07 Mar AI enabled medical devices where are we innovating
Insight by: Debbie Lin
Calling all biohackers. Your apple watch will soon be able to non-invasively track glucose continuously. Dubbed E5, Apple hit a major milestone recently in their moonshot project to develop a non-invasive glucose monitor. Current wearable glucose monitors detect glucose levels through the body fluid that runs between cells. This normally requires inserting a needle under the skin which can be painful and have other side effects. Thus far, it has been difficult to develop a technology small enough to be noninvasive and wearable; however, sensors in your clothing to measure movement and temperature, earbuds not only for music but to measure vitals signs and even contact lenses that measure blood pressure, will be common place. All of these devices will use artificial intelligence (AI) and machine learning (ML) trained on massive amounts of data to help clinicians and consumers/patients make actionable decisions regarding healthcare. In the recent years the rate of FDA approval of AI/ML-based medical devices, has skyrocketed. These devices have the potential to revolutionize the way doctors diagnose and treat patients. Their use is expected to grow even further in the coming years. Curious as to exactly what the rate of approval was for such devices, I did a quick analysis using data on the FDA website.
Since 1995, there were a total of 521 AI/ML FDA device approvals. The number grew from 5 approvals in 2015 to 91 in last year in 2022. 2021 was a record year, 115 approvals. This means over a period of 7 years the rate of approvals increased approximately 20 fold.
A few other observations from my short analysis of over the last 5 years:
- The majority of devices approved are intended for us by healthcare professional as compared a patient/consumer.
- Over 70% of the products focused on radiology, followed by cardiology (between 6-17%).
- Other areas of focus included oncology, neurology, orthopedics, ophthalmology, respiratory, and gastrointestinal.
- Radiological technologies covered a wide range of applications such as diagnostic imaging, image analysis, and image-guided interventions.
What does this analysis tell us? It appears that a great deal of innovation and thus FDA submissions in AI/ML enabled medical devices, has come from the area of radiology. Why radiology? Imaging data is easier to access at a large scale than other types of clinical and health information. It comes in a structured form as opposed to doctors notes and other clinical information which can come unstructured. AI/ML can be powerfully applied to recognized patterns in images to develop insights. In the past years, as investors and innovators, over 70% focus in radiology shows we certainly seem to have over-indexed in radiology.
What about in the future? In their recent report, Rock Health’s 2022 top funded clinical indications demonstrates that mental health followed by cardiovascular and then oncology are the top funded indications in 2022. Not all of these products have an AI/ML component and nor are they all medical devices. But given the amount of investor spend in these categories, we should be seeing more FDA approved AI/ML devices in mental health, oncology, primary care, maternal and reproductive health and diabetes, into the future. I’m looking forward to seeing more innovations come through in other areas and I am looking forward to tracking my glucose using my new Apple Watch.