Ankit Jha - Abhinav - Parth - Keerat Chakshu_Rakshak

Identification of Blindness-Causing Eye Disease: Cataract

• Among all disorders, the leading cause of complete blindness is Cataract.

​

According to the National Blindness Survey (2006-2007), the prevalence of blindness in India is 8% among those over the age of 50. The following are the main reasons for prevalence of blindness in India:

​

• Lack of community awareness and poor health seeking behavior
• In India, a large majority of the elderly population lives in rural areas with limited access to eye care.
• Inadequately qualified health staff are available which are not able to accurately examine people’s health. This is particularly the case in rural areas, where, despite the government's provision of required technology and equipment, no one is available to examine the images of people’s eyes (generated using fundoscopic exam) and identify ailments.
• People have to go to distant places so as to get their eyes checked. This has created a feeling of reluctance and hesitance amongst the people about eye check ups and has increased the blindness rates in India.
• Working in remote places is not appealing to ophthalmologists. As a result, despite the government's provision of critical equipment, citizens do not have access to sufficient facilities.

​

We wanted to solve this problem by creating an AI Computer Vision Model that would make identification of Cataract easily accessible for rural people and the entire population. This is how ‘Chakshu_Rakshak’ was born.

​

Reference: Blindness — Vikaspedia

This pictorial representation depicts that the rate of Cataract detection is less in regions with a greater proportions of rural areas (like Bihar, Jharkhand, Jammu and Kashmir etc.).

Pictorial Representation

Project Overview

• Using Fundoscopic Images of the Retina, ‘Chakshu_Rakshak’ determines whether or not a person has Cataract.

​

• Cataract detection could become faster, more accurate, and more accessible as a result of this.

​

• This will be especially useful in locations where ophthalmologists are few, as persons even with only a high school diploma can be educated to take fundoscopic images.

​

• This would serve another purpose of generating employment in the less educated sections.

​

• This Artificial Intelligence Model may then scan the fundoscopic images provided by a fundus camera and make a prediction.

​

• This would dramatically minimise the amount of time required for an eye examination.

​

• It will also address the issue of lack of availability of ophthalmologists.

​

• The salary of an ophthalmologist ranges from 1.5 lakh INR to 7.5 lakh INR. Hence, this AI system would be an economical alternative in regions of less accessibility.

Reference: https://www.payscale.com/research/IN/Job=Optometrist/Salary

Research ( a. Lack of Ophthalmologists)

• Nearly 86% of all the medical visit in India are made by ruralites with majority still travelling more than 100 km to avail health care facility of which 70-80% is born out of pocket landing them in poverty.

​

• The only way which could lead to the goal of health inclusion is by incorporating impoverish needy rural population through community participation.

​

• India has one of the largest population of unnecessarily blind and vision impaired individuals in the world; totaling 15 million people.

​

• In one of the study, it was indicated that 143 public facilities found absenteeism of 45% doctors from PHCs with 56% of time found to be closed with an unpredictable pattern of closure and absenteeism during regular hour visit.

​

• India is manifesting a rise in cataracts primarily because of a dearth of ophthalmologists and not enough nourishment among its population, said Dr. Sridhar Prasad, senior manager, applications – LenSx, Alcon India.

​

• Lack of manpower when it comes to the number of ophthalmologists is one of the main reasons for the crisis. India accounts for an estimated 15,000 ophthalmologists which is  well below the World Health Organization’s recommended ratio of 1 ophthalmologist per 20,000 population.

​

• Although the number of health facilities in rural areas of India have increased during the past decade, convincing doctors to work in them remains a challenge. Dinesh C Sharma reports.

​

• This could be tackled by focusing on technological up gradation, capacity development and capability reinvigoration and limiting the scope for practice of illicit and unqualified practitioners.

​

​

​

Research ( b. Cataract in India)

• Cataract has been documented to be the most significant cause of bilateral blindness in India.

​

• In India cataract has been reported to be responsible for 50-80% of the bilaterally blind in the country.

​

• Based on the findings of National Program for Control of Blindness survey, control of cataract blindness was given the highest priority.

​

• “Given that technology can make early detection and treatment of cataract possible” stated Dr Prasad who works with the doctors to train them on the LenSx procedures.

​

• Indians are at a greater risk of cataracts at a comparatively younger age than Europeans or Americans. On average, an Indian could have a cataract almost 14 years before his American or European counterpart. But the gratifying part is that cataracts can be corrected if discovered and treated on time.

​

Reference:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4014652/

http://www.pharmabiz.com/NewsDetails.aspx?aid=109562&sid=1

report_February

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2612994/#:~:text=Cataract%20has%20been%20documented%20to%20be%20the%20most,the%20%E2%80%B3Vision%202020%3A%20The%20right%20to%20sight%20

​

Hypothesis:

​

Using our Computer Vision Artificial Intelligence Model named ‘Chakshu_Rakshak’ can make Cataract detection faster, more accurate, and more accessible.

Procedure

Variables

Dataset

Data/Observations

• Case 1

​

optimizer = ‘adam’

epochs = 18

data augmentation – No

Average Training Accuracy = 33%

Average Validation Accuracy = 25%

​

​

​

​

​

​

​

​

​

​

• Case 2

​

optimizer = RMSprop

learning rate = 0.01

data augmentation = No

Average Training Accuracy = 75%

Average Validation Accuracy = 83%

​

• Case 3

​

optimizer = RMSprop

learning rate = 0.1

data augmentation = No

Average Training Accuracy = 53%

Average Validation Accuracy = 50%

​

Data/Observations

• Case 4

​

optimizer = ‘adam’

steps_per_epoch = 68

data augmentation – No

Average Training Accuracy = 50%

Average Validation Accuracy = 50%

​

​

​

​

​

​

​

​

​

​

• Case 5

​

optimizer = RMSprop

learning rate = 0.001

steps_per_epoch = 68

data augmentation = No

Average Training Accuracy = 60%

Average Validation Accuracy = 66%

​

• Case 6

​

optimizer = RMSprop

learning rate = 0.001

steps_per_epoch = 3

data augmentation = Yes

Average Training Accuracy = 80%

Average Validation Accuracy = 80%

​

Final Model

• batch_size = 3
• layers.Dropout(0.2)
• Data Augmentation – Yes
• optimizer = RMSprop
• learning rate = 0.001
• steps_per_epoch = 3
• epochs = 50
• Average Training Accuracy = 90%
• Average Validation Accuracy = 83%

13

Community Aspect

• We may deduce from the statistics that Chakshu_Rakshak has the potential to become a revolutionary technological tool in the eye examination sector.
• It can make Cataract detection faster, more accurate, and more accessible. The model has an accuracy of 80%
• It has the potential to lessen the number of people who become blind as a result of cataracts.
• It has the potential to make cataract screening more accessible and affordable in rural locations which are suffering from the acute shortage of ophthalmologists.
• This will be especially useful in rural areas as persons even with only a high school diploma can be educated to take fundoscopic images. This would also serve the aim of creating jobs for the less educated parts of the population.
• It will also ease the process for senior citizens and physically challenged people who are incapable of travelling to long distances to access the facility of eye care by professional ophthalmologists.
• The salary of an ophthalmologist ranges from 1.5 lakh INR to 7.5 lakh INR. Hence, this AI system would be an economical alternative in regions of less accessibility.
• It will also address the issue of lack of availability of ophthalmologists.

​

​

In conclusion, Chakshu_Rakshak will be a boon to society. It would be a gleam of hope for reducing cataract-related blindness. It's a fantastic innovation that supports SDG 3: Good Health and Well-Being.

​

​

Future Developments

• In future, we want to make this model well trained by feeding more and more training data.

​

• We can collaborate with the government and incorporate ‘Chakshu_Rakshak’ in the Fundoscopy Equipment as well making the machines self reliant and more accurate.

​

• We also want to set up some self-capable machines in rural areas which suffer from the acute shortage of Ophthalmologists and are devoid of this facility

​

• Another development that we want to make in this model is to incorporate identification of other diseases like Glaucoma, Uveitis etc.

​

• We even want to add an option where the person can see his history and analyse the reports.