In April 2018, the U.S. Food and Drug Administration (FDA) approved the first artificial intelligence-powered diagnostic system, a software program used to detect diabetes-related vision loss.

Since then, the industry has seen explosive growth of AI in medical device manufacturing, which is transforming every step in the process from design to production to postmarket surveillance.

Where is the industry headed, and where can we expect to see the greatest effects? In this article, we look at why and how AI is revolutionizing medical device manufacturing, and what the FDA has signaled about the compliance issues that lie ahead.

The rise of AI in medical device manufacturing

The market for AI in medical devices is experiencing massive growth and is expected to surge from $15 billion in 2023 to a whopping $97 billion by 2028.

More so than other industries, AI is playing a transformative role in medical devices through a combination of strict regulatory requirements, growing demand for personalization, and the need for precision.

What’s more, as companies mature in their digital journeys, a vast wealth of data is being generated that AI can help harness in new ways. Below, we discuss some emerging applications of AI in medical device manufacturing that are driving improvements in efficiency, safety, and compliance.

AI-powered supply chain management

Medical device manufacturing requires seamless coordination of raw materials, components, and finished products across a complex supply chain. Traditional supply chains are linear and vulnerable to disruption, giving rise to digital supply networks (DSNs) that enable real-time visibility and collaboration.

While DSNs represented a major advance for supply chain resilience, AI has taken their potential power to a whole new level. AI enables manufacturers to leverage the large amount of data contained in DSNs, allowing for innovations such as:
• Predictive analytics for demand forecasting: AI tools can integrate ERP data to automate demand forecasting and suggest changes to production and inventory levels.
• Self-healing supply chains: AI plus technologies such as internet of things (IoT) sensors and radio frequency identification (RFID) allow companies to continuously monitor inventory levels, production schedules, and shipping routes to mitigate disruption by automatically triggering corrective actions.
• AI risk modeling: AI risk modeling can help rapidly simulate various supply chain scenarios to identify potential bottlenecks or resource shortages and show where companies need to diversify to ensure resilience.
• Inventory optimization: AI algorithms can assess consumption patterns to ensure the right amount of inventory is stocked at each stage to reduce waste and storage costs.

Predictive maintenance

Predictive maintenance is another leading application of AI in medical devices, both in terms of enhancing production efficiency as well as ensuring optimal performance of the devices themselves.

In the manufacturing environment, equipment breakdowns and maintenance delays can lead to costly downtime while increasing the risk of product quality issues.

AI systems can analyze real-time machine performance data collected by IoT sensors to identify early warning signs of machine wear or impending malfunction.

AI systems can analyze real-time machine performance data collected by IoT sensors to identify early warning signs of machine wear or impending malfunction. This allows teams to take proactive steps to repair a machine before it breaks down. Predictive insights based on historical and real-time performance data can also help determine the optimal scheduling of maintenance activities with minimal disruption to production.

In terms of the medical devices themselves, AI is leading the way in helping ensure more reliable performance of devices in the field.

One example is GE Healthcare’s OnWatch Predict for MRI, which uses a digital twin of MRI machines to monitor critical components and detect any potential issues quickly. The result has been an average increase in MRI uptime of 4.5 days per year, reducing customer service requests by up to 35%.

AI vision detection systems

AI-based vision detection systems for quality control are currently one of the most well-developed AI technologies on the market today. These systems utilize deep learning to detect defects in medical devices at high levels of precision.

In many cases, these systems can identify defects that might not otherwise be detected through traditional quality inspections, reducing the risk of human error. For instance, vision detection systems can detect:
• Micron-level cracks or defects in surgical instruments or implantable devices
• Improper or incomplete seals on sterile packaging that could allow for contamination
• Printed circuit board (PCB) defects such as voids or the presence of foreign material
• Errors in final assembly of medical devices

Generative AI for product design

Where medical device development once involved lengthy cycles of design, prototyping, testing, and iteration, AI can significantly accelerate the process. The technology holds huge implications for bringing innovative devices to market faster, as well as enhancing partnerships between device manufacturers and contract manufacturing organizations (CMOs).

The technology holds huge implications for bringing innovative devices to market faster, as well as enhancing partnerships between device manufacturers and contract manufacturing organizations.

For example:
• AI algorithms can analyze input parameters like material properties and dimensions to generate thousands of potential designs optimized for performance and manufacturability.
• Integrating AI with additive manufacturing or 3D printing is enabling faster innovations in product design, allowing manufacturers to produce prototypes at speed to meet demand.

Virtual prototyping using digital twins, AI, and generative design allows designers to quickly develop and test highly complex virtual designs that can be taken to production without a physical prototype.

Personalized medical devices

AI is driving a shift from one-size-fits-all medical devices to more patient-centric solutions. This personalization is particularly evident in areas like orthopedics and wearables, where individual anatomy can determine device performance.

For example, AI tools can analyze patient data like CT or MRI scans to design implants or prosthetics that perfectly align with the patient’s unique anatomy. Examples include orthopedic implants customized to an individual’s bone structure, and cardiac stents tailored to a patient’s vascular structure. Surgery results are improved while recovery time and postsurgical complications are reduced.

AI for regulatory compliance

AI has several potential applications in the compliance arena, emerging as a powerful tool in helping navigate the complex regulatory landscape of medical device manufacturing. EU MDR, FDA QMSR, MDSAP—these regulatory programs and others involve meeting an ever-expanding number of regulatory requirements, underscoring the need for more sophisticated compliance tools.

Examples include:
• Regulatory prediction tools: AI systems can analyze historical trends in regulatory approvals to predict potential risks for new devices. For instance, reviewing previous rejections of similar devices can help proactively identify likely issues with new devices.
• Automated documentation: AI can automate the generation of technical files and compliance documents to ensure that QMS documentation is complete and aligns with regulatory requirements.
• Data auditing and traceability: By tracking every aspect of the manufacturing process from raw materials to finished goods, AI can help create a digital audit trail that simplifies FDA and third-party audits and certifications.

AI for postmarket surveillance

Regulations such as EU MDR and the forthcoming FDA Quality Management System Regulation (QMSR) place special emphasis on postmarket surveillance, which has been a weak spot historically for many device manufacturers. Part of the issue is that manufacturers often have to sift through numerous sources of feedback, from formal complaints to emails to issues reported in the field.

Part of the issue is that manufacturers often have to sift through numerous sources of feedback, from formal complaints to emails to issues reported in the field.

AI technology is transforming postmarket surveillance by improving several areas of the process, such as:
• Data analysis and pattern recognition: Machine learning algorithms can analyze vast amounts of data from sources like adverse event reports, social media, complaint management systems, and clinical studies to detect safety signals and potential risks.
• Extracting insights from unstructured data: Natural language processing (NLP) tools are being used to glean meaningful insights from unstructured data sources like incident reports, user feedback, and medical literature.
• Proactive risk mitigation: AI-powered tools can be integrated with real-time monitoring systems to predict and prevent potential adverse events. These tools can analyze device performance data collected in real-world settings, flagging potential problems before they escalate.
• Vigilance and recall management: By monitoring devices and field data in real time, AI enables companies to respond faster to adverse events and recalls. This reduces risks to patient safety and limits the scope and impact of potential recalls, both for the business and for patients.

The FDA perspective

The FDA regulates AI in medical devices through a risk-based framework, with many AI-powered devices classified under software as a medical device (SaMD). Devices that directly control hardware, such as imaging devices, are categorized as software in a medical device (SiMD), with the level of oversight depending on risk classification and potential patient impacts.

In collaboration with international regulatory agencies like Health Canada and the U.K.’s MHRA, the FDA has established good machine learning practices (GMLPs) emphasizing:
• Transparency in algorithm training and performance
• Robustness of AI systems to changes in input data
• Ongoing monitoring to ensure algorithms are safe and effective over time

Machine learning models that evolve over time are regulated under the FDA’s predetermined change control plan (PCCP) framework, requiring manufacturers to:
• Submit plans detailing future algorithm modifications
• Validate the safety and effectiveness of changes
• Continuously monitor real-world performance

As the FDA points out, traditional medical device regulations weren’t designed for artificial intelligence and machine learning technology. This means that changes to AI and machine learning-based devices may need to undergo premarket review.

Conclusion

AI and machine learning are driving groundbreaking transformation in medical device manufacturing, ushering in a new era of innovation, efficiency, and patient-centric care. More than just AI-powered medical devices, the technology has the potential to change the entire manufacturing process, helping streamline compliance and unlock rapid innovation while ensuring patient safety.

Published Dec. 5, 2024, in the AssurX blog.