Our Challenge

• The total number of blind people is about 39 million, and of the visually impaired is approximately 285 million. The leading causes are cataract, glaucoma, macular degeneration, corneal opacities, diabetic retinopathy and eye conditions in children. The preventable causes of visual impairment are approximately 80% of the total cases. Early diagnosis plays an invaluable role in such cases.
• Diabetic retinopathy (DR) is the major cause for the blindness in the working-age population. Diabetic retinopathy is caused by diabetes. Statistically more than 75% of people who had diabetes for more than 20 years will have some form of DR.
• The early detection of DR could prevent blindness. The problem is that DR usually shows no symptoms before the later stages where the blindness cannot be avoided. The early diagnosis in the blood vessels of the retina can be made through regular retina monitoring. The regular retina analysis of diabetic patients is a worldwide problem, because nearly 3 million eyes will need to be evaluated by the ophthalmologists each day by 2030 (35 exams per second).
• Another problem is the decrease of the number of ophthalmologists. We have a 54 % increase in the diabetes population against 2% growth in the number of ophthalmologists for 2030.
• Besides, most people (80%) with diabetes live in low-income countries where the ophthalmologists are hardly accessible. In developed countries diabetic patients are supposed to visit an ophthalmologist each year, but in practice they do it even each 2-5 years which prevents early detection of the retina disorders.

Our technology

• Our technologies are based on state-of-the art machine learning techniques, called deep learning. These algorithms were partly developed from within the company DreamQuark. At DreamQuark, as former particle physicists, we use our expertise of Big Data analysis and efficient algorithm design to reinvent how data analysis is performed in the healthcare sector. We know how to deal with complex images, and how to detect even the smallest, invisible effects. We work closely with healthcare professionals who give us invaluable medical expertise.
• These days we live in the era of Artificial Intelligence revolution via developments done in the field of deep learning algorithms. Computers have learned to recognize objects in pictures, they learned to write like humans, they can drive cars, they are able to read emotions in our faces and in our voice, they are even able to dream and they have learned how to paint a-la Van Gogh. What is the difference between Deep Learning techniques and more traditional cognitive technologies? Let's say we want to recognize different objects on an image. Before the appearance of deep learning, a lot of feature engineering was needed in order to sum up the information contained in an image - or data in general - to simplify the decision process. Indeed, the classic Machine learning tools have difficulties to deal from complex data and human expertise is required to extract what is meaningful in the data, before feeding this preprocessed data to the algorithm. With Deep Learning on the other hand, the algorithm learns itself what is the best simplified representation of the data and what decision to take. Deep Learning Techniques allows to outperform the more traditional methods in performance and to address issues that were not available to the machine before.

Our solutions