Abstract

Early prediction of adenocarcinoma esophagus is very challenging because of the impreciseness of its spread all over its esophagus tube. The compute aided diagnosis along deep learning paradigm improve the endoscopic assessment and presence of Adenocarcinoma Esophagus. Still, the parameter used in the standard deep learning are often based on the gradient descent based learning which often results in overfitting when there is a high degree of class imbalance in adenocarcinoma esophageal image detection. This paper aims to construct a novel adenocarcinoma esophagus prediction using Barrett’s esophagus image at its earlier stages. This work introduced a novel deep belief network to predict the presence of adenocarcinoma in esophagus images. The stacked Restricted Boltzmann Machine learns the pattern of the Barrett’s Esophagus images using their belief network structure with two different layers Visible and Hidden. The pre-trained information of the esophagus image using the quantum grid-based clustering extracts the significant features and it is fed as input to the DBN. The parameters of the Deep Belief Network (DBN) is fine-tuned using the behavioral inspiration memetic model known as Raven Roosting Optimization (RRO) Algorithm. The strategy of Ravens food searching strategy by computing the best solutions based on the best fittest value. The roosting site and the current position update based on optimized food resource searching is used for optimized weight value selection. The learning rate of the DBN is mainly relies on the weight values assigned to the hidden nodes to activate the appropriate output and to reduce the error rate. The simulation results proved that the proposed RRO–DBN produce best accuracy rate in early detection of adenocarcinoma esophagus while comparing with other deep learning models. The parameter fine tuning is the essential factor which improves the functionality of the Deep Belief network by inducing the Raven Roosting Optimization.

Published
2021-09-20
How to Cite
Dr.A. Hema, G. V. (2021). Raven Roosting Optimized based Deep Belief Network for Esophageal Adenocarcinoma Prediction. Design Engineering, 12887-12900. Retrieved from http://www.thedesignengineering.com/index.php/DE/article/view/4515
Section
Articles