Diabetic retinopathy (DR) is the leading cause of blindness among adults worldwide. This project encompasses the development of an artificial intelligence convolutional neural network (CNN) that learns to diagnose DR at its earliest stages. The multi-class fundoscopic retinal image dataset is trained on several networks. The loss function is computed to determine the accuracy of the predictions. The network has learned via stochastic gradient descent and backpropagation, exhibiting a performance accuracy of 96.2% post-data augmentation. While being implemented in a mobile and web framework for clinical evaluations, this outperforms current diagnostic methods in literature and benefits the worldwide medical field.

Infos

Participants

• Jordan Levett

Année : 2019

Région : Montreal Regional Science & Technology Fair (Sec/Coll)

Type de projet : E

Classe : S2

Categorie de projet : SBSS

Volet : Secondaire

Niveau scolaire : Secondaire 5

École : Herzliah High School

Prix et distinctions

Finale Québécoise

Longueuil,Collège Charles-Lemoyne

• Exposciences Internationale du Milset à Abu Dhabi 2019
• Médaille du Réseau Technoscience - Argent Senior
• Prix de reconnaissance Francis-Boulva

Photo