The esophageal cancer is the sixth most lethal cancer with an elevated mortality rate. It is the fastest-growing form of cancer globally. The Early Esophageal Cancer diagnosis is a challenging task for the clinicians. There is a 5- fold increase in Esophageal Cancer patients initially diagnosed with Esophagitis. The persistence of Esophagitis can lead to Barrett’s Esophagus and then to Esophageal Cancer. Barrett’s Esophagus is one of the key precursors for the development of EC. The Convolution Neural Network has a significant role in the diagnosis of Early Esophageal Cancer. Most of the pre-trained networks can be efficiently used in Computer-Aided Diagnosis of cancer diseases. The proposed model performs the classification of the precancerous stage, Barrett’s Esophagus and Esophagitis which is asymptomatic in nature. The proposed work concentrates on the use of the blending of attributes acquired from several Deep Learning networks. A Correlation based Feature Selection is utilized for selecting the relevant attributes using suitable search algorithms. The selected feature set is then classified using a Bayesnet classifier. The proposed multi-CNN model outperforms all the existing methods with an accuracy of 99.61% and an AUC of 100%. We can infer that the multi-CNN model is more efficient than the models with individual pretrained networks from the experimental analysis.

Barrett’s Esophagus, Bayesnet Classifier, Esophagitis, Feature Selection, Multi -CNN networks