Greetings from Preceptra Regulatory Affairs. Last December marked the introduction of the new ISO 42001:2023, the Artificial Intelligence Management System (AIMS). This landmark publication signifies the growing importance of AI within organizations, underscoring its transformative potential across various sectors. In the healthcare industry, the integration of AI into medical devices is becoming increasingly indispensable, promising a new era of innovation and enhanced patient care. As AI continues to evolve, it is set to redefine the landscape of medical technology.
In this blog, we will explore the principles of responsible AI development and deployment, and discuss the key considerations you need to be aware of before adopting AI solutions in your organization, specifically for the healthcare industry.
Responsible Development: The Cornerstone of AI in Healthcare
AI as a medical device sometimes feels like a black box, and many questions arise: Is it practical? Is patient data safe? How exactly does AI make these decisions?
One critical challenge lies in the very foundation of AI, the data. Biases can infiltrate AI algorithms if the training data they rely on isn’t diverse and representative of the real-world population. Imagine an AI system trained primarily on data from a specific demographic group. This could lead to inaccurate diagnoses or biased treatment recommendations for patients with different backgrounds. As highlighted in “Lessons Learned from AI in Clinical Use,” AI excels at identifying patterns, but its accuracy can vary depending on the data it’s trained on.
Data security is another paramount concern. Hospitals hold a wealth of sensitive patient information, and a data breach can have devastating consequences.
AI is unlike other traditional medical devices, their internal decision-making processes can be opaque, even for experts. This lack of transparency can hinder trust and create hesitation among healthcare professionals to fully embrace AI’s capabilities.
A Medical Device Quality Foundation and A New Dawn in Responsible AI Management
As AI developers, demonstrating a commitment to patient safety throughout a device’s lifecycle is paramount. Certifications from accredited bodies provide a clear path to achieving this.
Existing international standards like ISO 13485:2016 offer a strong foundation. This standard outlines best practices for quality management systems specific to medical devices, ensuring a focus on patient safety from design and development to production and ongoing monitoring.
In December 2023, the first-ever standard for Artificial Intelligence Management Systems (AIMS) emerged. This standard is known as ISO/IEC 42001 providing a comprehensive framework for organizations developing AI medical devices. The core of this new standard goes beyond just quality, by focusing on building trustworthy AI.
Knowing Before Adopting
There are several factors to consider before adopting AI, especially in healthcare. Here are the broad perspectives, you’ll need to understand:
AI Algorithm: This encompasses the training data used, the model’s quality, and the validation process – all of which directly impact AI accuracy. You can delve deeper into this topic in our previous blog post, “Navigating the Waves of AI Technology in Healthcare.“
AI System: Beyond the core algorithm, factors like user experience, reliability, technical support, and security are essential. Our blog post, “System Reliability: Beside the Model, Why We Should Focus on the System,” explores this further
Required Standards: Ensure the AI medical device adheres to established quality management standards (ISO 13485:2016) and optionally the new standard for AI management (ISO/IEC 42001:2023). This ensures the device meets the baseline safety and reliability requirements and adheres to responsible AI development practices.
Thailand Medical Device Registration - Requirements for AI in Medical Devices
While Thailand currently mandates ISO 13485 quality management systems for medical devices, that’s no longer enough for AI-powered tools. To ensure responsible development, the Thai Medical Device Control Division now requires additional crucial documents for AI medical devices. This includes information on:
- Input data and AI datasets: Transparency regarding the data used to train the AI system.
- AI Model/algorithm description: A clear understanding of the underlying algorithms powering the AI.
- AI performance: Data on the effectiveness and limitations of the AI system.
- Re-training AI model process: Demonstrating a plan for ongoing improvement and adaptation of the AI.
For more details, visit the Thai Medical Device Control Division website: https://en.fda.moph.go.th/
