← Back to Explore

Supervised Binary Convolutional Neural Networks with Model Stacking for Diagnosis of Fundus & Eyelid Diseases

JSHS · 2023

Overview

Retinal health is often overlooked and seen as a privilege, not a necessity. Although there are a variety of machine learning-based solutions and retinal specialists to diagnose these ailments, they require expensive equipment and resources, and oftentimes overlook specific retinal diseases. The following comprehensive paper details the creation of six binary convolutional neural networks to diagnose Diabetic Retinopathy, Age-Related Macular Degeneration, Myopia, Glaucoma, Ocular Hypertension, and Cataracts with fundus images, along with a seventh Trachoma neural network through inner under eyelid images. T o produce a conclusive diagnosis, model stacking was implemented, and the classification results were compared to a trial multi-classification neural network. Analysis was also completed on the sklearn test data reports and confusion matrixes for the test dataset. The neural networks produced promising results for both training (90% +) and testing data accuracies (65% +) along with model classification and have the ability to identify new features in the fundus image. 40 After analyzing the convolutions, it was found that the models were focusing on the optic disc and fovea region of the fundus along with pigmentation and fundus spots, often overlooked with traditional diagnosis. Additionally, the Trachoma CNN may be able to provide further insights into what specific features of the inner eyelid to focus on when diagnosing this ailment. Traditionally, inner eyelid scarring and magnitude of the puss are used. However, the feature maps depict that pigmentation of eyelid scarring along with puss, eyelash rheum, and trichiasis [abnormal unhealthy eyelash placement] in the inner eyelid were fixated on.

Competition history

  • JSHS 2023 Category not listed

Resources

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Source: Junior Science and Humanities Symposium

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google