Skip to content

Radiology Oncology

UC-ONC-010: Extract Findings from Reports

Purpose: Extract key findings from CT/MRI/PET reports.

Property Value
Actor Radiology NLP Service
Trigger Radiology report ingested
Priority P0

Main Success Scenario: 

1. Parse radiology report text
2. Extract findings: lesion location, size, characteristics
3. Identify suspicious findings vs stable disease
4. Structure findings in oncology timeline

Alternative Flows:

Alt-1: Unstructured Report - Report lacks standard sections - Apply NLP extraction heuristics - Flag for manual review

Observability:

  • Metric: onc_rad_findings_extracted, onc_lesion_count_detected
  • Log: {"event": "rad_findings", "lesion_count": 3, "largest_size_mm": 25, "modality": "CT"}

Acceptance Criteria:

  1. [ ] Extracts lesion count and size
  2. [ ] Identifies new vs existing lesions
  3. [ ] Flags critical findings

UC-ONC-011: Detect RECIST Lesions

Purpose: Extract RECIST target lesion measurements.

Property Value
Actor RECIST Extraction Service
Trigger Radiology report with RECIST data
Priority P1

Main Success Scenario: 

1. Identify target lesions from report
2. Extract measurements (diameter in mm)
3. Calculate sum of lesion diameters
4. Store in RECIST tracking table

Acceptance Criteria:

  1. [ ] Captures up to 5 target lesions
  2. [ ] Measurements in mm accurately extracted
  3. [ ] Baseline vs follow-up distinguished

UC-ONC-012: Track Lesion Progression

Purpose: Track lesion changes over time.

Property Value
Actor Lesion Tracking Service
Trigger New RECIST measurement available
Priority P1

Main Success Scenario: 

1. Compare current lesion measurements vs baseline
2. Calculate percentage change
3. Determine trajectory: Growing, Stable, Shrinking
4. Visualize on timeline

Acceptance Criteria:

  1. [ ] Longitudinal tracking supported
  2. [ ] Growth rate calculated
  3. [ ] Timeline visualization available

Purpose: Link imaging studies to oncology treatment history.

Property Value
Actor Oncology Correlation Service
Trigger Imaging study ingested
Priority P1

Main Success Scenario: 

1. Match imaging study date to treatment cycles
2. Tag scan as: Baseline, Mid-treatment, Response assessment, Surveillance
3. Link to relevant chemo cycles or surgeries

Acceptance Criteria:

  1. [ ] Imaging linked to treatment context
  2. [ ] Supports multiple cancer diagnoses
  3. [ ] Timeline shows imaging + treatment correlation

UC-ONC-014: Auto-score Response

Purpose: Auto-calculate response (PR/CR/SD/PD) per RECIST.

Property Value
Actor RECIST Scoring Service
Trigger Follow-up RECIST measurements available
Priority P2

Main Success Scenario: 

1. Calculate % change in sum of target lesions
2. Apply RECIST 1.1 criteria:
   - CR: Complete Response
   - PR: Partial Response (≥30% decrease)
   - SD: Stable Disease
   - PD: Progressive Disease (≥20% increase)
3. Assign response category
4. Flag for physician review

Acceptance Criteria:

  1. [ ] RECIST 1.1 compliant
  2. [ ] Requires physician confirmation
  3. [ ] Response tracked over time