Downstream Integration — Dependency Chain and Responsibilities#
Four-Layer Dependency Chain#
Layer 1: MADDENING (open source, LGPL, general-purpose framework)
↓
Layer 2: MIME (open source, LGPL, microrobotics physics engine)
↓
Layer 3: MICROBOTICA (open source, LGPL, research simulator)
↓
Layer 4: [Commercial Product] (regulated, CE-marked, built by commercial entity)
Layer |
Status |
Regulatory Obligation |
|---|---|---|
MADDENING |
Open-source research tool |
None (provides SOUP documentation voluntarily) |
MIME |
Open-source research tool |
None (provides SOUP documentation voluntarily) |
MICROBOTICA |
Open-source research tool |
None (provides SOUP documentation voluntarily) |
Commercial Product |
Regulated medical device |
Full EU MDR manufacturer obligations |
Per-Layer Responsibility Details#
Layer 1: MADDENING#
Role: General-purpose JAX-based multiphysics simulation framework.
Provides:
Core simulation infrastructure: graph management, JIT compilation, multi-rate scheduling, adaptive timestepping
Physics node library: heat diffusion, rigid body dynamics, spring-damper, LBM fluid
Neural surrogate framework: training, validation, architecture library
Compliance infrastructure:
NodeMeta, anomaly registry, verification benchmark registry,@stabilitydecorator,HealthCheckNodeSOUP documentation package (
docs/validation/soup_package.md)
Does NOT provide:
Clinical claims or intended clinical use
Domain-specific physics for microrobotics (CSF flow, Brownian motion, magnetic actuation)
Risk management file (ISO 14971) — only
hazard_hintsfor downstream risk analysisQMS, clinical evaluation, or post-market surveillance
SOUP classification: When used in a regulated product, MADDENING is IEC 62304 SOUP. Its safety class is determined by the downstream manufacturer based on the context of use.
Layer 2: MIME (MIcrorobotics Multiphysics Engine)#
Role: Domain-specific physics engine for microrobot simulation, built on MADDENING.
Provides:
CSF flow simulation nodes (wrapping or extending LBMPipeNode)
Magnetic field and actuation models
Particle dynamics for microrobots (Brownian, drag, contact)
Domain-specific
HealthCheckNodeconfigurations (e.g., CSF density bounds, velocity limits)Own SOUP documentation, anomaly registry, and verification benchmarks (using
MIME-prefix)
Inherits from MADDENING:
Compliance schema types (
NodeMeta,AnomalyRecord, etc.) viamaddening.complianceAnomaly registry validator CLI
@verification_benchmarkand@stabilitydecoratorsHealthCheckNodebase class
SOUP classification: When used in a regulated product, MIME is IEC 62304 SOUP. MADDENING is MIME’s own SOUP dependency (SOUP-of-SOUP).
Layer 3: MICROBOTICA#
Role: Open-source research simulator for microrobot-assisted drug delivery in CSF.
Provides:
Pre-configured simulation scenarios for cerebrospinal fluid geometries
Visualization and analysis tools for researchers
Training data pipelines for reinforcement learning research
Own SOUP documentation, anomaly registry, and verification benchmarks (using
MICROBOTICA-orMBOT-prefix)
Inherits from MIME: All MIME compliance infrastructure, plus MIME-specific physics and verification evidence.
SOUP classification: When used in a regulated product, MICROBOTICA is IEC 62304 SOUP.
Layer 4: Commercial Product#
Role: CE-marked medical device software built by a downstream commercial entity.
This is the ONLY layer subject to EU MDR. The commercial entity is the EU MDR manufacturer (Article 2(30)).
Commercial Responsibility Statement#
The Layer 4 commercial entity is the EU MDR manufacturer (Article 2(30)) and is solely responsible for all regulatory obligations, including but not limited to:
Establishing and maintaining an ISO 13485 QMS certified by a Notified Body
Performing the conformity assessment procedure appropriate to the device class (Class IIb: Annex IX or Annex X+XI; Class III: Annex IX or Annex X+XI with full technical file review)
Engaging and satisfying the Notified Body for initial conformity assessment and annual surveillance audits
Building and maintaining the clinical evaluation report (CER) with sufficient clinical evidence
Operating post-market surveillance including PSUR (annual for Class III), vigilance reporting, and PMCF studies
Bearing manufacturer liability for the device placed on the EU market
Performing SOUP assessment of all open-source layers (MADDENING, MIME, MICROBOTICA) per IEC 62304 Clause 5.3.3/5.3.4
Maintaining risk management file (ISO 14971) incorporating SOUP failure modes
MADDENING provides SOUP documentation to support the commercial entity’s regulatory obligations. This provision is voluntary and does not make MADDENING subject to EU MDR or IEC 62304.
Classification-by-Use-Mode Analysis#
The EU MDR classification of the Layer 4 commercial product depends on its intended use:
Use Mode |
Classification |
Rationale |
|---|---|---|
Research / pre-clinical simulation |
Not a medical device |
No clinical decision-making; general-purpose tool (MDCG 2019-11 §3.2) |
Treatment planning (pre-operative) |
Class IIa |
Software providing information used to make decisions with diagnosis or treatment purposes (Rule 11, Annex VIII); non-immediate clinical context |
Intraoperative digital twin |
Class IIb minimum, likely Class III |
Real-time surgical decision support; erroneous output during live procedure in CSF spaces adjacent to striatum/substantia nigra could cause irreversible neurological harm (Rule 11(b)) |
RL policy generation for microrobot control |
Class III |
Software directly informs an active control loop for a device operating in the CNS; failure could result in death or serious injury (Rule 11(b), Rule 22) |
The MADDENING documentation architecture is designed to support the most demanding classification (Class III / IEC 62304 Class C) so that less demanding use cases are automatically covered.
Commercial Boundary Documentation#
What MADDENING provides to the commercial entity#
SOUP Package (
docs/validation/soup_package.md): Software identification, functional description, known anomalies, verification evidence, lifecycle activities, configuration management, anomaly management policy, dependency listKnown Anomalies Registry (
docs/validation/known_anomalies.yaml): Machine-readable, versioned, validated by CIVerification Evidence: Registered benchmarks with analytical solutions, convergence studies, and regression tests
Hazard Hints: Per-node qualitative risk information for integration into the manufacturer’s ISO 14971 risk management file
Algorithm Guides: Per-node documentation of governing equations, discretization, assumptions, limitations, and validated regimes — with CI-validated bibliography citations (
docs/bibliography.bib) and CI-validated implementation-to-code mappingsHealthCheckNode: Configurable execution-layer fault detection (NaN/Inf, bounds, moment checks) for use in downstream safety monitoring
Compliance Schema Types: Importable Python types (
NodeMeta,AnomalyRecord, etc.) for downstream libraries to use and extendSBOM: CycloneDX-format software bill of materials attached to each release
What MADDENING does NOT provide#
Clinical safety assessment
Intended use for any specific clinical application
Risk management file (ISO 14971)
Clinical evaluation report (CER)
Post-market surveillance plan
QMS documentation (ISO 13485)
Notified Body relationship
End-user training materials for clinical use
Cybersecurity risk assessment for the deployed product (MADDENING assumes trusted inputs; see
docs/regulatory/intended_use.md)
LGPL Replaceability#
MADDENING is licensed under LGPL-3.0-or-later. The LGPL “replaceability” obligation is satisfied via Python’s module system: the end user can replace the maddening package with a modified version by installing it into the same Python environment, with no special linking or build steps required. The commercial entity should document this replaceability mechanism in their technical file.