全球医疗影像：来自硅谷的AI创新

Susannah Ludwig+41 582 723 127susannah.ludwig@bernsteinsg.comCourtney Breen+1 917 344 8407courtney.breen@bernsteinsg.comLee Hambright+1 917 344 8429lee.hambright@bernsteinsg.comNandan Kulkarni+91 22 6842 1436nandan.kulkarni@bernsteinsg.comDelphine Le Louet+33 1 42 13 92 93delphine.le-louet@bernsteinsg.comRebecca Liang, Ph.D.+852 2123 2656rebecca.liang@bernsteinsg.comWilliam Pickering, MD+1 917 344 8340william.pickering@bernsteinsg.comJustin Smith+44 20 7762 5899justin.smith@bernsteinsg.comMiki Sogi, Ph.D.+81 3 6777 6991miki.sogi@bernsteinsg.comLance Wilkes+1 917 344 8501lance.wilkes@bernsteinsg.comJeffrey Walch+1 917 344 8613jeffrey.walch@bernsteinsg.com As AI and machine learning excel at pattern recognition and anomaly detection, image analysis is one of the healthcare segments best suitedfor adoption. Over the past decade, leading imaging companies, major technology firms, and a growing number of start-ups have introducedAI-driven tools aimed at improving both the speed and accuracy of lesion detection. In today’s Weekend Healthcare Pulse, we examine severalprominent start-ups and recently IPO-listed players in radiology, focusing on how they are using AI to enhance clinical workflows and thecompetitive dynamics between these players and the incumbent OEMs. By Susannah Ludwig, Estelle Pang and Richard Hombach diagnosis. RADIOLOGY IS THE MOST ADVANCED SUBMARKETFOR AI ADOPTION IN MEDTECH Radiology is widely regarded as the most advanced submarketfor AI adoption within medtech, primarily because it involves apattern recognition problem applied to vast volumes of medicalimages, supported by highly structured and repeatable workflows.Radiologists routinely interpret thousands of images acrossmodalities such as CT, MRI, and X-ray, making the specialtyparticularly amenable to algorithmic assistance. In this context, AIoffers clear and measurable efficiency gains by helping cliniciansmanage rising scan volumes, reducing reporting workloads, andprioritizing time sensitive cases more effectively. AI applications are already being embedded across multiplestages of the radiology workflow. These include image acquisition,workflow coordination, post-processing analysis, as well as follow-up management and training (Exhibit 1). In the long-term, there ispotential in transforming radiology from a discipline of qualitativeinterpretation to one of quantitative analysis. This shift is oftenreferred to as radiomics. But in imaging, and in medtech morebroadly, AI's main impact in the near to medium-term will likely bemore around speeding up workflows, increasing the efficiency ofradiology suites, reducing errors, and improving the accuracy of COVID-19 HAS BEEN A WAKE-UP CALL FOR HOSPITALSTO INVEST FOR EFFICIENCY Prior to the pandemic, equipment and healthcare IT vendors oftenstruggled to persuade hospitals to adopt new technologies. Manyadvanced software solutions were seen as useful but not essential, making it difficult to justify investment during procurement. Thismindset shifted in 2020, when COVID-19 forced hospitals torapidly expand ICU capacity and operate more efficiently to managesurging patient volumes. In the aftermath, radiology departmentsfaced significant backlogs, with the pressure persisting as long-term trends continued to intensify. An aging population, as babyboomers move into higher-care years, is increasing demand forimaging services while ongoing staff shortages are pressuringcapacity. As a result, the need for technologies that improveefficiency and streamline workflows in the radiology suite persists. volumes and reducing backlogs, meaning that efficiency gains willbe critical to meeting growing demand. EXHIBIT 4:The gap between U.S. demand and supply ofphysicians in radiology is expected to reach -10% by 2034 Estimated supply and demand for radiology physicians (2024-2034E) AGING POPULATION SUSTAINING RADIOLOGY DEMAND Demographic trends are set to provide a sustained tailwind fordiagnostic imaging demand over the coming decade. Agingpopulations, in particular, are a primary driver, as older individualstypically require more frequent imaging procedures to managechronic conditions and monitor disease progression. According tothe World Bank, the population aged 65 and above is projected togrow at roughly +3% annually through 2030 in the US (Exhibit 2),while Europe is expected to see a growth rate of around +2% overthe same period (Exhibit 3). As a result, the underlying demand forbasic diagnostic imaging services is likely to remain robust in themedium-term, reinforcing the importance of workflow optimizationsolutions to help providers keep pace with rising volumes. AI driven workflow innovation represents one of the most promisinglevers for alleviating these pressures. Rather than replacingclinical roles, AI has the potential to augment staff productivityby automating repetitive and time intensive tasks across the carepathway. According to a McKinsey industry report, AI could freeup to 4