Artificial Intelligence (AI) in Healthcare

AI is changing healthcare by providing software that democratizes access to state-of-the-art technology.

1. Intro

2. Classification

3. Radiology for cancer

4. Blood testing

Theranos founder, Elizabeth Holmes captured the interest of many with her now disbanded single finger prick blood test. What was once thought to be a product that could democratize healthcare is now at the heart of an ongoing legal battle. Through advancements in artificial intelligence and machine learning technology in radiology, it’s possible a few more years could’ve been the solution Holmes needed to save her company. We’ll discuss the ways that AI/ML is currently being used within healthcare and its potential future for democratizing the industry.

One of the mission statements of Theranos was to democratize healthcare. Through a home-sized blood testing kit, patients all around the world could have access to information about their health; typically accessible to only wealthy countries. Despite a team of engineers and scientists, Holmes never made this blood testing system a reality. Medical experts and AI engineers in healthcare have the same mission statement as Theranos; the only difference is that AI engineers are using machine learning. Using

based-algorithms for image processing, experts have been able to identify health related issues using only photographs.

Google’s

project is using this technology to address blindness brought on by diabetes in India. With a growing number of diabetes patients in India, the goal for this project was to discover as many early cases as possible, so patients could seek out treatment to prevent blindness. India lacked the amount of trained nurses it needed to identify diabetes patients through a photograph. By loading in thousands of images and scans of healthy and diabetic eyes, Google’s algorithm learned how to train itself in identifying patients with the eye disease, reaching an accuracy rate equal to retinal specialists. What Google’s ML tool accomplished was providing poor people in rural India the same healthcare services wealthy patients in the United States would receive. As Seeing Potential rolls out into more countries and areas, the algorithm on this project will have more images to be trained on and become more accurate.

AI advancements have not only shown its promise in eye scans, but are changing the way radiologists do their job. By feeding a machine thousands of medical scans, the machine improves upon itself in identifying abnormalities within these scans. To make the algorithm run, radiologists will first sort through and label thousands of photos with and without a particular disease. In a very similar way to other image scans, the machine will train itself by identifying which images show signs of disease.

This technology is currently in its infancy, and much more will need to be seen and tested before this type of machine learning can be dispatched for widespread use. Some radiologists fear that it will replace the need for radiologists all together, which will have dire consequences for patients. While some skeptics agree to this assumption, the reality is that currently, this technology is not strong enough to be used on its own. Most diseases look different across people and age ranges, and not until the machine can be fed enough data will the fear of replacement be a concern.

Even with the increase in data, human radiologists will still need to be used to check the machine’s effectiveness and communicate with their patient. Most specialists agree that AI/ML is not being used as a replacement to radiology, but a tool that radiologists can use to better understand their patients in a quicker, more accurate way. Similar to Google’s Seeing Potential, ML radiology scans are yet another way that AI technology is expanding access to state-of-the-art healthcare.

Elizabeth Holmes’s goal of democratizing healthcare through blood testing proved to be smoke in mirrors. Yet, similar to the use of AI in scanning medical images, AI blood testing could bring about the reality Holmes failed to create. Blood can tell us a lot about overall health; the shape of blood cells can be a good indicator for what might be going wrong. The human eye has a difficult time identifying these microscopic cells, even when blown up on a computer. Research teams have created an algorithm that objectively identifies differences in these blood cells, which can tell patients more about their health.

While there are many research projects using AI/ML for blood cell analysis, none are more similar to Theranos than the Tel Aviv based startup

Sight Diagnostics created their own black box, OLO which was developed by a PhD scientist with a background in AI and biology. OLO uses AI to analyze blood cells by taking a finger prick sample which is then placed into the box which determines abnormalities within the blood. OLO has shown promising results in testing for malaria in India and Kenya, recently passed UK’s health and safety regulations, and is set to be deployed at Oxford University hospital.

Much like other industries (

,

,

, etc.), AI’s integration into healthcare has unlimited potential to change the industry through revolutionizing the accuracy at which medical scans can be read to democratize the way healthcare is distributed around the world. What remains of Theranos and much of Elizabeth Holmes’s reputation is currently being played out in the legal system and court of public opinion. However, Theranos’s philosophy of bringing medical tests to everyone is one which will outlive its reputation.