Why it matters
Without AI interpretation, reliance on manual reading risks missed internal features, inconsistent documentation, delayed treatments, and unresolved safety concerns that undermine loan readiness and insurer confidence.
From design to safety
Design
AI-supported interpretation goals
Define data needs, treatment implications, and documentation standards from X-ray interpretation, including internal structure maps, weave patterns, and pentimenti, guiding method selection, reversibility checks, and PACR-compliant reporting. Specify data governance, provenance trails, and linked condition reports for every plate. Establish acceptance criteria for AI-generated annotations, including confidence thresholds and human-review gates.
- X-ray cabinet
- Digital detector panel
- Canvas weave sampling grid
Build
Integration and configuration
Install imaging-analysis interfaces, calibrate detectors, and establish reporting templates, ensuring compatibility with PACS, condition reports, and treatment proposals. Implement version-controlled templates, audit logs, and role-based access to imaging archives, with on-site test objects for validation. Define safe-fail paths and backup imaging routes for on-site or remote review.
- Lead shielding cabinet
- Radiation dosimeter badge
- Canvas weave sampling grid
Operation
Routine interpretation workflow
AI models process X-ray images, annotate internal features, compare related plates, and generate preliminary notes for condition and proposed actions. Include cross-object stitching, plate-to-plate comparison, pentimenti correlation maps, and a reversible-change log linked to treatment proposals and condition reports.
- X-ray cabinet
- Digital detector panel
- Lead shielding cabinet
Maintenance
Calibration and data hygiene
Regularly recalibrate detectors, validate AI outputs with expert reviews, and archive radiographs with metadata for future condition checks. Maintain a formal data-drift monitoring regime, run quarterly revalidation with benchmark cases, and preserve full imaging lineage from capture to final report.
- Radiation dosimeter badge
- Digital detector panel
- Canvas weave sampling grid
Safety
Radiation safety governance
Monitor exposure records, ensure shielding integrity, and enforce COSHH practices with sign-off from the accredited conservator. Maintain radiation safety drills, PPE logs, and incident reporting, with annual COSHH review tied to laser and X-ray operations in the imaging suite.
- X-ray cabinet
- Lead shielding cabinet
- Radiation dosimeter badge
Part by part
- 1
X-ray cabinet
- Watch for
- Detector drift or chamber leakage that alters image quality.
- AI action
- AI flags image quality anomalies and prompts recalibration task.
- 2
Digital detector panel
- Watch for
- Pixel patch inconsistencies or artifacts masking features.
- AI action
- AI highlights artifacts and suggests re-acquisition angles.
- 3
Canvas weave sampling grid
- Watch for
- Weave disruption patterns or cross-weave distortions.
- AI action
- AI maps weave and correlates to internal features.
- 4
Lead shielding cabinet
- Watch for
- Shield integrity erosion or door mis-seal.
- AI action
- AI alerts safety officer if shielding fails.
- 5
Radiation dosimeter badge
- Watch for
- Cumulative exposure approaching limits.
- AI action
- AI logs exposure and notifies supervisor and COSHH file.
- 6
Canvas pentimenti reference slate
- Watch for
- Undetected pentimenti correlating with X-ray features.
- AI action
- AI annotates pentimenti candidates for expert review.
Worked example
Detecting hidden composite structure in a panel portrait
Illustrative scenario with sample data.
A mid-17th century panel painting requires internal examination to assess previous restorations and structural support.
Signal in
plate_id: P-2048; detectorID: DPT-11; exposure: 40kV; exposureTime: 6s; energy: 120 keV; objectID: OBJ-741; current_condition: stable; operator: tech2; date: 2026-04-18
System out
annotation: internal re-lamination near center; pentimenti region identified; weave disruption map cross-referenced with pentimenti; recommended: confirm with on-site tomography and update treatment proposal; actionItem: schedule tomography session; responsible: Head Conservator; dueDate: 2026-05-02
Human sign-off: Head conservator approves the interpretation notes, requests a follow-up imaging plan, and assigns tomography and reporting tasks to the imaging team.
How to roll it out
Phase 1
Weeks 1 to 2
Baseline integration of AI interpretation with existing X-ray workflows and documentation templates.
Phase 2
Weeks 3 to 6
Pilot on two objects, calibrate models against expert readings, and refine reporting outputs.
Phase 3
Weeks 7 to 12
Full deployment with archival metadata, QA procedures, and formal COSHH alignment.
Where people stay in charge
- Only a qualified conservator or conservation scientist may approve final treatment decisions based on AI outputs.
- All radiation-related risk assessments must be signed off by the accredited conservator before imaging begins.
- AI outputs must be reviewed against high-resolution images and traditional expert readings prior to reporting to clients.
Questions owners and crews ask
What data formats does the AI system require for X-ray interpretation?
The system accepts standard radiography formats with embedded exposure data, calibration data, and metadata, enabling consistent comparison across sessions and objects.
Who validates AI-generated annotations during treatment planning?
A licensed conservator with relevant training validates AI annotations, cross-checks with physical examination, and ensures reversibility and ethics compliance.
How is safety managed in AI-assisted X-ray workflows?
Radiation safety follows COSHH, with calibrated dosimeters, shielding checks, exposure logging, and sign-off by the responsible conservator before imaging.
Can AI outputs affect insurance or loan decisions?
Yes, AI-augmented condition reporting and imaging interpretations feed into insurer and lender assessments, subject to professional review and archival documentation.