We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.IV

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: DiaRet: A browser-based application for the grading of Diabetic Retinopathy with Integrated Gradients

Abstract: Diabetes is a metabolic disorder that results from defects in autoimmune beta-cell destruction in Type 1, peripheral resistance to insulin action in Type 2 or, most commonly, both. Patients with long-standing diabetes often fall prey to Diabetic Retinopathy (DR) resulting in changes in the retina of the human eye, which may lead to loss of vision in extreme cases. The aim of this study is two-fold: (a) create deep learning models that were trained to grade degraded retinal fundus images and (b) to create a browser-based application that will aid in diagnostic procedures by highlighting the key features of the fundus image. Deep learning has proven to be a success for computer-aided DR diagnosis resulting in early-detection and prevention of blindness. In this research work, we have emulated the images plagued by distortions by degrading the images based on multiple different combinations of Light Transmission Disturbance, Image Blurring and insertion of Retinal Artifacts. These degraded images were used for the training of multiple Deep Learning based Convolutional Neural Networks. We have trained InceptionV3, ResNet-50 and InceptionResNetV2 on multiple datasets. These models were used to classify the fundus images in terms of DR severity level. The models were further used in the creation of a browser-based application, which demonstrates the models prediction and the probability associated with each class. It will also show the Integration Gradient (IG) Attribution Mask superimposed onto the input image. The creation of the browser-based application would aid in the diagnostic procedures performed by ophthalmologists by highlighting the key features of the fundus image based on an educated prediction made by the model.
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2103.08501 [eess.IV]
  (or arXiv:2103.08501v1 [eess.IV] for this version)

Submission history

From: Samyak Prajapati [view email]
[v1] Mon, 15 Mar 2021 16:19:56 GMT (476kb)
[v2] Tue, 16 Mar 2021 19:26:55 GMT (465kb)
[v3] Sun, 11 Apr 2021 16:21:45 GMT (406kb)
[v4] Fri, 13 Aug 2021 20:16:15 GMT (408kb)

Link back to: arXiv, form interface, contact.