How Generative AI Empowers Medical Imaging: Subtle Medical Provides the Answer

On October 18, 2023, jointly hosted by VCBeat, VB100, and the Endoscopy Industry Alliance, with support from Subtle Medical“2023 Future Medical Imaging Innovation Forum”The event concluded successfully at the Nanjing Yangtze River International Conference Center. The forum brought together research experts and distinguished entrepreneurs from China’s medical imaging sector. Through in-depth exchanges and discussions, participants jointly sought replicable pathways and models for industrial integration and innovative development, thereby providing broader perspectives and greater scope for industrial upgrading and high-quality growth.

Dr. Xiang Lei, CTO of Subtle Medical China, Shares “Generative AI Fully Accelerates Image Acquisition and Optimizes Imaging Workflows”

Xiang Lei, CTO of Subtle Medical ChinaPh.D.Attend the forum, and"Generative AI Comprehensively Accelerates Image Acquisition and Optimizes Imaging Workflows"delivered an insightful presentation on the topic. As the sole third-party solution provider in the field of AI-powered medical imaging acceleration and enhancement to have obtained certifications from the U.S. FDA, the European CE, and China’s NMPA, Subtle Medical’s every move has drawn significant attention from the industry.

Thus, in the face of the new opportunities and challenges that generative AI currently presents to the medical imaging industry, Dr. Xiang Lei, CTO of Subtle Medical China, has outlined a new development blueprint for the sector through his recent remarks, which, when considered alongside Subtle Medical’s recent strategic moves, offer a clear vision of the industry’s future trajectory.

In recent years, as China’s healthcare industry has entered a phase of deeper structural reform, addressing critical clinical needs has become increasingly vital. This trend is particularly pronounced in the rapidly evolving field of medical imaging, where both company founders and investors are now conducting more meticulous and comprehensive analyses of demand-side dynamics.

In this keynote address, Dr. Xiang Lei explicitly stated, “"How to integrate existing pain points and problems with generative AI capabilities will determine the future direction of the medical imaging industry."

So, what exactly are the pain points? Dr. Xiang Lei highlighted three key aspects.

First, from the patient perspective, the main issues are prolonged waiting times and radiation exposure caused by contrast agent deposition during the examination.

As is well known, the vast majority of MRI and PET-CT scanners in China are currently operating at full capacity. Patients often face waiting times of at least three days, and in some cases several weeks, to schedule an MRI scan; in major cities such as Beijing and Shanghai, the wait is even longer. This bottleneck stems primarily from two factors: first, a significant shortage of diagnostic radiologists and technologists; and second, limited throughput of imaging equipment, with each scan taking anywhere from 30 minutes to one hour.

Furthermore, to meet the imaging requirements for certain specific examinations, contrast agents must be administered to patients. The use of these agents can lead to their deposition in the patient’s body and cause radiation-related harm; therefore, strategies to minimize contrast agent usage must be considered. Additionally, reducing the dosage can also lower examination costs.

The second major pain point lies on the hospital side, primarily manifested in significantly increased costs and investments, as well as patient wait times and attrition due to limited service capacity.

To address the aforementioned patient queuing issues, hospitals need to procure imaging equipment with higher scanning speeds and better quality. However, a single high-end imaging device can cost tens of millions of yuan, and even upgrading existing equipment requires an investment of several million yuan. Beyond the financial burden, the time cost is also significant. It is reported that the cycle for upgrading or procuring imaging equipment typically ranges from six months to one year. During this period, hospital service capacity becomes constrained, leading some patients to choose other hospitals due to unwillingness to continue waiting. Furthermore, adding new equipment imposes additional requirements on space and supporting medical staff, thereby incurring further costs.

The third point concerns downstream service providers, primarily manifested in the fact that low-quality medical images can compromise subsequent quantitative diagnostic accuracy, thereby adversely affecting final treatment outcomes.In response, Dr. Xiang Lei stated, “When downstream service providers perform quantitative analysis based on imaging data, if the quality of the medical images is suboptimal, the resulting quantitative analysis will be significantly compromised.”

Thus, enhancing the quality and efficiency of clinical imaging is a widespread global demand at the current stage.

According to Dr. Xiang Lei, the current applications of generative AI in healthcare are primarily manifested in three specific segments: medical Q&A, AI-driven drug discovery and development, and acceleration of clinical validation. By breaking down these application scenarios, it becomes evident that the challenges are concentrated inData, Quality, and EfficiencyThree Major Ports: This Precisely Represents the Application Capabilities of Generative AI Technology.

Therefore, based on the current pain points in the medical imaging industry,DeepSubtle Medical focuses on optimizing the efficiency and quality of medical imaging examinations at the source through generative AI technology.. So, how is this specifically reflected? Dr. Xiang Lei also provides a detailed introduction and explanation in this regard.

To address the challenges of prolonged patient wait times and radiation exposure, Subtle Medical has achieved a fourfold increase in scanning speed while reducing the contrast agent dosage to 10%.In response, Dr. Xiang Lei stated, “Our first step is to accelerate the process by optimizing scanning parameters, reducing the scan time to one-quarter of the original duration. Of course, this approach results in compromised image quality. Therefore, our second step is to implement image enhancement to optimize image quality.”

It is reported that Subtle Medical’s flagship products at this stage are SupPET™ and SupMR™. The former, designed for PET (positron emission tomography), accelerates PET imaging by 4–10 times while still delivering medical images of equivalent quality. The latter, tailored for MR (magnetic resonance imaging), reduces imaging time to less than half of the conventional duration, yet yields final image quality that is consistent with or even superior to that obtained from standard-time acquisitions.

Next is addressing the hospital side, that is, how to enhance medical imaging service capabilities.Based on this, Subtle Medical is enhancing product compatibility. It is reported that SupMR™ currently covers all MRI scan regions and sequences, and is compatible with MRI imaging equipment of all brands and models. In China, Subtle Medical offers a pure on-premises deployment model, which ensures compatibility while safeguarding patient data security and privacy. In the majority of global markets, Subtle Medical adopts a cloud-based SaaS deployment model to help hospitals improve imaging efficiency, quality, and revenue.

Finally, it empowers downstream applications, primarily by enhancing imaging equipment to achieve superior image quality and more accurate downstream analysis.Unlike the strategic approaches of other domestic AI companies, Subtle Medical’s product line, while situated within the medical imaging AI sector, does not focus on computer-aided diagnosis. Instead, it specializes in front-end AI-powered image enhancement. During the image enhancement process, Subtle Medical’s AI solutions do not alter physicians’ existing workflows. Without requiring any modifications to imaging hardware, its intelligent algorithms facilitate image enhancement or optimized reconstruction, thereby reducing scan times, lowering radiation doses, and improving image quality.

In this regard, Dr. Xiang Lei stated, ““Subtle Medical’s AI products are not designed to replace physicians, but rather to empower them with access to a greater volume of higher-quality imaging data.”

Currently, Subtle Medical’s AI products are accelerating their strategic rollout across markets in China, the United States, and Europe. Over 500 healthcare institutions and imaging centers worldwide have completed deployment. The company’s global order volume reached nearly RMB 100 million in 2022, and it secured another nearly RMB 100 million in orders during the first half of 2023. If this momentum is maintained while controlling costs, Subtle Medical will achieve break-even within this year.It is expected to achieve profitability within two years,Breaking the Profitability Deadlock in Medical AI.

Subtle Medical has even greater goals for the future.

According to Xiang Lei, development based on generative AI models,Subtle Medical will also continue to optimize its existing products,Not only must the image quality be consistent with the gold standard, but it should even surpass it.Currently, Subtle Medical has completed preliminary validation, with actual performance meeting expectations. The next step is to expand the sample size to further enhance model stability.

In addition to further improving image quality,Another major initiative by Subtle Medical is leveraging generative AI to synthesize unacquired sequence images from scanned ones, thereby further accelerating scan times and improving examination efficiency. Preliminary results indicate that the fully generated images not only ensure diagnostic accuracy but also exhibit higher image quality than those obtained through conventional scanning.In fact, refining AI for contrast enhancement is not an overnight endeavor. Magnetic resonance imaging (MRI) scans of different anatomical regions and disease types require corresponding adjustments to various MRI sequences. These sequences also involve complexities related to different manufacturers, models, and field strengths, all of which necessitate that AI companies specializing in contrast-enhanced imaging devote greater efforts to gradually achieve comprehensive coverage.

According to Dr. Xiang Lei, when Subtle Medical’s initial product took shape in 2019, it primarily focused on the six most frequently scanned anatomical regions. However, after several years of research and development, the company has now achieved coverage for all anatomical regions. Additionally, since 2017, it has been continuously refining its intellectual property portfolio through dozens of global patents.

Through continuous iteration, the product value of Subtle Medical and the company’s competitive advantages in the industry have become increasingly pronounced.Dr. Gong Enhao, Founder and CEO of Subtle Medical“In an interview, it was stated that for AI technology itself, enterprises should optimize the technology to minimize costs and ensure stability, while accelerating technological iteration and product deployment. This constitutes a virtuous cycle capable of delivering commercial value, rather than blindly engaging in price reductions or continuous expansion.”