Our team consists of individuals who have traveled and worked in many different countries around the world. We have witnessed the challenges primary healthcare providers face in low-resource settings. For instance, Can’s mother is an ophthalmology professor and surgeon in Turkey, who is struggling with huge patient loads in her hospital. She wishes there was a more efficient way to deal with these huge loads of patients. Schahrouz on the other hand was exposed to multiple issues, such as patients getting diagnosed so late that they suffered permanent retinal damage, optic nerve damage or other severe intraocular diseases. These experiences are our inspiration. We want to help healthcare workers around the world to become more efficient in diagnosing critical diseases through ophthalmic imaging, and thereby help millions of patients in need. Considering the exponential growth of smartphone use around the world, we decided to leverage this broadly applicable tool by building an attachable low-cost device, in order to diagnose various eye diseases non-invasively.
What it does
The device uses the existing smartphone camera in combination with a fundus / infrared camera module to capture a retinal image. This retinal image is used to predict the risk of diabetes and heart failure - two of the leading causes of death and disability worldwide. It can help patients take preventive steps before the disease progresses, thereby saving costs for the healthcare system, time for the healthcare worker and valuable years of life for the patient.
How we built it
First, we have researched papers about various Machine Learning approaches to diagnose a patient with a certain eye disease. Then, we used the Messidor dataset to train a convolutional neural network in order to determine if a patient has a chance of having Diabetic Retinopathy. We used SketchUp to design a 3D-prototype. We decided to focus on rural communities, where this device can be more helpful, since severe eye diseases result in a higher death rate or permanent vision loss in those areas.
Challenges we ran into
The biggest challenge we ran into was the sparse public availability of retinal images in the World Wide Web to train our convolutional neural network. Furthermore, we were not able to determine the feasibility of our prototype from a physical point of view yet. A 3D-printer on-site would have given us a better proof of concept.
Accomplishments that we're proud of
We were able to determine if a person has the possibility of having Diabetic Retinopathy with an accuracy of 78.7% by using the Messidor dataset. Moreover, we outlined some of the challenges that we would face once we start to build the real product. We already tried to address some of these challenges during the ideation phase.
What we learned
Our product FixEye can be a potential non-invasive device to detect chronic diseases using retinal images.
What's next for FixEye
Our project has the potential to revolutionize the eyecare globally. Given that various Machine Learning techniques can be extremely efficient in detecting and diagnosing certain diseases via retinal images, we can potentially change millions of lives in rural areas.