A Novel Transfer Learning Approach for Eye Tumour Detection using Pre-Trained CNN Models

According to a recent study, using retinal imaging to identify the tumour in the eye is difficult. When an image is analyzed, the affected region of the lesion impacted retinal picture is not discovered in the conventional system due to uneven illumination, low contrast, and blurring, which are symptoms of an eye disease. The proposed methodology was created to identify the eye tumour in the earlier stage with enhanced image quality and increased accuracy. The conventional system has a few drawbacks such as lower accuracy and more computational time to analyze the image. Convolutional neural networks (CNN) have recently demonstrated its result of their capacity to learn various category image denotions that aid in image classification in various sectors. These networks have a tremendous ability to extract the proper features from input images with training on millions of images, which produces accurate categorization. The deep learning-based transfer learning approach is suggested for earlier detection of eye tumours. The five pre-trained models such as AlexNet, ResNet, DensNet, VGG 19 and Inception model were used for efficient eye tumour classification. The performance analysis of the proposed approach has shown that higher accuracy of 98.63% and this model shows enhanced performance in better classification.