Index
3-min read
Front Matter
tech-feasibility
lab-hygiene-sentinel
index
A clinical lab technician must wash their hands between handling any two instruments. Today, no system enforces this — violations go undetected and undocumented. This document is a 3-minute summary of the full feasibility analysis for an automated enforcement system.
What It Does¶
When a technician picks up an instrument, the system starts a 30-second window. If a second instrument is picked up before the technician washes their hands, the system fires an in-lab alarm, sends a push notification to a supervisor, and attaches a video clip of the violation to the incident record. The rule repeats for every instrument transition, all shift.
How It Works (One Paragraph)¶
RFID tags on every instrument + RFID wristbands on each technician identify who picked up what. A soap sensor at the wash station confirms a hand-wash event. Cameras at the bench and wash station provide a second confirmation layer and supply the evidence clip on violation. A deterministic state machine — not ML inference — evaluates the sequence and triggers alerts. The protocol logic is a simple three-state machine: idle → holding → wash-window → idle (clean) / violation.
Verdict¶
| Dimension | Assessment |
|---|---|
| Instrument detection (RFID + CV) | Feasible with caveats — on-metal tag bench test required |
| Hand-wash detection (soap sensor) | Straightforward |
| Person attribution (~5 technicians) | Feasible — wristband RFID recommended |
| Real-time alarm | Straightforward |
| Supervisor push + video clip | Feasible — clip latency must be measured |
| Regulatory compliance (ISO 15189/GLP) | Unknown — needs one conversation with quality manager |
| Optional: compliance history + reports | Straightforward once core log is stable |
The system is buildable. The core protocol logic is deterministic and low-risk. The two genuine risks are: (1) on-metal RFID tag compatibility per instrument — resolved by a bench test before deployment; (2) unconfirmed regulatory obligation — resolved by asking the lab's quality manager one question.
The longest pole is not code. It is physically tagging every instrument in the lab and maintaining that registry over time.
Key Assumptions to Confirm¶
All parameters below are design assumptions — none were confirmed by the lab. The architecture holds under each one, but they must be validated before sprint 1.
| Assumption | Default used | Confirm with |
|---|---|---|
| Wash window duration | 30 seconds | Lab protocol documentation |
| Number of concurrent technicians | ~5 | Lab operations |
| Alert channel | Supervisor phone app | IT + operations |
| Regulatory obligation | ISO 15189 / GLP may apply | Quality manager |
| Wristband RFID acceptable | Yes | Lab operations team |
What Gets Built, In Order¶
Spine: Sensor Edge → Event Broker → State Machine → Alert Dispatcher (~9 weeks)
Parallel tracks once spine nodes unlock: CV redundancy layer, clip store, compliance history, shift reports.
For the full build sequence and Gantt, see analysis.md.
The Four Open Questions¶
- Does ISO 15189 or GLP apply? (determines storage design)
- Will technicians wear wristbands? (determines attribution approach)
- Is there existing camera infrastructure? (determines Sensor Edge scope)
- iOS, Android, or web dashboard for supervisor alerts? (determines app platform)
Each is unblocked by a single stakeholder conversation — none require engineering investigation.