AI in Healthcare: Regulatory and Patenting Trends

Regulatory documents in China and the United States

China issued “Regulations for the Management of Artificial Intelligence-Assisted Diagnosis Technology (Trial)” in November 2009, and “Guidelines for Classification and Demarcation of Artificial Intelligence Medical Software Products” in July 2021. These documents specified the scope and limitations of AI-assisted diagnosis technology to ensure that it would not be used as the final clinical diagnosis without being ascertained by a qualified clinician. In the meanwhile, they offered guidance as to how such technology should be harnessed and properly used, by introducing a high threshold for AI medical software with low algorithm maturity.

In November 2024, China issued “Reference Guidelines for Artificial Intelligence Application Scenarios in the Health Care Industry”, greenlighting 84 AI-utilisation scenarios falling under 4 categories of “AI + Medical Service Management”, “AI + Community-level Public Health Services”, “AI + Health Sector Development”, and “AI + Medical Teaching and Research”.

The U.S. Food & Drug Administration (FDA) has so far formulated 24 Guidance Documents (23 finals, 1 draft) with Digital Health content, as well as a plurality of discussion papers, guiding principles and the like. The latest one is “Draft Guidance: Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations” published on January 6, 2025.

The FDA has been continuously updating the “AI/ML-Enabled Medical Devices List,” which includes medical devices that incorporate AI/ML across medical disciplines. The most recent update of the List was made in the end of 2024, with the current list including 1,016 devices. The List reveals that the AI/ML-enabled medical devices are aggregated in medical diagnostics field, with 76% concentrating on Radiology, around 10% on Cardiovascular and approximately 4% on Neurology.

Patent Landscape in AI-Enabled Medical Devices

Exploring the FDA “AI/ML-Enabled Medical Devices List” reveals key players in the USA market. These include GE (GE Healthcare, GE Medical Systems et al.), Siemens (Siemens Medical Solutions, Siemens Healthcare GmbH et al.), Canon (mainly Canon Medical Systems Corporation), Aidoc Medical, and Shanghai United Imaging. Other than Aidoc Medical, an Israeli clinical AI solution provider, the other four companies filed a considerable number of patent applications in AI medical technologies ranging from several hundred to over a thousand in various countries. The USPTO and CNIPA received most applications, indicating the importance of the two markets.

General Electric (GE) strategically allocated its AI-health-related patent filings in a few major markets, led by the USA (40-50%), followed by China (20-30%), trailed by Europe and Japan (cumulative 10-20%). This geographical distribution is indicative of GE’s approach to leveraging local market opportunities and navigating the regulatory environments that are most receptive to healthcare innovation. GE’s patents in this domain include several noteworthy innovations:

• Distributed Learning Platforms: A patent family from GE highlights the use of distributed learning systems that integrate AI informatics into healthcare systems. This approach facilitates real-time data sharing and processing across various locations, which enhances adaptability and efficiency in healthcare services. The patents granted in the USA (US10957442B2), China, and Europe, underscore the global applicability of this technology.
• MRI Image Enhancement: GE has also developed methods and systems for enhancing magnetic resonance imaging. For instance, US10712416B1 and its Chinese counterpart CN111513716B utilize an extended sensitivity model and a deep neural network to improve the clarity and precision of MRI scans, aiding in more accurate diagnostics and better patient prognosis.
• Normative Imaging Data Generation: Another patent family focuses on generating normative imaging data for medical image processing using deep learning. This includes patents such as US11195277B2, which standardize data analysis across healthcare systems, crucial for consistent and reliable diagnostic assessments.

Siemens, on the other hand, is re-aligning its patent portfolio to emphasize AI-health-related technologies, filing 30-40% of such applications in the USA, 25-30% in China, and 15-20% in Europe. By pivoting away from its headquarters and traditional filing bastion of Germany, Siemens embraces a broader international patent portfolio presence, especially in China, highlighting the region’s growing importance in the AI-health market. Siemens’ key patent families include:

• Enhanced CT Measurement: Siemens harnesses machine learning techniques, such as artificial neural networks, to significantly enhance the accuracy of CT measurements. The international protection and recognition of this methodology are evident from patents like EP3672688B1.
• Optimized Contrast Agent Administration: Another series of patents, including US10925565B2, improve the administration of contrast agents. This technology allows for better image quality with significantly lower doses, reduces scan times, and tailors injection protocols to individual patient needs.
• Anatomical Structure Estimation: Additionally, Siemens has developed methods like those in US11478212B2 and its Chinese counterpart CN108433700B, using machine learning to estimate internal anatomical structures from surface data. This technology reduces radiation exposure and streamlines the imaging positioning process.

Canon has also adjusted its strategic patent filings, allocating AI-health-related filings in Japan (over 40%), USA (over 30%), and China (over 10%). This shift shows an increased focus on filings in the USA and China, with a relative decrease in Japan, underscoring the pivotal role of these markets in the AI-health sector. Key patent families in Canon’s AI-driven portfolio include:

• Medical Data Processing Innovations: One Canon patent family uses deep neural networks (DNN) to enhance the functionality of medical data processing apparatuses and magnetic resonance imaging (MRI) devices, as evidenced by patents such as JP6545887B2.
• Medical Signal Restoration: Canon has also patented systems that adapt machine learning models to specific imaging conditions, enhancing diagnostic accuracy. Patents such as JP7126864B2 showcase this technology.
• Pre-detection Model for Medical Examinations: Further, Canon is developing medical examination assistance apparatuses that evaluate pre-detection models for adverse events based on machine learning techniques, such as in JP7438693B2.

Shanghai United Imaging demonstrates a clear strategic shift directed to AI-health-related patent filings, with a dominant focus on the Chinese market (60%), followed by strategic filing activities in the USA (20%) and Europe (5-10%). This distribution suggests an increasing intention to target the gigantic home and thriving global market. Notable patent families from Shanghai United Imaging are as follows:

• ROI Determination: Patents such as CN109727234B describe AI algorithms for determining regions of interest in diagnostic imaging, enhancing accuracy and patient-specific analysis.
• Artifact Removal: This family focuses on systems and methods for removing Gibbs artifacts in medical imaging, using machine learning to improve image clarity. Patents like GB2557382B and EP3465247B1 in the UK and Europe are indicative of the wide applicability of this technology.
• Tumor Grading: The company also focuses on tumor classification using trained detection models based on deep learning architectures such as YOLO and Fast R-CNN, as seen in patents like CN109903280B and US11200668B2.

These comprehensive filings by GE, Siemens, Canon, and Shanghai United Imaging not only showcase their innovations in leveraging AI to transform healthcare diagnostics and treatments but also underpin their strategic approaches to protecting these technologies in key global markets.