|
Novarum DX Ltd
|
Document ID: IFU059
|
Summary
2026-11-04
RARE Portal v3.1.0 extends the Performance Report with phone-variability, phone-effect, population-accounting, environmental, and temporal usability reporting. The release supports evaluation of analytical performance across the intended smartphone device matrix by adding per-phone qualitative and quantitative outputs, confidence intervals, availability states, and clearer limitations on what each section can and cannot demonstrate.
The change is intended to support IVDR Annex XIII §1.3, ISO 5725-2, and CLSI EP05-A3 style performance evaluation activities. Stored raw result values are not changed by this release.
New Features
Phone variability report modules
-
Fixed report structure: Performance Reports now use an 11-section layout covering provenance, overview, pooled qualitative and threshold analysis, analytical performance, batch variability, phone variability, phone-effect estimates, environmental factors, and capture timing/usability.
-
Population accounting: A new data-quality funnel shows movement from captured results through classifier and batch selection, ground-truth linking, technical validity, and analysed records. Missingness and phone × batch × concentration design-crossing matrices help users identify sparse or incomplete study designs before interpreting the charts.
-
Qualitative phone variability: Reports can now show per-phone confusion matrices, overlaid ROC curves, AUC with confidence intervals, and threshold-stability plots. These outputs help identify whether classifier performance changes by smartphone model.
-
Quantitative phone variability: Reports can now generate per-phone 4PL, 5PL, or linear fits within batch and test-line slices, with curve overlays, parameter summaries, concentration-level precision, and residual summaries.
-
Phone-effect and margin comparison: The report can estimate phone-model effects against a configured reference phone or pooled comparator and compare confidence intervals with saved lower and/or upper margin bounds.
-
Environmental and usability diagnostics: New sections show exploratory lux, pitch, and yaw relationships with signal, lux-banded precision, and PMF temporal plots showing frame state, orientation, and lighting through capture.
-
Inspectable calibration-fit failures: Failed batch calibration fits now retain structured failure diagnostics and concentration-versus-signal scatter inspection data, so users can investigate data quality or concentration-response anomalies rather than receiving a silent report failure.
Improvements to Existing Features
Performance report layout, filtering, and export
-
Report navigation: Reports now follow the fixed section numbering defined for v3.1.0, making screenshots, PDF exports, and review comments easier to cross-reference.
-
Result filter bar: Users can slice report outputs by test line and classifier split. A report-wide population selector remains outside this release scope.
-
Density toggle: Report content can be presented in Executive / Triage mode for compact review or Full Regulatory / DHR mode for expanded tables and matrices.
-
PDF export: PDF output is expanded and paginated with running headers and a table of contents. Available test-line and classifier-split outputs are included, and Section 11 retains the active phone-model selection in the report view.
Data preservation and telemetry handling
-
Telemetry preserved during reprocessing: Result documents can now preserve optional environmental telemetry, including device model, operating system version, lux, and orientation data recovered from PMF Story frames or linked source results.
-
No change to raw measurements: The release extends report data structures and reporting outputs but does not rewrite stored raw result values.
-
Section-specific availability: Phone-comparison sections are available only where the required phone, classifier, batch, ground-truth, and sample-count prerequisites are met. Environmental and temporal sections use their own telemetry-based availability gates.
Verification and release assurance
-
The change was assessed as low risk with no safety escalation required.
-
Verification evidence includes risk assessment, SRS, TDS, unit test reports, merged code branch, code review, manual test reports, and completed linked issues.
-
Existing pooled qualitative outputs for single-phone datasets remain unchanged in regression coverage, while v3.1.0 responses are extended with additional metrics, confidence intervals, states, and telemetry-derived outputs.
Bug Fixes
-
Corrected the Analytical Performance chart label so the reference batch is identified as Reference Batch rather than the more general Batch.
-
Fixed frozen table columns appearing above the side bar in Portal table views (Safari), including Results and Ground Truth tables.
-
Fixed batch configuration view/edit behaviour so all configured test lines are visible, rather than only one line.
-
Fixed the Show Regions display in View Results so enabled graph regions are shown correctly.
Known Issues & Limitations
-
Open bugs: One open issue is currently listed for this release-note scope: signup can show an unhelpful “Unknown error” notification instead of actionable error detail.
-
Limitations summary: The new Performance Report phone, margin, environmental, fit-diagnostic, and usability outputs are intended to support review and investigation. They should not be treated as standalone proof of device equivalence, environmental robustness, or full calibration suitability without the configured limits, study design, sample coverage, and verification evidence needed for those claims.
-
Deferred scope summary: Paired inter-phone agreement, Cohen’s kappa, per-device curve correction and its effectiveness evaluation, level 2 lux-inclusive curve correction, comprehensive comparison-specific equivalence enforcement, and automatic inclusion of every phone model in Section 11 PDF output remain outside v3.1.0.