Analytics 4 Life (A4L) is an artificial intelligence company developing non-invasive, cost-effective medical devices. In January of 2017, the company moved into new office space at JLABS @ Toronto, where it employs a team of approximately 20 people, with another 10 located at its U.S. headquarters in Research Triangle Park, North Carolina, and around the Eastern U.S. “We’re really lucky to be at JLABS inside of MaRS Discovery, and the fact that the Vector Institute for Artificial Intelligence recently launched within the same building, it’s very fortunate for us,” says Shyam Ramchandani, Ph.D., Vice President of Clinical Affairs at A4L.
“The idea for A4L came from the company’s Founder and Chief Scientific Officer, Sunny Gupta, who spent time with the Canadian Armed Forces in Kingston,” says Ramchandani. “Canada has strengths in signal analysis due to its involvement in NORAD.” Canada’s signal intelligence facility at Canadian Forces Station Alert, the most Northern permanently inhabited settlement in the world, is where advanced technology is deployed to warn against military threats, such as the presence of nuclear submarines and torpedo launches.
“My family has a history of heart disease,” recalls Gupta, “And one day while working at the Canadian Armed Forces, it struck me: why couldn’t advanced signal analysis apply to physiologic signals emitted by the heart to better diagnose disease?” The company launched officially in July 2012, set to tackle the #1 cause of death in the developed world: heart disease. But as Ramchandani explains, they actually started about a year and a half earlier out of a 100 sq. ft. administrative office above a church in Kingston, Ontario. “It was just three of us and our laptops, exploring ways to analyze biological signals using machine learning,” he recalls. A4L went on to leverage the SOSCIP R&D consortium, which gave them access to IBM cloud computing infrastructure and highly qualified local talent through the TalentEdge Program.
A4L sees machine learning as a promising opportunity to assist doctors in diagnosing disease
The American College of Cardiology defines significant coronary artery disease as when a patient has at least one major coronary artery blocked to 70% or greater, and the best way to diagnose the disease is through catheterization. During the procedure, a specialist feeds a catheter through a patient’s vein, usually from their leg, to their heart. A dye is then released into their bloodstream and x-ray images are taken in real time, so a physician can see where the dye is going, with the pinch points representing blockages within the vessels. Doctors can then go in and open up the vessels using stents. “The problem is, you don’t just send patients into the cath lab without just cause,” says Ramchandani. “While it is a fairly routine procedure, there still is a mortality rate, and patients are exposed to radiation.”
While there are several different tests doctors send patients in a workup to catheterization, one of the most common is the Nuclear Perfusion Test or Nuclear Stress Test. Every year, between six to eight million nuclear stress tests are performed in the U.S. on patients identified as at risk of having heart disease. It’s similar to a routine exercise stress test but with images taken so they can detect whether areas in your heart are receiving less blood flow. After fasting for 6 to 12 hours, the patient is injected with radioactive dye. Images are then taken before and after intense exercise in order to compare blood flow. “The entire procedure takes approximately 4 to 6 hours,” Ramchandani explains, “and usually a follow-up appointment to review results takes place days or weeks after the procedure, as the images need to be read by a radiologist or cardiologist. What’s more, the radioactive dye is equivalent to receiving about 600 chest x-rays, and the test records false positives about 60% of the time as demonstrated in the PROMISE study.”
In terms of infrastructure, all you really need is a good 3G Wi-Fi connection
– Shyam Ramchandani on the simplicity of A4L’s solution
With A4L’s test, the patient lies down and six electrodes are applied to the chest and one on the back. The data is pushed to the cloud, where a machine-learned algorithm crunches the signals. By the time the patient’s shirt is back on, the results are ready for the doctor and patient to review. “In terms of infrastructure, all you really need is a good 3G Wi-Fi connection,” says Ramchandani. “There are actually three components of the product at work. 1) The Phase Signal Recorder device collects the physiologic signals and transmits the signal package to the cloud for analysis. 2) Phase signals are analyzed by the cloud-based, machine-learned algorithm. 3) And results are available on our web-based physician portal.”
Opportunities abound for A4L’s proprietary machine-learning platform
The information within these signals was previously hidden, but that is where the machine learning comes into play. “We take the signals from patients who we know have coronary disease and those who do not, and we use that data to train our algorithm to separate the positives and negatives. We then test that algorithm prospectively on the next set of data to assess how well it runs,” explains Ramchandani. He believes the opportunities for this platform are substantial. Not only can it be used for diagnosing other types of cardiovascular disease and other forms of medical imaging, but it could also be applied to other industries. “You could literally use this in mining for precious metals and gas if you wanted to,” he says. “The mining industry uses different types of seismic signals in the discovery process. If you have the right training data, you could use artificial intelligence to accurately predict whether or not you’re sitting on top of a mineral deposit.” A4L’s device is not yet FDA approved or available for commercial use, as the company plans to share early results of their first clinical trial in late 2017.
SOURCE: Invest in Ontario