Machine learning in diagnostic imaging will grow into a $2 billion market by 2023, according to analysts at Signify Research.
Signify predicts the market will quadruple as healthcare systems adopt machine learning tools at every step in the imaging pathway, from scheduling to follow up, to improve care and mitigate radiologist shortages.
While noting that regulatory barriers and a lack of validatory data from large studies could hold the sector back, Signify thinks the drivers o1f growth are too powerful to constrain the nascent market.
Healthcare professionals began exploring the use of algorithms to support image analysis 20 years ago, but these early applications had neither the accuracy nor the ease of use to go mainstream. The breakthroughs facilitated by deep learning around the start of the decade, including the application of the machine learning algorithms to medical images, raised hopes that the technology was nearly ready for widespread use.
Signify, a healthcare technology research group, thinks the technological advances of recent years have brought the market for machine learning in diagnostic imaging to an inflection point. Today, Signify values the market for AI-based medical image analysis software at less than $500 million. By 2023, Signify thinks the market will top $2 billion.
The forecast reflects Signify’s belief that the market for detection and automated diagnosis tools, which is virtually non-existent today, will blossom into a near $1 billion field. In parallel, demand for quantitative tools, which account for most of the market today, will take off, adding around another $1 billion to the total size of the AI-based medical image analysis sector. Signify expects demand for decision support tools to grow, too, pushing the overall AI imaging sector past the $2 billion mark.
Those forecasts, however, are based on assumptions that could prove to be wrong. The reasons why hospitals may want AI-based technologies, such as shortages of radiologists and the need to reduce errors, are well established. It is less clear whether AI-based imaging software will meet their needs and be adopted in the near term, though.
Signify’s belief that the software can break through is underpinned by its improved performance, heavy investment in the field, rising use of quantitative imaging in clinical care and the emergence of supporting storage, compute and networking infrastructure.
These factors lay the groundwork for explosive growth but, as Signify acknowledges, barriers remain. Signify picks out the regulatory process, lack of validation data from large studies and the challenge of integrating new software into existing workflows as some of the main roadblocks. The analysts also note that healthcare providers are reluctant to buy AI tools from multiple vendors as they fear it will lead to integration challenges and increased administrative overheads.
Work is underway to tackle some of the challenges. Signify thinks the FDA's precertification program may "expedite the regulatory process to some extent." Other challenges may prove to be persistent.
The uneven and at times slow rate of adoption of healthcare technology could be a factor. The uptake of MRI equipment illustrates the point. Australia, Canada and France have between 10 and 15 MRI machines per 1 million people, according to data from the Organisation for Economic Co-operation and Development. In contrast, Italy, Germany and the U.S. hit that level in the early 2000s and have continued to add machines. The U.S. now has around 38 MRI machines per million people, almost four times as many as Canada.
Adoption of AI-based imaging tools may be faster and more uniform as they will likely require smaller upfront investments and promise long-term savings. Equally though, uptake could be constrained by long sales cycles, legal issues and other barriers.