Between 2010-2014, the field of image recognition and analysis was revolutionised by the introduction of deep learning, which enabled unprecedented performance leaps. These rapid advancements are fuelling the development of automated, accurate, accessible, and cost-effective medical diagnostics. Since 2010, over 60 entities including 40 new firms globally have set out to capitalise on these technological advances, seeking to commercialise AI-based diagnostics services in fields such as cancer and cardiovascular disease (CVD). More than $2.2 billion has been invested in new start-ups, with the investment since 2017 being 200% higher than the total since 2010.
Market for AI-enabled image-based medical diagnostics to grow by nearly 10,000% until 2040 whilst the global addressable market (scan volume regardless of processing method) will grow by 50%.
Technical threshold for automation reached, but is it a point of differentiation?
Algorithms are faster than humans and can be implemented on a massive scale with access to sufficient computing power via the cloud or even at the edge. Until recently, traditional hand-crafted algorithms would fail to meet the fundamental technical pre-requisite that they match or exceed the performance of human experts. The chart below shows that this is no longer the case and that this minimum technical milestone has already been surpassed, clearing away important technical barriers that had long held back automation in this field.
The report constructs realistic roadmaps, quantitatively outlining the current status, showing what challenges the technology still faces, and discussing how it is likely to evolve. The report identifies key commercial and technological issues currently limiting the uptake of image recognition AI and provides roadmaps describing when and how they are likely to be overcome over the next decade.
Rapid rise in company formation and investment
Since leaps in AI-based image recognition technology opened the market gates, over 60 entities including 40 new firms globally have set out to commercialize AI-based medical diagnostics based on various imaging modalities. The inset below shows the trend in company formation. Interestingly, this trend clearly correlates with the annual improvement reported in image recognition error rates between 2010 and now, highlighting how the technical and commercial developments are fully intertwined.
The chart also shows that money has been flowing into the start-up scene. Interestingly, the invested amounts have risen in the immediate past, partly reflecting the fact that the post-tech-demonstration companies now need larger financial reservoirs to pursue a scale-up strategy and to survive a potential consolidation phase. The inset also shows the more popular focus areas. In short, the disease segments with high scanned volumes and/or high value (e.g., preventing mortality with early detection) have received the most money thus far.
Companies are trying one or multiple of the following approaches to succeed:
- Towards wider applicability: The days of leaps in performance of image recognition are over, barring radical innovation in algorithm techniques. The gains in precision, recall and other metrics will henceforth be incremental. As such, the emphasis has shifted to other points. Of importance is showing that the AI is applicable to as wide a population set (ie: gender, age, ethnicity, tissue density, etc.) as possible.
- Evolving beyond simple abnormality identification towards super-human insights: Whilst there is a spread in what different algorithms are offering, most are positioned as decision support tools. The next evolution will be to provide further information and explanations alongside the detection and segmentation. Some are even aiming to suggest treatment options, hoping to evolve beyond the radiologist scope and to encroach into the doctors’ sphere of competency, although this is generally further down the line. In short, the goal is to raise the AI complexity beyond anomaly detection.
- Scale: Our view is that scale will matter in this business. Large scale, if done right, (a) means more access to data, which translates into an ever widening performance gap against competitors in terms of algorithm accuracy, versatility, and applicability; (b) creates a one-stop-shop proposition, helping with the sales and customer acquisition process; (c) results in larger technical teams that can aid the on-site into-work-flow integration process, which in turn boosts installed base and acts a lock-in mechanism. In general, scale can help the winners drive consolidation.
Current and future market trends
AI in medical image diagnostics is already in existence and numerous companies are past clinical stages in many segments. Our assessment is that the inflection point is near and likely to be reached around 2023-2025.
Indeed, it is estimated and forecasted the total addressable market per disease type in scan volume. It is leveraged out technological understanding and market insight to develop realistic market penetration and cost evolution projections per disease type, allowing us to create segmented market forecasts in scan volume and market value. The report highlights the current and future market share and size for 12 different applications and markets, including the detection of five types of cancer and four aspects of CVD. Insights into pricing strategies, uptake-limiting roadblocks and commercial opportunities complement our analysis to provide a comprehensive perspective of the market today and in the future.