Innovating Safely: Software-as-a-Medical Device Compliance in the Era of Generative AI

By Phane Mane and Brian Peet

According to a recent report from BCG on Generative AI in health and opportunities.., “the healthcare industry is expected to experience some of the most significant benefits from and growth in GenAI investments in the coming years.” ?The report goes on to say that “GenAI is projected to grow faster in healthcare than any other industry. With an estimated compound annual growth rate of 85%, by 2027 the market value is expected to reach $22B”

This may not come as a surprise to anyone given the transcending impact that Generative AI is expected to have across all industries. That said, not all opportunities in Healthcare are equally opportunistic, let alone feasible given the underlying complexity, impact on patients, and of course the strict regulatory compliance requirement that ensues.

In this blog, we will discuss the application of Generative AI with Software-as-a-Medical Device (SaMD) aspects of the Healthcare industry which are typically associated with firms that specialize in the design and manufacturing of MedTech and/or Life Science products.

Before we dive into the potential benefits and/or impact, let us get a baseline understanding of what Software-as-a-Medical Device (SaMD) is.? The U.S. Food and Drug Administration (FDA) defines SaMD as "software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device."

Based on that definition, SaMD can include any Software that acquires, processes, or analyzes a medical image OR provides recommendations to HealthCare professionals (HCPs), and patients with prevention, diagnosis, treatment, or mitigation of a disease condition OR control other scenarios where a piece of software is directly part of the solution. As an example, the use of dosimetry for calculation and determination of radiation dosage (eg: vials) that directly integrates with other systems used in the treatment process (such as liver cancer).

Given the high adverse impact of getting things wrong and the relative infancy of Generative AI, we propose identifying use cases that do not touch any patient data or interact with medical procedures involving SaMD interfaces to avoid patient risks and ensure compliance with regulatory bodies like the FDA, Health Canada, or the European Medicines Agency (EMA), etc.

Some examples could include

Producing Draft Documentation - a typical treatment procedure using SaMD can often involve significant regulatory and other compliance document generation. This is where Generative AI can be useful in creating content such as images, text, and even software code for parts of the system.? Additionally, Generative AI can produce the logic for diagnostic accuracy, statistical process control, identify treatment regimens, and streamline certain workflows.

Creating Software Requirements Specifications (SRS) – Generative AI can be very handy in creating SRS drafts for SaMD products to ensure completeness, consistency, and traceability of requirements documentation while facilitating communication between development teams and regulatory authorities.

Test Case Generation for Verification – by leveraging Generative-AIs strength to identify logical reasoning you can use it to validate and ensure comprehensive test coverage and adherence to regulatory requirements such as IEC 62304 while optimizing testing efforts and resource use.

Drawing from knowledge -?review and analyze more medical literature, studies, and clinical outcomes than any single person could in their lifetime. Leveraging advanced technologies like PubMed, ontologies, and the RAG (Retrieve, Aggregate, Generate) framework offers a comprehensive approach to reviewing and analyzing medical literature and clinical outcomes. PubMed provides access to vast biomedical literature, while ontologies organize and categorize data based on medical concepts. The RAG framework systematically processes data by retrieving relevant information, aggregating it to identify patterns, and generating actionable insights. Fine-tuning Large Language Models (LLMs) on domain-specific data enhances their understanding of medical terminology and clinical findings, making them valuable tools for healthcare providers.

In practice, healthcare professionals can access the results of this analysis through user-friendly interfaces or applications. This enables them to interact with the fine-tuned language model, gaining access to timely information, personalized recommendations, and decision support tools.

Overall, integrating advanced technologies in medical literature review and analysis streamlines the process, enhances accessibility to insights, and empowers healthcare providers to make informed decisions. This approach represents a significant advancement in healthcare delivery, offering unprecedented access to knowledge and improving the quality of care provided to patients.

Data augmentation -?the ability to?produce synthetic medical images to improve the datasets for machine learning algorithms. Data augmentation, specifically the generation of synthetic medical images, is revolutionizing the creation of control arms in clinical trials, expediting product and drug approval processes. Traditional control arm creation faces challenges like data scarcity and limited patient diversity. However, data augmentation addresses these issues by producing synthetic images resembling real patient data. Utilizing generative models such as variational autoencoders and generative adversarial networks, researchers can simulate diverse pathologies and anatomical variations encountered in clinical practice.

The integration of synthetic images offers several advantages. It overcomes data scarcity by supplementing real-world datasets with abundant synthetic data, ensuring the representation of diverse patient populations. Moreover, it enables customization of control arms to align with trial objectives, minimizing bias and improving reliability. Augmented control arms streamline trials by expediting patient recruitment and data collection, accelerating the generation of conclusive results.

This approach benefits pharmaceutical companies, medical device companies, regulatory agencies, and patients. It hastens drug and product approvals by reducing the trial duration and resources required. Additionally, it enhances patient access to innovative therapies that undergo rigorous evaluation. Ultimately, data augmentation advances precision medicine and improves patient outcomes by enhancing the efficiency, reliability, and generalizability of clinical trial results.

As such, Generative AI can help facilitate evidence-based decision-making, optimize your workforce, and lower costs related to human labor-intensive tasks that may improve patient outcomes.

I hope you find this blog useful, do let us know other topics that you would like us to discuss in the future.