Assuring the safety of on-demand medical cyber-physical systems

Andrew L. King, Lu Feng, Oleg Sokolsky, Insup Lee

TL;DR

This paper introduces a novel approach to ensure the safety of "on-demand" medical cyber-physical systems (MCPS) that are assembled at the bedside. They propose a model-based framework using a formal specification language and a Medical Application Platform (MAP) to verify safety properties of these systems before they are deployed with actual devices.

What Problem Does This Paper Address?

Modern healthcare increasingly relies on interconnected medical devices to deliver patient care. The trend is moving towards "on-demand" systems, where devices are assembled ad-hoc to meet specific patient needs in a given clinical scenario. These systems, termed Virtual Medical Devices (VMDs), are problematic because their components are not known a priori, making traditional safety assurance methods (which evaluate the complete, fixed system) insufficient. The challenge is to guarantee the safety of these dynamically assembled systems when the exact make and model of devices are unknown until the moment of assembly.

Methods

The authors propose a two-pronged approach:

  1. Formal Modeling Language: They developed a language to model specific clinical scenarios. This language allows for:
  2. Modular Specification: Defining components (devices and logic modules) independently.
  3. Formal Semantics: Enabling rigorous analysis and verification.
  4. Expressing Functional and Timing Variability: Distinguishing between "must" (required) and "may" (optional) behaviors and timing constraints of devices. This is crucial for handling the diversity of real-world devices.
  5. Compositional Refinement: Allowing models to be instantiated with actual devices that are compatible with the specified requirements.
  6. Medical Application Platform (MAP): This acts as a trusted base for VMD instantiation. Its key roles include:
  7. Device Compatibility Check: Verifying that selected physical devices meet the VMD's specification requirements.
  8. Logic Module Hosting: Deploying and executing the software components of the VMD.
  9. Resource and Timing Management: Ensuring that the VMD's performance and timing constraints are met during execution, using schedulability tests.

The core idea is that if a VMD model is verified to be safe, and the MAP ensures that the instantiated VMD correctly refines this model, then the safety properties will transfer to the actual running system.

Datasets

Not discussed in the paper. The paper focuses on a conceptual framework and a motivating example (PCA pain management) rather than empirical evaluation on specific datasets.

Key Insights and Ideas

  • Shifting Safety Assurance: The paper advocates for a paradigm shift in safety assurance for medical devices, moving from evaluating fixed systems to verifying dynamically assembled ones.
  • Virtual Medical Devices (VMDs): The concept of a VMD is central, representing a system assembled from interoperable devices at the point of care.
  • Handling Variability is Key: The formal language's ability to express both functional and timing variability ("must" vs. "may") is critical for bridging the gap between abstract models and diverse real-world devices.
  • MAP as a Trusted Enforcer: The MAP is not just a deployment platform but a crucial safety enforcer, guaranteeing that instantiated VMDs adhere to the verified model's properties.
  • Model-Based Safety Transfer: The fundamental claim is that safety properties verified at the model level can be transferred to the instantiated system if the instantiation process is correctly managed by the MAP.

Key Results

The paper presents a framework and a detailed specification for a PCA infusion safety interlock VMD. While no quantitative performance metrics are provided, the key result is the conceptual demonstration of how to:

  • Formally model a dynamic medical system.
  • Specify device requirements and logic module behaviors with timing constraints.
  • Outline a platform (MAP) that can instantiate these VMDs while ensuring safety.

The PCA example illustrates how a safety interlock can be modeled, specifying device interfaces, data flows, and the controller's logic, including timing constraints for safety.

Limitations and Future Work

The authors acknowledge several limitations and areas for future research:

  • Regulatory Approval: It is unclear how this model-based approach will fare in regulatory approval processes for VMDs. Extensive consultation with regulators is needed.
  • Faulty Behavior: The current models focus on nominal device behaviors. Assessing faulty behaviors is crucial for safety, and the language needs to be extended to capture and distinguish these effectively.
  • Risk Management Adaptation: Adapting current medical device risk management practices for VMDs is an open challenge. Even with accurate models, unforeseen implementation behaviors are possible, requiring specific risk management strategies for VMDs.
  • Concurrent VMD Execution: The MAP needs to support the concurrent execution of multiple VMDs, which is a complex engineering challenge.
  • Extension of Modeling Language: Further work is needed to fully integrate timing specifications and explore compositional property-preserving refinement in detail.

Summary updated

Abstract

We present an approach to establish safety of on-demand medical cyber-physical systems which are assembled to treat a patient in a specific clinical scenario. We treat such a system as a virtual medial device (VMD) and propose a model-based framework that includes a modeling language with formal semantics and a medical application platform (MAP) that provides the necessary deployment support for the VMD models.
BibTeX
@inproceedings{king2013assuring,
  author = {Andrew L. King and Lu Feng and Oleg Sokolsky and Insup Lee},
  title = {{Assuring the safety of on-demand medical cyber-physical systems}},
  booktitle = {CPSNA},
  year = {2013},
  doi = {10.1109/cpsna.2013.6614238},
}
Figure