Google researchers worked with doctors to create an AI which will be able to identify diabetic retinopathy. Diabetic Retinopathy is one of the main causes of blindness among adults. Google AI uses the same methodology with which it differentiates the faces of animals, human beings or other objects. They upload the objects and faces to Google Online services. The system will also detect the condition of a patient/individual by checking out the retinal photos. They are focused on doing something core to Google.
The objective of Google AI to read retinas
The aim behind this is not at all to replace the doctors. It is to prevent blindness. If you can catch diabetic retinopathy early, you can prevent the permanent blindness. The advantage is, technology can screen more people than a doctor can do single-handedly, especially in such countries where the healthcare is limited. According to Lily Peng, the physician and biomedical engineer at Google, the project started when a Google researcher realized that in his area, Doctors struggled to provide proper screening to the locals who needed it.
In a lot of places, doctors are treating patients on the basis of photos. Google AI can be of help for them. The technology can bring screening services to remote locations where the diabetic retinal scan is less, not available. The project is a part of a much bigger effort started by Peng. In this project, detection of illness is possible using deep neural networks and pattern recognition system.
The Google AI is capable of analyzing a vast amount of data in a short time. At Google AI lab, DeepMind, in London, researchers also teamed up with Britain National Health Services. They are working on developing technologies which will automatically detect which patients have what disease.
The device performed better than doctors in accuracy
Peng is also a part of Google Brain. It is a team inside the company that offers all forms of AI services starting from security, health, travel anything you name. Peng is leading a group of researchers now who are focusing on medical applications of AI. The computer scientists at Google and medical researchers from both US and India say that the algorithm was initially developed for analyzing retinal images.
The design was not to identify the features of diabetic retinopathy. The device learned on its own after Peng and its team fed 128000 images into neural network of Google AI. After that, the system identified the condition better than the group of doctors.
The system was able to detect and avoid the false negatives and false positives more than 91.5% of the times. The result exceeded National Institute of Health’s accuracy and precision of 80%. Journal of American Medical Association released the results. On the other hand, a professor at Carnegie Mellon, known as Yaser Sheikh says this might become difficult in the developing world. There are certain barriers. But Peng and her team are pushing forward. The new technology will be able to facilitate diagnosis and treatment in the areas where there are fewer doctors.