Not for diagnosis or clinical decision-making.
Back to CellSightTrace …
Data Approved sample data
- Image analysed
- Review action recorded
- Batch loaded
- Evidence opened
- Assistant answered
0 of 5 workflow steps
Release summary
Every status label and count on this page is derived from one typed release manifest. Frontend coverage is reported separately from full technical implementation.
Two separate conclusions
A frontend pass does not imply clinical, production or model-validation readiness.
Frontend proposal coverage
PASS
All 9 frontend-scope capabilities are demonstrated or classified as working prototypes and simulated concepts with acceptance evidence.
Full technical implementation
PARTIAL
7 capabilities remain blocked, missing or research candidates. They are non-frontend work and are classified transparently rather than marked complete.
Build and release metadata
Injected from the build environment. Unsupplied values are reported, never invented.
- Release version
- Unavailable — not verified
- Commit SHA
- Unavailable — not verified
- Build timestamp
- Unavailable — not verified
- Environment
- Unavailable — not verified
- Test-suite result
- Unavailable — not verified
- Acceptance-gate result
- Unavailable — not verified
Capability counts
Counted from the release manifest across all six canonical capability states.
Implemented & Demonstrated
7
Frontend capability that works in this build and has passing acceptance evidence.
Functional Prototype
1
Working deterministic frontend implementation that stands in for a future service.
Simulated Concept
1
Output produced from approved sample scenarios, never live model inference.
Research Candidate
4
Research or evaluation work that has not been carried out for this release.
Blocked
2
Requires a service that is not connected, so it cannot be demonstrated.
Missing
1
Required capability or evidence artifact that does not exist yet.
Release items
Each item carries its route, executed evidence, limitation and remaining dependency.
Image Workspace
P1 Image intake and technical quality checks
Sample intake, metadata, checksum and technical checks
Approved sample fields load with measured metadata, a verified SHA-256 checksum and five technical checks.
- Route
- /mvp/image-workspace
- Evidence
- src/test/demo-services.test.ts — approved sample analysis integrity
- Limitation
- Checks run in the browser on approved sample files only.
- Remaining dependency
- None for the frontend scope.
Implemented & DemonstratedP2 Research output, raw score and review state
Research output, uncalibrated raw score and review state
Outputs come from approved sample scenarios and are labelled as approved sample output wherever they appear.
- Route
- /mvp/image-workspace
- Evidence
- src/test/demo-services.test.ts — deterministic scenario replay
- Limitation
- Raw scores are uncalibrated scenario values and are not probabilities.
- Remaining dependency
- Score calibration study before any score can be interpreted.
Simulated ConceptP3 Review actions and activity trail
Review actions recorded against the session trace
Accept, flag and expert-review actions are written to the session activity log.
- Route
- /mvp/image-workspace
- Evidence
- src/test/demo-session.test.ts — review actions and activity log
- Limitation
- The activity log lives in this browser tab only.
- Remaining dependency
- None for the frontend scope.
Implemented & DemonstratedP4 Analysis of user-supplied images
Model inference on user-supplied images
The typed analysis interface exists; no model service is connected, so user images receive preview, metadata and technical checks only.
- Route
- /mvp/image-workspace
- Evidence
- src/test/demo-services.test.ts — live adapter returns not-connected
- Limitation
- No biological or model output is produced for user images.
- Remaining dependency
- Hosted inference service plus integration test evidence.
Blocked
Batch Analysis
P5 Deterministic batch analysis with computed metrics
36-record approved batch with computed charts and metrics
Every summary figure and chart is computed from the records held in the session.
- Route
- /mvp/batch-analysis
- Evidence
- src/test/demo-services.test.ts — derived 36/26/6/4 batch metrics
- Limitation
- Records are approved sample records, not measured runs.
- Remaining dependency
- None for the frontend scope.
Implemented & DemonstratedP6 Batch import with validation
CSV and JSON import preview with row-level validation
Invalid rows are listed with row number and reason before the reviewer confirms.
- Route
- /mvp/batch-analysis
- Evidence
- src/test/demo-services.test.ts — CSV and JSON import validation
- Limitation
- Imported files are parsed locally and never transmitted.
- Remaining dependency
- None for the frontend scope.
Implemented & Demonstrated
Evidence Hub
P7 Evidence corpus with citations and gaps
Twelve approved records with limitations, gaps and citation deep links
Records carry citation ID, version, owner, review date, state, limitations, dependencies and related records.
- Route
- /mvp/evidence-hub
- Evidence
- src/test/demo-services.test.ts — evidence deep links by citation ID
- Limitation
- The corpus is product and policy knowledge, not clinical evidence.
- Remaining dependency
- None for the frontend scope.
Implemented & Demonstrated
AI Assistant
P8 Bounded assistant with citations and refusals
Deterministic bounded retrieval with citations and refusals
Answers are extracts from approved records. Insufficient evidence is declared and clinical questions are refused.
- Route
- /mvp/ai-assistant
- Evidence
- src/test/demo-services.test.ts — supported answer, insufficient evidence, clinical refusal
- Limitation
- Deterministic lookup, not a language model and not generative.
- Remaining dependency
- None for the frontend scope; generative answering is tracked separately.
Functional PrototypeP9 Generative retrieval-augmented answering
Live generative retrieval-augmented answering
The typed assistant interface exists; no generative service is connected.
- Route
- /mvp/ai-assistant
- Evidence
- src/test/demo-services.test.ts — live assistant returns not-connected
- Limitation
- No generated text is produced anywhere in this build.
- Remaining dependency
- Server-side generative endpoint with grounded-answer acceptance tests.
Blocked
Traceability
P10 Session trace, provenance and exports
One session trace, activity log and provenance-labelled exports
One trace identifier connects every area; exports carry trace, capability status, data mode and version fields.
- Route
- /mvp/release
- Evidence
- src/test/demo-session.test.ts — export provenance fields
- Limitation
- Exports describe approved sample activity only.
- Remaining dependency
- None for the frontend scope.
Implemented & Demonstrated
Guided Demo
P11 Connected guided demonstration
Connected image → review → batch → evidence → assistant → release journey
Workflow signals are derived from real session state and persist across navigation.
- Route
- /mvp/guided-demo
- Evidence
- src/test/demo-session.test.ts — guided workflow signals and reset
- Limitation
- Signals reflect demonstration activity, not clinical workflow.
- Remaining dependency
- None for the frontend scope.
Implemented & Demonstrated
Evaluation
P12 Independent evaluation on real specimens
Independent evaluation on real specimens
No dataset, protocol or reader study exists for this release.
- Route
- /mvp/release
- Evidence
- Unavailable — not verified
- Limitation
- No performance figure of any kind may be stated.
- Remaining dependency
- Ethics approval, specimen dataset and reader study protocol.
Research CandidateP13 Score calibration study
Score calibration study
Raw scores cannot be interpreted as probabilities until calibration is completed.
- Route
- /mvp/release
- Evidence
- Unavailable — not verified
- Limitation
- Scores remain uncalibrated scenario values.
- Remaining dependency
- Held-out calibration set and calibration report.
Research CandidateP14 Site and instrument generalisation
Site and instrument generalisation study
Behaviour across scanners, stains and sites is unknown.
- Route
- /mvp/release
- Evidence
- Unavailable — not verified
- Limitation
- No generalisation claim may be made.
- Remaining dependency
- Multi-site acquisition programme.
Research Candidate
Research
P15 Backbone, OOD/QC and explainability research
EfficientNet backbone, DinoBloom features, OOD/QC and explainability research
Named research directions with no trained artefact, benchmark or model card in this release.
- Route
- /mvp/technology
- Evidence
- Unavailable — not verified
- Limitation
- Attribution regions shown in the workspace are approved sample overlays.
- Remaining dependency
- Research programme with model cards and benchmark reports.
Research Candidate
Regulatory
P16 Clinical and regulatory evidence pack
Clinical validation and regulatory evidence pack
No clinical validation, quality-system or regulatory artifact exists for this build.
- Route
- /mvp/release
- Evidence
- Unavailable — not verified
- Limitation
- The demonstrator is not a medical device and is not clinically validated.
- Remaining dependency
- Quality management system and regulatory strategy.
Missing
Non-frontend work
Blocked services, research candidates and missing evidence. These are not frontend passes.
- Blocked
Model inference on user-supplied images
Dependency: Hosted inference service plus integration test evidence.
- Blocked
Live generative retrieval-augmented answering
Dependency: Server-side generative endpoint with grounded-answer acceptance tests.
- Research Candidate
Independent evaluation on real specimens
Dependency: Ethics approval, specimen dataset and reader study protocol.
- Research Candidate
Score calibration study
Dependency: Held-out calibration set and calibration report.
- Research Candidate
Site and instrument generalisation study
Dependency: Multi-site acquisition programme.
- Research Candidate
EfficientNet backbone, DinoBloom features, OOD/QC and explainability research
Dependency: Research programme with model cards and benchmark reports.
- Missing
Clinical validation and regulatory evidence pack
Dependency: Quality management system and regulatory strategy.
Not claimed in this release
Not for diagnosis or clinical decision-making.
- • Diagnostic accuracy, sensitivity or specificity
- • Generalisation to real specimens or clinical workflows
- • Regulatory clearance, certification or clinical validation
- • Any interpretation of a raw model score as a probability or clinical risk
- • Live model inference or generative answering of any kind
Session snapshot
This browser session's trace and activity. Nothing is transmitted off-device.
- Trace ID
- …
- Session start
- …
- Data mode
- Approved sample data
- Model version
- Not supplied
- Preprocessing version
- Not supplied
- Analyses
- 0
- Review actions
- 0
- Batch records
- 0
- Evidence records opened
- 0
- Assistant turns
- 0
This record never leaves the browser except as this download.
Session activity log
Most recent 15 events, newest first.
No activity yet
Session events will appear here as you use the demonstration.