Medical technology, or med-tech, is a field with enormous potential for doing good. The field is filled with low-hanging fruit rotting on the branches because there are not enough people attempting to pick them. More specifically, there is huge potential in the analysis of medical data.
Our research team looked at retinal images (the retina is the inner tissue layer that lines the back of the eye), which offer a non-invasive and simple way to see the blood vessels and nerves in the eye. Our main focus was to develop an autonomous diagnostic test for the detection of Diabetic Retinopathy (DR), a diabetes complication that is the leading cause of blindness amongst working adults in the US. During the development of disease detection models, one of our machine learning experts found that the models could be trained to detect the number of years a patient has had diabetes and also distinguish between a person who has been diagnosed with diabetes and a healthy individual. Sensitivity and specificity (measures of accuracy) for detecting the age of the disease and whether a person was healthy or sick were over 90%.
With these fantastic results in hand, we proceeded to examine the potential impact that this discovery could have. Every year, in the US alone, there are about a million and a half new diabetic patients, a quarter of which are diagnosed unintentionally by optometrists during routine eye exams (they observe signs of diabetic retinopathy).
Research shows that a blood test every five years enables earlier detection of diabetes by an average of 3.3 years , other estimates suggest that the diagnosis is advanced by an average of six years, with the help of regular testing .
A different study on the long-term effects of diabetes showed that a three-year delay in diagnosis resulted in a 1.8% increase in mortality in the following ten years, and a six-year delay of 3.6% . The study population had a mean age of 60 which translates into an estimated life expectancy of 23.6 years.
Suppose that only 40% of the US population who are short-sighted  go to an optometrist once every five years and during that visit also have their retinas looked at and analyzed for diabetes using our machine learning models. If we multiply all these numbers, we will find that approximately 300,000 (between 200,000 - 400,000) person-years can potentially be saved every year in the US alone. This is an incomprehensible number of human life years that can easily be saved using a simple test.
For comparison, it is the equivalent of saving all the residents of Sedona, Arizona from certain death, every single year or saving all the people who died from COVID-19 in 2020 in the US. And that's only in the US!
We already know that retinal images could be used to detect numerous diseases including various retinal diseases, cardiovascular diseases, and even neurological conditions such as Alzheimer's and Parkinson's diseases. A couple of images from your eyes could provide early detection for a broad range of conditions that can be used to improve and even save your life.