Eye Tumour Detection Using Deep Learning

Abstract

Iris melanocytic tumours, are the most dangerous tumours in the eye , commonly known as eye tumours. This includes freckle, nevus, melanocytoma, Lisch nodule, and melanoma. The detection of eye tumour is very difficult in early stages. Many research works are being carried out to detect eye diseases. But few research works in the eye tumour were published. Most of the system needs specific data acquisition devices to capture the region. This is very expensive. To diagnose eye melanoma, doctors recommend PET - CT, eye ultrasound, angiogram, optical coherence tomography, etc. Here, a new approach is presented to detect the eye tumour from eye images using deep learning technique. The deep network model created with modified LeNet architecture. The model created with the segmented eyeball images. Hough circle transformation could predict the eyeball and iris regions. As the deep learning technique needs more data for training, the number of image data has been increased with image augmentation method. Successful testing of this method with an accuracy of 95% shows that this method can be implemented in real time applications.