AI Enabled DR Diagnosis FYP

Diabetic retinopathy (DR) is a disease not fatal but proves to be if not treated on time it will cause complete blindness. In Pakistan, due to the lack of professional ophthalmologists, many people suffer from this disease as machinery to examine DR is too heavy to be transported to the rural areas

Project Title

AI Enabled DR Diagnosis FYP

Project Area of Specialization

Artificial Intelligence

Project Summary

Diabetic retinopathy (DR) is a disease not fatal but proves to be if not treated on time it will cause complete blindness. In Pakistan, due to the lack of professional ophthalmologists, many people suffer from this disease as machinery to examine DR is too heavy to be transported to the rural areas of Pakistan. Since the internet is also not widely available around Pakistan and especially in the less developed regions, an offline smartphone-based device is introduced to diagnose those people who mostly live in less developed areas. With the help of software that has an AI model fed into the smartphone-based device, it should be able to output a result indicating the diagnosis determining to what extent does the patient has DR and at which stage. With this device, there will not be a need for a professional retinal specialist to be present at that moment because it will be efficient enough to diagnose patients with the help of the AI model giving accurate results. Such devices have been programmed before and tested on separate datasets proving themselves as reliable to an extent, they were made cost-efficient as they used the camera of cell phones and worked offline using the GUI of the cell phones instead of online servers the AI algorithm performed well on the device without the need of internet. However, they are still not accurate enough to be marketed or used widely. Dataset of the images taken from the portable device, with the help of Vista View camera will go through a process of cleaning and pre-processing to well organize and enhance the images, furthermore, the images will be augmented in order to remove the biasedness of the dataset. After augmentation, the images will be separated into portions for training and testing to form an AI model as accurately as possible. lastly, a mobile application compatible with the portable device will be developed to detect the stages of DR.

Project Objectives

• Data Pre-processing will be performed on the collected dataset to clean our real-world data into an understandable form.
• Data augmentation technique to meet the need of balance data availability for the algorithm.
• To devise an algorithm with a reliable network accuracy for the detection of the level of tests for Diabetic Retinopathy.
• To convert our research into an app so that ophthalmologists can use it easily.

Project Implementation Method

The fundus image data will be collected through a portable device. The data will go through the pre-processing stage to achieve a clean, formative, and organized dataset. There is a high chance the real-world dataset collected will be unbalanced. Therefore, to remove biases from the dataset, a data augmentation technique will be applied to it. Then, ML architectures will be used for modeling to get DR disease classification. Consequently, an app will be developed that can work with the portable device to detect the stages of DR.

Benefits of the Project

Diabetic retinopathy (DR) is a serious sight-threatening disease, that damages the retina of an eye. Pakistan is now ranked 3rd in the prevalence of Diabetes in China and India. According to International Diabetes Federation (IDF), 33M people live with diabetes in Pakistan. Which is alarming for health experts. Pakistan is now ahead of the USA, which has 32M diabetics in 330M people. Data is now more valuable than anything. All policies are being prepared on basis of data, but unfortunately, Pakistan lacks data collection. Also, we don’t have enough ophthalmologists, healthcare centers, and light machinery. What we’d are heavy fundus cameras that can’t be carried everywhere. To overcome this an easy, portable & cheap device is required. It aims to design a system that can perform DR screening from a portable device. A designed algorithm with a suitable model will be proposed with cleaned & augmented data. Lastly, an application will be designed that health workers can easily carry from place to place to perform DR screening.

In rural areas where technologies could not make their way till now. But the disease is increasing rapidly and to control this issue an early diagnosis is sufficient. To capture the retinal images or to do diabetic screening requires a high-performance fundus camera and machine, which is too huge and costly and could not make it to commute to different places. To minimize this issue, we are going to work through a portable device. So that we can use it as a gadget and can take it to different places for the screening purpose.

Technical Details of Final Deliverable

An application that diagnoses the human eye with the help of the specific type of lens then compares with the collected dataset and will result in us whether the penitent has diabetic retinopathy or not. Whether it is treatable or required immediate assistance etc. This user-friendly application cannot use the camera lens of mobile phones since they are not of high resolution to capture the fundus images. Only a doctor can use this app with the help of a special lens needed to take detailed images. In simplified form, broken down into four different tasks following are the outcomes of this project:

• Identify the best Data augmentation technique to resolve class unbalancing of the local Dataset.
• Visualization/Comparison of Augmented Data/Real Data.
• Train any CNN architecture using a balanced Dataset to achieve the best outcome.
• Simple low weight mobile app for diagnosing DR; based on ML algorithm.

Final Deliverable of the Project

Software System

Core Industry

Medical

Other Industries

IT

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Good Health and Well-Being for People, Industry, Innovation and Infrastructure, Partnerships to achieve the Goal

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com