DR_2021

KO | EN

gitlite — search

#hacktoberfest #ai #llm #go #golang #machine-learning #rust #macos #windows #awesome #vue #awesome-list #ios #openai

(★ 20)

Diabetic Retinopathy Feature Extraction and Binary Diagnosis

DR_2021 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1> Segmentation of Blood Vessels, Optic Disc Localization, Detection of Exudates and Diabetic Retinopathy Diagnosis from Digital Fundus Images </h1>

[Soham Basu](https://scholar.google.com/citations?user=VFrBx88AAAAJ&hl=en), [Sayantan Mukherjee](https://orcid.org/0000-0001-9385-7369), [Ankit Bhattacharya](https://orcid.org/0000-0003-1434-7892), [Anindya Sen](https://scholar.google.com/citations?user=vA6hgasAAAAJ&hl=en)

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/segmentation-of-blood-vessels-optic-disc/retinal-vessel-segmentation-on-drive)](https://paperswithcode.com/sota/retinal-vessel-segmentation-on-drive?p=segmentation-of-blood-vessels-optic-disc)

[![paper](https://img.shields.io/badge/Springer-Paper-rgb(50,50,200))](https://doi.org/10.1007/978-981-16-1543-6_16) 
[![paper](https://img.shields.io/badge/arXiv-Paper-brightgreen)](https://arxiv.org/abs/2207.04345)
[![dataset](https://img.shields.io/badge/DRIVE-Dataset-red)](https://drive.grand-challenge.org/) 
[![dataset](https://img.shields.io/badge/IDRiD-Dataset-rgb(165,10,175))](http://dx.doi.org/10.21227/H25W98) 
[![slides](https://img.shields.io/badge/Presentation-Slides-yellow)](https://drive.google.com/file/d/1xo0QhifZ2juq-dHoke1nMhcgLlhQJKsj/view?usp=sharing)

> **Abstract:** *Diabetic Retinopathy (DR) is a complication of long-standing, unchecked diabetes and one of the leading causes of blindness in the world. This paper focuses on improved and robust methods to extract some of the features of DR, viz. Blood Vessels and Exudates. Blood vessels are segmented using multiple morphological and thresholding operations. For the segmentation of exudates, k-means clustering and contour detection on the original images are used. Extensive noise reduction is performed to remove false positives from the vessel segmentation algorithm's results. The localization of Optic Disc using k-means clustering and template matching is also performed. Lastly, this paper presents a Deep Convolutional Neural Network (DCNN) model with 14 Convolutional Layers and 2 Fully Connected Layers, for the automatic, binary diagnosis of DR. The vessel segmentation, optic disc localization and DCNN achieve accuracies of 95.93%, 98.77% and 75.73% respectively.*

<hr />

<h2>Proposed Methods</h2>

<h3>1. Blood Vessel Segmentation</h3>

<div align="center">
<table>
  <tr >
    <td width=25%><img src = "./figures/Figure_2a.png" width="100%"><br>(a) </td>
    <td width=25%> <img src = "./figures/Figure_2b.png" width="100%"><br>(b) </td>
    <td width=25%> <img src = "./figures/Figure_2c.png" width="100%"><br>(c) </td>
    <td width=25%> <img src = "./figures/Figure_2d.png" width="100%"><br>(d) </td>
  </tr>
  <tr>  
    <td width=25%> <img src = "./figures/Figure_2e.png" width="100%"><br>(e) </td>
    <td width=25%> <img src = "./figures/Figure_2f.png" width="100%"><br>(f) </td>
    <td width=25%> <img src = "./figures/Figure_2g.png" width="100%"><br>(g) </td>
  </tr>
</table>
(a) Original Image, (b) Green channel component of (a), (c) CLAHE applied image, (d) Background estimated after Alternate Sequential Filtering, (e) Image (d) subtracted from (c) and CLAHE applied again, (f) Median blur and thresholding, (g) Final segmentation output.
</div>

<div align="center">
<img src = "./figures/Figure_3.png" width="70%">
<p>Blood vessel segmentation algorithm </p>
</div>

<br>

<h3>2. Optic Disc Localization</h3>

<div align="center">
<table>
  <tr >
    <td width=25%> <img src = "./figures/Figure_4a.png" width="100%"><br>(a) </td>
    <td width=25%> <img src = "./figures/Figure_4b.png" width="100%"><br>(b) </td>
    <td width=25%> <img src = "./figures/Figure_4c.png" width="100%"><br>(c) </td>
    <td width=25%> <img src = "./figures/Figure_4d.png" width="75%"><br>(d) </td>
  </tr>
  <tr>  
    <td width=25%> <img src = "./figures/Figure_4e.png" width="1005"><br>(e) </td>
    <td width=25%> <img src = "./figures/Figure_4f.png" width="100%"><br>(f) </td>
    <td width=25%> <img src = "./figures/Figure_4g.png" width="100%"><br>(g) </td>
  </tr>
</table>
<p>(a) Original image, (b) Grayscale of (a), (c) Result of k-means clustering, (d) Generated Template, (e) Template Matching result (using NCCOEFF; notice the OD region has highest similarity), (f) Marking OD and its center, (g) Masking OD region. </p>
</div>

<div align="center">
<img src = "./figures/Figure_5.png" width="70%">
<p>Optic Disc Localization algorithm </p>
</div>

<br>

<h3>3. Detection of Exudates</h3>

<div align="center">
<table>
  <tr >
    <td width=20%> <img src = "./figures/Figure_6a.png" width="100%"><br>(a) </td>
    <td width=20%> <img src = "./figures/Figure_6b.png" width="100%"><br>(b) </td>
    <td width=20%> <img src = "./figures/Figure_6c.png" width="100%"><br>(c) </td>
    <td width=20%> <img src = "./figures/Figure_6d.png" width="100%"><br>(d) </td>
    <td width=20%> <img src = "./figures/Figure_6e.png" width="100%"><br>(e) </td>
  </tr>
</table>
<p>(a) Original Image, (b) K-means clustering result, (c) Extracting the exudates from (b) and thresholding, (d) Logical OR of (c) and the images containing the smallest exudates, (e) Final segmentation result after OD masking. </p>
</div>

<div align="center">
<img src = "./figures/Figure_7.png" width="100%">
<p>Exudates Detection algorithm </p>
</div>

<br>

<h3>4. Binary Diagnosis of Diabetic Retinopathy using a Deep Convolutional Neural Network</h3>

<div align="center">
<img src = "./figures/Figure_8.png">
<p>Proposed Neural Network architecture with the corresponding kernel/Filter sizes (k), number of feature maps (n) and strides (s) specified for each convolutional layer.</p>
</div>

<hr />

<h2>Results </h2>
<h3>1. Blood Vessel Segmentation</h3>

The DRIVE dataset was used for the evaluation of the proposed method.

<div align="center">
<img src = "./figures/Figure_9.png" width="50%">
<p>The segmentation result for the best case (left) with ground truth (right)</p>
</div>
<br>

<h3>2. Optic Disc Localization</h3>

The proposed method had an accuracy of 98.77% on the IDRiD Segmentation Dataset.

<div align="center">
<img src = "./figures/Figure_10.png" width="50%">
<p>Optic Disc masked image (left) with the original image (right)</p>
</div>
<br><br>

<h3>3. Detection of Exudates</h3>

We used the IDRiD Segmentation Dataset for testing our proposed exudates detection algorithm.

<div align="center">
<table>
  <tr>
    <td width="50%"> <img src = "./figures/Figure_11a.png" width="100%"><br>(a) </td>
    <td width="50%"> <img src = "./figures/Figure_11b.png" width="100%"><br>(b) </td>
  </tr>
</table>
<p>(a), (b) Result of proposed exudates detection method (left) with original
image (right).</p>
</div>
<br><br>

<h3>4. Binary Diagnosis of DR using Deep CNN</h3>

The IDRiD Disease Grading dataset was used to train and evaluate the proposed network. It had an accuracy of 75.73% on a test set comprising 25% of the entire dataset.

<div align="center">
<img src = "./figures/Figure_12.png" width="70%">
<p>DCNN training results with Training Accuracy, Test Accuracy and Training Loss</p>
</div>
<br><br>

<hr />

Basu, S., Mukherjee, S., Bhattacharya, A., Sen, A. (2021). Segmentation of Blood Vessels, Optic Disc Localization, Detection of Exudates, and Diabetic Retinopathy Diagnosis from Digital Fundus Images. In: Pan, I., Mukherjee, A., Piuri, V. (eds) Proceedings of Research and Applications in Artificial Intelligence. Advances in Intelligent Systems and Computing, vol 1355. Springer, Singapore. https://doi.org/10.1007/978-981-16-1543-6_16

<hr />

<h2>Contact</h2>
For any queries, please contact: <a href="mailto:soham.basu07@gmail.com">soham.basu07@gmail.com</a>

DR_2021

DR_2021 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

DR_2021

DR_2021 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README