Introduction
The intersection of artificial intelligence and machine learning (AI/ML) with healthcare is nothing short of revolutionary. Yet, this fusion of technology and medicine brings forth a labyrinthine regulatory landscape. While FDA decisions offer a glimpse into what’s possible, they also highlight the challenges of bringing AI/ML medical devices to market. This article aims to describe the most up-to-date landscape for the top 10 markets for AI/ML-enabled medical devices that have received a final decision from the FDA.
Dataset Description
In our descriptive analysis of the publicly available data for AI/ML-enabled devices, we examined a curated dataset that was compiled by the FDA that contained the following information: date of final decision, submission number, name of device, name of company, lead panel (i.e., field of medicine), and the primary product code. The range of dates included in this dataset was from November 8, 1995, to July 27, 2023, totaling 692 records. Of these records, 691 had a unique submission number and 597 had a unique device name. There were a total of 407 unique companies in this list, unique panel leads, and 100 unique primary product codes. This article will primarily focus on the top 10 markets, product codes, and companies due to the large spread in data. For a more comprehensive overview of all available data, please see the included supplementary materials.
Temporal Trends
The surge in FDA decisions around the early 2010s is conspicuous. The establishment of the Software as a Medical Device Working Group (WG) by the International Medical Device Regulators Forum (IMDRF) in 2013 suggests an industry coming of age. Chaired by the FDA, the Software as a Medical Device WG aimed to create guidelines that would foster innovation and ensure timely access to safe and effective Software as a Medical Device globally.
Trends by Panel Lead (Field of Medicine)
While radiology’s overwhelming dominance — 531 devices which were found to have met the FDA’s applicable premarket requirements — signals a red-hot market, it also raises questions about market saturation, with cardiovascular at a distant 2nd place with only 71. Conversely, underrepresented panels like microbiology (n = 5), gastroenterology/urology (n = 5), and general and plastic surgery (n = 5) may offer relatively untapped avenues. Could these areas be the next frontier in AI/ML healthcare applications? The data suggest that diversification into these less crowded panels might currently be a strategic move.
Most Common Primary Product Codes
The frequent appearance of certain product codes “LLZ” (System, Image Processing, Radiological), “JAK” (System, X-Ray, Tomography, Computed), and “QIH” (Automated Radiological Image Processing Software) depicts how the market has been leaning toward specific device types over the past few years. However, one must question if this frequency is a result of technological limitations or if it reflects genuine market demand. For sponsors and developers, it is crucial to gain a better understanding of why these codes are as prevalent as they are before entering the market for the first time or attempting to stay on the market. Is it currently easier to get approval for these types of products, or do they accurately represent areas of high clinical need?
Conclusion
The world of AI/ML-enabled medical devices is a complex tapestry of opportunity and challenge. While trends point toward a radiology-focused market, savvy players may find lucrative niches in less crowded fields.
Key Takeaways
- Radiology’s market dominance may both invite and deter new entrants
- Established players like Siemens, Canon, and Aidoc carve out significant market share
- The temporal trends in FDA decisions reveal an overall increasing trend in the number of decisions, particularly after 2015; this uptick likely reflects the maturation of AI/ML technologies and their growing acceptance in medical applications
*The source of the data presented in this report is from the October 19, 2023, update, which is still current at the time of writing and publication of this report.