Is AI changing the rules for medical imaging?

Yohann Tschudi, PhD. Technology & Market Analyst, Computing & Software added: “At Yole, we estimate the total market in 2025 for software generated revenues through the sale of AI tools will reach $2.9bn with a 36% CAGR between 2019 and 2025. These revenues can be shared between the main applications including improved image capture, noise reduction, image reconstruction, screening, diagnostic and treatment planning.”

Yole Group of Companies including Yole and KnowMade, have been investigating the computing & software domain for a while. Its aim is to develop a deep understanding of the impact of AI on the semiconductor industry, with a special focus on software development. 

Since the beginning, with dedicated teams, the market research, strategy and IP consulting companies have developed an impressive expertise with both perspectives, software and applications including automotive, consumer and medical. AI, cryptocurrencies, machine learning and block chain are the key words of their researches and are well analysed in a dedicated collection of reports.

AI is clearly one of the biggest questioning today. Lots of companies invest a lot of money and develop innovative technologies to answer the market demands and follow the industry evolution. The semiconductor industry is part of this revolution. Yole and KnowMade have announced a special focus on the medical imaging applications with two dedicated reports, respectively ‘AI for Medical Imaging market & technology report’ and ‘AI in Medical Diagnostics - Patent Landscape’.

With its new technology and market report, ‘AI for Medical Imaging’, Yole has offered a comprehensive overview of the AI market in the field of medical imaging with the companies involved. This new report proposes a deep analysis at different levels of the supply chain, from device to platform including the development of the related algorithms. 

To complete this technology and market approach, KnowMade described the patent landscape with the time-evolution of published patents, and countries of patent filings as well as a relevant analysis of the IP strategies of the key players in its new ‘AI in Medical Diagnostics - Patent Landscape report’. In addition to the presentation of a detailed ranking of main patent assignees, KnowMade’s analysts identified over 90 IP newcomers including startups, described their operations and listed their patents.

AI is based on the training of algorithms. Deep learning is a type of AI technology based on artificial neural networks which can detect more precise details in the data. This technology has initially been implemented for recognition models and is specialised for the study of images.

“Radiology is mutating with the adoption of deep learning models for the recognition of lesions in the body, to prioritise cases for the direct treatment of patients at risk, to predict the evolution of pathologies,” added Villien. “Furthermore, AI affects all the imaging modalities in particular MRI , CT scanning, X-rays and Ultrasound imaging. These are the ones at the centre of Yole’s study.”

Not every type of modality requires the same algorithm. In fact, modalities can be organised into two types of procedures: quality procedures, which include MRI and CT scans, and fast imaging procedures, which include ultrasound and X-rays.

“The professionals’ needs depend highly on the imaging modality used,” continued Villien.

On the one hand, MRI and CT scans are intensive procedures able to acquire high quality images. With the addition of annotations on the images, the model can reach very high accuracy to classify pathologies or to segment objects. 

Furthermore, the execution speed of the model does not need to be very fast, as the imaging procedure is usually long. On the other hand, models trained on ultrasound images are in need of very fast execution to be able to process real-time images. Those models are then used to detect abnormalities faster and prioritise cases, implying an important productivity gain.

The application of the models empowered by AI can be classified as within three parts: the screening models, in charge of the detection of abnormalities, the diagnostic models which is, from its side, in charge of the evaluation of the disease and the treatment planning models. These latest ones are able to predict the most pertinent treatment according to the pathology and the physical condition of the patient. The value generated by the use of such models in hospitals depends on their applications.

According to Yole’s AI reports, more than $2.05bn has been invested since 2010 by companies working on the development of artificial intelligence for medical imaging. Companies such as Heartflow received $476m investment in the past ten years. The main expected players in this market are the medical diagnostic systems manufacturers, General Electric, Philips and Siemens, but also AI-guru companies like IBM or Microsoft. 

“Beside these big companies, the number of IP newcomers is important and growing,” said Brice Sagot, CTO and Co-founder at KnowMade. “Unlike the development of new medical devices, AI software development costs are moderate. As a result, the number of IP newcomers developing innovative software is likely to continue to rise sharply in coming years.”

Thereby, with emergence of many new companies like Aidence, Bay Labs and doc.ai, and given the many advantages and new applications of AI for medical diagnostics, it is crucial to understand the IP position and strategy of these different players. This analysis helps detect business risks and opportunities, anticipate emerging applications, and enables strategic decisions to strengthen one’s market position. 

In addition, the analysis of the time evolution of patent publications points out the development of medical diagnostic systems with built-in computer-assisted detection features. According to KnowMade’s AI report, this trend is not new.

“The first patents related to this topic were published in the 1980s,” explained Olivier Thomas, Patent & Technology Analyst at KnowMade.“In the 1990s Japanese medical imaging system manufacturers like Toshiba, Fujifilm, Topcon, Fujitsu and Hitachi started to investigate this topic soon followed by European companies like Siemens and Philips and then by American companies like IBM, Medtronic and General Electric.”