Medical devices and SaMD platforms are not collections of siloed documents; they are complex living systems of interconnected physics, software, patient physiology, and regulatory mandates. The PMX Knowledge Graph establishes a mathematically verifiable, high-dimensional semantic graph connecting every engineering component directly to patient safety and global regulatory submissions.
For three decades, MedTech engineering teams have attempted to satisfy 21 CFR 820.30 and ISO 13485 design controls using document folders, PLM tables, and relational SQL databases. Here is why those tools inevitably collapse under product evolution.
A static Excel matrix or Word table gives the illusion of traceability on the day of submission. However, documents are dead snapshots. The moment an engineer changes a sensor resistor value on a schematic, the risk spreadsheet stored in SharePoint remains untouched, creating invisible compliance divergence.
Relational tables require predefined schemas and rigid foreign keys. In MedTech, dependencies span 12+ degrees of transitive depth (Component → Requirement → Risk → Verification → Supplier → Submissions). Traversing these across SQL tables results in fragile 15-table JOIN queries that grind systems to a halt.
Without directed acyclic graph (DAG) invariants, modern agile engineering creates hundreds of orphaned objects: software unit tests testing obsolete requirements, risk controls mitigating deleted hazards, and supplier specs certifying out-of-production silicon.
Live visualization of the CardioSense Pro v3.8.2 design control graph. Click any node to open the real-time metadata inspector drawer.
Use mouse drag to pan across the topological space. Use scroll wheel or (+) (-) buttons to zoom. Filter by semantic object type above.
Every node and edge in PMX OS conforms to strict mathematical rigor, enforcing 21 CFR Part 11 integrity and automated ISO 13485 design control synthesis.
Heterogeneous typed vertices representing regulated entities: Physical Components (HW), Logical Modules (SW), Design Inputs (REQ), Failure Modes (RSK), Verification Protocols (VER), Clinical Claims (CLN), and Regulatory Clauses (REG). Each node possesses a globally unique SHA-256 identifier and canonical schema.
Strongly-typed semantic vectors establishing causality: implements, mitigates, verified_by, submitted_under, polled_by. Edges maintain mathematical directionality to prevent circular logic and enable bi-directional depth traversal.
Nodes do not store arbitrary strings. They enforce parametric engineering values: numerical thresholds (<41.0°C), sampling rates (100 Hz), tolerances (±0.1°C), biocompatibility standards (ISO 10993-10), and unit dimensions verified at runtime.
Full bi-temporal graph indexing. PMX OS tracks both system time (when a change was recorded) and effective time (when an engineering change goes live on the manufacturing line). Reconstruct the exact DHF baseline for any day in history.
Every verification edge anchors directly to an immutable raw data artifact: machine execution logs, thermal chamber thermocouple telemetry, digital oscilloscope waveforms, and clinical biostatistical outputs, hashed with SHA-256 for audit immutability.
Nodes enforce strict ownership hierarchies across engineering, quality, and regulatory roles. A software engineer cannot approve a hazard risk control without the cryptographic digital co-signature of the Lead Risk Assessor (21 CFR Part 11 compliant).
State transitions follow formal statecharts: Draft → In-Review → Verified → Approved → Obsoleted. If an upstream node enters Draft due to an ECO, all downstream edges automatically transition to Suspect.
Every addition, modification, or severance of an ontological edge is recorded in an append-only cryptographic event log. Auditing inspectors can trace the complete provenance of any design decision back to original patient safety requirements.
Observe how a single thermistor sensor change cascades through the entire medical product lifecycle, establishing continuous mathematical proof for global regulators.
Transducer specification: Negative Temperature Coefficient (NTC) thermistor mounted directly on the patient-contacting lower chassis. Sourced from Supplier Y (MicroSens Corp) under QMS Spec QS-409.
"The patient-contact surface of the CardioSense Pro enclosure shall not exceed 41.0°C during 36 hours of continuous operation under normal conditions and single-fault thermal boundary conditions (ambient 38.0°C)." Mandated by IEC 60601-1-11 Clause 11.1.2.2.
Hazard: Uncontrolled dissipation resulting in low-grade skin thermal erythema or burn during continuous multi-day cardiac monitoring. Initial Severity: Critical (4) | Occurrence: Remote (2). Mitigation Requirement: Hardware dual-sensor cutoff + firmware auto-shutdown at 41.5°C. Residual Risk: Minor (1) | Improbable (1) — ALARP Accepted.
Bench test execution on calibrated skin phantom with ambient incubator held at 38.0±0.5°C for 36 continuous hours. 432 data points collected. Peak surface temperature recorded: 40.2°C (±0.15°C margin of safety). Cryptographic Raw Data Hash: e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855.
The entire design chain — from thermistor silicon characteristics to bench test V-0183 and clinical trial CLN-088 — is compiled into eSTAR Volume 9 and EU MDR Annex II Technical File Section 4. Regulatory reviewers can audit the unbroken provenance chain in seconds.
Transition your medical device design controls from static dead documents to a living, verifiable knowledge graph. See how your engineering changes propagate in real time.