aimsdog
Create
SystemAIS-001Medical device and SaMD

Retinal Screening Assistant

Support ophthalmology screening services by detecting referable diabetic retinopathy in colour fundus images and prioritising cases for specialist review.

Device class
Class IIa (MDR)
FDA
Class II, 510(k)
EU AI Act
High-risk (Annex I)
Status
Approved for operation
State
Deployed
Owner
Cirocco Jones
Readiness
0
Blockers
12

Controlled evolution states, in advance, how this AI System is allowed to change. It is the source of the predetermined change control plan, not a separate document.

Where the change plan stands
3 planned modifications in the boundary, 2 with an impact assessment. The boundary, its constraints and its reassessment triggers are recorded.

Key details

The evolution boundary as agreed. Everything below is read from it.

Evolution statement
AIS-001 evolves through periodic retraining and bounded threshold adjustment. Architecture, intended purpose, inputs, outputs, and clinical workflow are outside the declared boundaries and require a new authorisation.
Strategy
Performance optimization
Review frequency
Annually, and after any breached monitoring threshold
Last updated
2026-07-14
Objectives
  • Improve sensitivity in under-represented screening populations
  • Reduce subgroup performance spread
  • Hold the referral rate within service capacity
Planned changes
  • Retraining on a newer released Dataset Version
  • Referral threshold adjustment within the declared range
  • Recalibration of the confidence output
Evolution constraints
  • No change to the intended purpose or the clinical claim
  • No change to model architecture or input modality
  • No change to the level of human oversight
  • Evaluation always on an independent Dataset Version not used in training
Regulatory reassessment triggers
  • Any change outside the declared boundary
  • Sustained subgroup sensitivity below 0.90 across two monitoring periods
  • A new user group or a new market
  • A serious incident attributable to the model output
Evidence requirements
  • Retraining verification report for each released AI Model Version
  • Site-stratified performance report against the acceptance criteria
  • Updated data governance statement when the population changes
  • Change authorisation review with an approver independent of the author
Declared boundaries
Permitted: retraining on released Dataset Versions of the registered datasets; threshold adjustment within 0.30 to 0.55. Not permitted: architecture change, new input modality, new clinical claim, autonomy change.
Monitoring plan
Monthly subgroup sensitivity and specificity by site. Monthly image-statistics drift metric. Quarterly review of override rate.
Rollback plan
Previous approved AI Model Version remains deployable for 12 months. Rollback is authorised by the AI System Owner and recorded as a Change authorisation review.

Linked records

The records that give the boundary its context.

Use caseUC-001
Primary screening triage
ModelMOD-001
Referable disease classifier
RiskRSK-001
Missed referable disease in under-represented population
RiskRSK-002
Automation bias in reviewing clinicians

Planned modifications

Part 1 of a predetermined change control plan. Each row is one specific, planned change.

ChangePM-001

Periodic retraining on extended screening archive

Implemented
Revision
2
Inside boundary
Yes
Authorised by
REV-010

ScopeRetrain the referable classifier on a newer Dataset Version of the screening archive without changing architecture, inputs, outputs, or intended purpose.

ImpactNo change to intended purpose or clinical workflow. Risk profile unchanged. Labelling unchanged.

What this feeds in the plan
Part 1Part 2(a)Part 2(b)Part 2(c)Part 2(d)Part 3

Each band is a section of the predetermined change control plan. Open the plan

ChangePM-002

Site expansion retraining

Assessed
Revision
1
Inside boundary
Yes
Authorised by
Not authorised

ScopeRetrain on a Dataset Version that adds two northern screening sites, in response to the observed drift signal.

ImpactNo change to intended purpose. Population coverage widened. Requires an updated data governance statement.

What this feeds in the plan
Part 1Part 2(a)Part 2(b)Part 2(c)Part 2(d) missingPart 3

Each band is a section of the predetermined change control plan. Open the plan

ChangePM-003

Confidence threshold adjustment

Draft
Revision
1
Inside boundary
Not assessed
Authorised by
Not authorised

ScopeAdjust the referral confidence threshold within a declared range in response to service capacity.

What this feeds in the plan
Part 1Part 2(a) missingPart 2(b)Part 2(c)Part 2(d) missingPart 3 missing

Each band is a section of the predetermined change control plan. Open the plan

Modification protocol

Part 2 of the plan, in the four parts the FDA expects: data management, re-training, performance evaluation and update procedures.

ChangePM-001Periodic retraining on extended screening archive
2(a) Data management
2(b) Re-training
2(c) Performance evaluation
2(d) Update procedures
2(a) Data management
Training data drawn only from released Dataset Versions of the registered screening archive. The evaluation set DV-002 is held out and never enters training. Site, device and demographic composition are recorded for each version.
2(b) Re-training
Retraining is triggered by the annual cycle or by a breached monitoring threshold. Frozen training protocol TP-002 with identical hyperparameters, identical preprocessing and a fixed random seed.
2(c) Performance evaluation
Sensitivity and specificity with 95 percent confidence intervals on DV-002, reported overall and stratified by site, device and age band, against the acceptance criteria below.
2(d) Update procedures
Release is authorised by a Change authorisation review with an approver independent of the author. Users are notified through the release note before deployment. The previous AI Model Version stays deployable for 12 months.
Acceptance criteria
Sensitivity not below 0.92 and specificity not below 0.86 on DV-002, with no subgroup falling more than 0.04 below the overall sensitivity.
Affected records
ModelMOD-001Model versionMV-002
Evidence
DHF-2291-RV-002
ChangePM-002Site expansion retraining
2(a) Data management
2(b) Re-training
2(c) Performance evaluation
2(d) Update procedures
2(a) Data management
Adds two northern screening sites to the archive under the same consent and provenance rules. A northern-site subset is carved out of the evaluation set and excluded from training.
2(b) Re-training
Frozen training protocol TP-002. Site-stratified sampling so no single site exceeds 30 percent of the training records.
2(c) Performance evaluation
Site-stratified sensitivity and specificity on DV-002 plus the northern-site subset, against the acceptance criteria below.
2(d) Update procedures
Not recorded
Acceptance criteria
Northern-site sensitivity not below 0.90 and overall specificity not below 0.86.
Affected records
ModelMOD-001Model versionMV-003Data versionDV-004
Evidence
None linked
ChangePM-003Confidence threshold adjustment
2(a) Data management
2(b) Re-training
2(c) Performance evaluation
2(d) Update procedures
2(a) Data management
Not recorded
2(b) Re-training
Threshold sweep on DV-002 with fixed operating-point reporting.
2(c) Performance evaluation
Sensitivity not below 0.92 at any permitted threshold.
2(d) Update procedures
Not recorded
Acceptance criteria
Sensitivity not below 0.92 at any permitted threshold.
Affected records
ModelMOD-001
Evidence
None linked

Impact assessment

Part 3 of the plan: the benefits, the risks introduced, the mitigations and the cumulative effect of the modifications together.

ChangePM-001
Assessment
No change to intended purpose or clinical workflow. Risk profile unchanged. Labelling unchanged.
Benefits
Recovers sensitivity lost to population drift and reduces the subgroup performance spread.
Risks introduced
A retrained model can shift the operating point and change the referral rate.
Mitigations
Fixed acceptance criteria, site-stratified reporting, and a monitored referral rate for the first 90 days after release.
Cumulative and collective impact
Second retraining under this plan. Cumulative sensitivity change since authorisation is within 0.01 and the intended purpose is unchanged.
ChangePM-002
Assessment
No change to intended purpose. Population coverage widened. Requires an updated data governance statement.
Benefits
Extends validated performance to the northern screening population.
Risks introduced
Population coverage widens without an updated data governance statement.
Mitigations
Not recorded
Cumulative and collective impact
Not recorded
ChangePM-003
Assessment
Not recorded
Benefits
Not recorded
Risks introduced
Not recorded
Mitigations
Not recorded
Cumulative and collective impact
Not recorded

Change authorisation

The change authorisation gate reads the same findings shown on the governance map.

GateOutcomeOpen blockers
Change authorisationHeldRISK-01, RISK-01, REV-01, MON-01, MON-01, CE-01, SIG-01, SIG-01

When the boundary, the modification protocol and the impact assessment are complete, the predetermined change control plan is assembled on the PCCP tab.