Accurate Detection Of Retinal Blastoma (Eye Cancer) Using Electroretinogram Erg - Image By Deep Learning (Vgg-19) Model

Abstract

Retinoblastoma is an uncommon yet life-threatening eye cancer that mostly affects children. Early and correct identification is crucial for avoiding visual loss and increasing survival chances. This paper describes an automated deep learning strategy for precisely detecting retinoblastoma using electroretinogram (ERG) data. The ERG picture files are analysed using a pre-trained VGG-19 model, which is noted for its exceptional feature extraction capabilities. The model uses transfer learning to tailor its architecture to the unique purpose of retinoblastoma detection, resulting in excellent sensitivity and specificity. To improve performance, the ERG picture dataset is pre-processed using techniques including normalisation, augmentation, and noise reduction, which improves the model's resilience and generalisability. The VGG-19 network accurately identifies pictures as healthy or ill, producing a high F1-score while minimising false positives and negatives. The findings show that the VGG-19 model has the potential to help ophthalmologists diagnose retinoblastoma earlier and more reliably. This work demonstrates the use of deep learning in medical imaging, demonstrating its effectiveness in detecting uncommon illnesses such as retinoblastoma.