Traditional healthcare systems are evolving rapidly due to the emergence of technological innovations. At Google, we are constantly in search of ways to improve the quality of healthcare with technology. For example, we are researching the uses and future of artificial intelligence (AI) in e-health. In this article, we will take a look at the promising practices and future of this innovative combination. We will show you how AI can help doctors and patients to prevent, predict, and diagnose disease faster and more accurately.
There is an increasingly big amount of digital medical data available that can be used in AI models. Through using machine learning (ML) we can already predict what treatment protocols are most likely to succeed. A more complex form of ML, neural network technology, associates input, output, and different variables to predict the likelihood of someone developing a specific disease. Furthermore, the most complex form of AI, deep learning, can reveal thousands of hidden features that are uncovered faster by quickly processing graphics, data units, and cloud architectures. At Google we are currently studying the use of these AI technologies to assist in diagnosing cancer, predicting patient outcomes, preventing blindness in people with diabetes, and much more. Let us tell you a little more about one of our studies.
“I had started thinking a lot about working on more fundamental problems. I wanted to use image recognition for something that would benefit society.”
— Varun Gulshan, Google Research Scientist
The AI team was joined by medical experts from the start to implement their own scans, medical knowledge, and experience in the field into the AI models. The doctors scanned the images for symptoms multiple times and coded them. The coded images were then used to create an algorithm. In this way, our models learned to recognize and scan the images of thousands of different eyes for signs that could predict risk for blindness, just like a doctor would. The digitalization of this process was incredibly important since, in India alone, 70 million diabetics needed screening quickly.
We tested our models and found that, in ideal settings, AI developed by Google Health could identify signs of blindness linked to diabetes with more than 90% accuracy. After these findings, we took the AI algorithms and started to build an ARDA-machine (Automated Retinal Disease Assessment). In this way, the AI systems became directly applicable in real-world clinical settings and could immediately help doctors and patients.
But we’re not done yet, real-world situations and people vary heavily and the challenge is up to us to make AI compatible with these differences without compromising on quality. We are excited to see what the future of AI in healthcare will bring and are determined to help doctors and patients around the world. To get inspired by our projects and learn more about how the future of health looks, take a look at the video below.