Categories
Digital Health

Artificial Intelligence and Machine Learning (AI/ML) in Medical Devices

Do you know that FDA already approved 521 Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices?  

PAPNET. The failure of the pioneering AI/ML-enabled test.

Five hundred apps may not surprise you in January 2023, given the noise around Open AI and its GPTChat. However, the first such device, PAPNET Testing System, was approved over 28 years ago. In 1995, the year of Johnny Mnemonic and Ghost in the Shell movies!

Fig.1 Neural net-based (PAPNET, Neuromedical Systems, Suffern, NY) display of squamous cells (Papanicolaou stain) from a balloon smear showing effects of radiotherapy. Marked cell enlargement and vacuolization of cytoplasm are easily recognized.
Source: Koss, Leopold & Morgenstern, Nora & Tahir-Kheli, Naveed & Suhrland, Mark & Schreiber, Katie & Greenebaum, Ellen. (1998). Evaluation of Esophageal Cytology Using a Neural Net–Based Interactive Scanning System (the PAPNET System): Its Possible Role in Screening for Esophageal and Gastric Carcinoma. American journal of clinical pathology. 109. 549-57. 10.1093/ajcp/109.5.549.

PAPNET was using a neural network to analyze and interpret cytology from Pap smears. While this early system generated a lot of interest and Google Scholar lists 217 peer-reviewed articles on PAPNET results, the business side of it was not that great. The cost-effectiveness of the system in comparison to manual screening by cytotechnician was not there. Neuromedical Systems Inc, the company behind PAPNET went bust in 1999, and now its intellectual property is a part of Becton, Dickinson and Company portfolio.

List of FDA-approved Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices

If you look at the list of approved AI/ML-enabled medical devices, you will notice that the vast majority (392 medical devices, 75% of the whole) are for Radiology. Cardiovascular (57,11%), Hematology (15, 3%), and Neurology (14, 3%) are the remaining three significant categories.

Fig. 2: Split of approved AI/ML-enabled medical devices by Specialty Panel.
Source: FDA.gov, graphics by disrupting.healthcare

There are only 15 companies that have more than five AI/ML-enabled medical devices approved. Five of those companies are actually subsidiaries of GE, which in total owns 42 AI/ML-enabled medical devices. Then there is Siemens with 27 devices, Canon with 15, Aidoc Medical with 13, and Zebra Medical Vision with 9 devices. Philips, which also submitted its devices via different subsidiaries has in total 10 approved AI-enabled devices

Table 1. Companies with over 5 approved AI/ML-enabled medical devices
CompanyAI/ML-enabled medical devices
GE Medical Systems42
Siemens Healthineers27
Canon17
Aidoc Medical13
Philips Healthcare10
Zebra Medical Vision9
Quantib BV6
Arterys Inc.5
Clarius Mobile Health Corp.5
HeartFlow, Inc.5
RaySearch Laboratories AB5
Viz.ai, Inc.5
Source: FDA.gov, disrupting.healthcare

The future of AI-enabled and data-driven MedTech

The market for Artificial Intelligence / Machine Learning – enabled Medical Devices seems to be poised for growth. At a recent HLTH 2022 conference in Las Vegas, Michelle Wu, CEO of NyquistData, a company offering an AI-supported intelligence platform dedicated to MedTech companies discusses the advantages of using AI to unlock the potential of unstructured data from medical devices.

Leveraging Artificial Intelligence, Data to Improve Medical Devices
Source: Xtelligent Healthcare Media

Cleerly. An example of an AI-enabled medical device for a heart-attack-free future.

A good example of upcoming Artificial Intelligence / Machine Learning – enabled Medical Device may be Cleerly. The startup has raised $279 million from investors including Fidelity, T. Rowe Price, Novartis and Peter Thiel.

Founded by cardiologist James Min, former professor at Weill Cornell Medical College and director of the Dalio Institute of Cardiovascular Imaging at New York-Presbyterian, Cleerly uses AI to improve diagnostics cutting down on the time it takes to flag patients at risk.

Its proprietary AI algorithms analyze CCTA images to generate a 3D model of patients’ coronary arteries, identify their lumen (the cavity or channel within a tube or tubular organ such as a blood vessel) and vessel walls, locate and quantify stenoses, as well as identify, quantify and categorize plaque.

Source: Cleerly

By Piotr Wrzosinski

For over a decade I focus on digital technology in healthcare.
Good stuff, but it is all too slow.
I thought I will write a few words that may inspire you to disrupt this industry. Thanks for reading!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.