Overview of AI-enabled SaMD and SiMD devices

Artificial Intelligence (AI)-based software medical devices are revolutionizing the healthcare sector in diagnosis, treatment, and patient management. AI-powered Software as a Medical Device (SaMD) and Software in a Medical Device (SiMD) applications cover a wide range of solutions, from medical decision support systems to patient monitoring tools.

If a software is a medical device and contains AI, it must comply with the AI Act.

What are SaMD , SiMD and AI System ?

• Software as a Medical Device (SaMD): Software that operates independently and is not dependent on hardware while serving medical purposes.
• Software in a Medical Device (SiMD): Software that enhances or controls the functionality of a medical device.

• AI System: According to AI Act, ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments.

• Samples for AI based SaMD: an AI-powered eye screening application or a deep learning model used for cancer diagnosis.
• Sample for AI based SiMD:  an AI-powered pacemaker algorithm or AI software used in robotic surgery systems.

Relationship between AI-dnabled SaMD and SiMD with the AI Act

The AI Act is a horizontal regulation that applies across all sectors. MDR, on the other hand, focuses on ensuring the safety and effectiveness of medical devices. While MDR regulates AI-enabled medical devices, the AI Act imposes additional requirements on high-risk AI systems. Since MDR is a more specific regulation compared to the AI Act, medical devices must first comply with MDR and then adhere to the AI Act.

• The AI Act introduces additional security and transparency requirements for high-risk AI systems.
• A significant portion of AI-enabled SaMD and SiMD are classified as high-risk AI systems.
• AI models must be trained with ethical, reliable, and unbiased datasets.
• AI-enabled devices must meet additional requirements regarding data security, algorithm performance, and human oversight.

Classification of SaMD and SiMD under MDR

MDR (Medical Device Regulation 2017/745) regulates software medical devices through a risk-based classification system. The conformity assessment process for AI-enabled medical devices depends on the device’s classification under MDR and the additional requirements introduced by the AI Act. While classification is always directly related to the device’s intended use, it can generally be outlined as follows:

• Class I (Low Risk): Software that does not have a direct impact on patient health.
• Class IIa (Moderate Risk - Low): Software that provides diagnostic or patient management support.
• Class IIb (Moderate Risk - High): Software that directly influences medical decisions.
• Class III (High Risk): Software that manages diagnostic or therapeutic processes with life-threatening implications.

When classifying deivce, all implementing rules in Annex VIII of Regulation (EU) 2017/745 shall be considered. Special considerations on Implementing Rule 3.3 and 3.5:

The first sentence of implementing rule 3.3 of Annex VIII clarifies the regime applicable to software driving or influencing the use of a device:

‘Software, which drives or influences the use of a device, shall fall within the same class as the device’

The second sentence of implementing rule 3.3 of Annex VIII clarifies that the regime applicable to Independent MDSW:

 ‘If software is independent of any other device, it shall be classified in its own right’

In addition to that, The role of AI in the decision-making process can also be decisive in the classification:

• Decision Supportive Systems (Assistive AI): AI systems that support the doctor's decision-making process (e.g., radiology image analysis systems).
• Semi-Autonomous Systems: Systems that work with the doctor's approval and provide recommendations (e.g. AI-powered diagnostic software).
• Fully Autonomous Systems: AI systems that offer diagnosis or treatment advice without human intervention (e.g. AI algorithms that detect medical emergencies).
• Since adaptive AI systems, unlike traditional medical devices, can be continuously updated, they may be subject to additional regulations such as a “Change Control Plan”.

The development of SaMD and SiMD software in compliance with regulations is crucial for patient safety and data security. The conformity processes under MDR and the AI Act require manufacturers to ensure that their medical devices meet both technical and ethical standards. AI-based medical device developers must meticulously manage the conformity assessment processes to maintain compliance.