Cancer Diagnosis¶
UC-ONC-001: Extract Tumor Location¶
Purpose: Identify primary tumor site from pathology/clinical notes.
| Property | Value |
|---|---|
| Actor | Oncology NLP Service |
| Trigger | Pathology report ingested |
| Priority | P0 |
Main Success Scenario:
1. Parse pathology report text
2. Extract tumor location using NER (e.g., "Right breast", "Colon sigmoid")
3. Map to ICD-O-3 topography code
4. Update patient oncology record
Alternative Flows:
Alt-1: Ambiguous Location
- Multiple possible sites mentioned - Flag for manual review - Log ambiguity for model retrainingAlt-2: Multiple Primary Sites
- Detect synchronous primaries - Create separate tumor records - Link tumors in patient oncology bundleObservability:
- Metric: onc_tumor_location_extracted, onc_icd_o3_mapped
- Log: {"event": "tumor_location", "site": "breast", "icd_o3": "C50.9", "confidence": 0.92}
Acceptance Criteria: 1. [ ] Accuracy > 95% for common cancer sites 2. [ ] Supports ICD-O-3 coding 3. [ ] Handles multiple primary sites
UC-ONC-002: Extract Histopathology Findings¶
Purpose: Extract HPE diagnosis and grade.
| Property | Value |
|---|---|
| Actor | Oncology NLP Service |
| Trigger | Pathology report ingested |
| Priority | P0 |
Main Success Scenario:
1. Extract histology type (e.g., "Adenocarcinoma", "Squamous cell")
2. Extract grade (Well/Moderately/Poorly differentiated)
3. Map to ICD-O-3 morphology code
4. Store in oncology bundle
Alternative Flows:
Alt-1: Grade Not Specified
- Report lacks explicit grade - Mark as "Grade Unknown" - Flag for pathologist review if criticalObservability:
- Metric: onc_histology_extracted, onc_grade_detected
- Log: {"event": "hpe_extracted", "histology": "Adenocarcinoma", "grade": "Moderately differentiated"}
Acceptance Criteria: 1. [ ] Captures grade and histology 2. [ ] ICD-O-3 morphology mapping 3. [ ] Handles variant histologies
UC-ONC-003: Extract Cancer Stage (TNM)¶
Purpose: Extract TNM staging and map to AJCC stage.
| Property | Value |
|---|---|
| Actor | TNM Extraction Service |
| Trigger | Staging report/note ingested |
| Priority | P0 |
Main Success Scenario:
1. Extract T (Tumor size), N (Nodes), M (Metastasis) from report
2. Map TNM to AJCC stage (I, II, III, IV)
3. Validate against AJCC 8th edition rules
4. Update staging in oncology record
Alternative Flows:
Alt-1: Incomplete TNM
- T, N, or M value missing - Derive stage with available data - Mark staging as "Provisional"Alt-2: Stage Revision
- New information changes stage - Update and preserve stage history - Alert tumor board if upstagedObservability:
- Metric: onc_stage_extracted, onc_tnm_complete
- Log: {"event": "tnm_staged", "T": "T2", "N": "N1", "M": "M0", "ajcc_stage": "IIB"}
Acceptance Criteria: 1. [ ] Supports AJCC 8th edition 2. [ ] Handles clinical vs pathological staging 3. [ ] Stage changes tracked longitudinally
UC-ONC-004: Detect Primary vs Metastatic¶
Purpose: Classify lesions as primary or metastatic.
| Property | Value |
|---|---|
| Actor | Oncology Classification Service |
| Trigger | Imaging/pathology finding detected |
| Priority | P1 |
Main Success Scenario:
1. Analyze lesion location against primary site
2. Check for keywords: "metastatic", "secondary", "primary"
3. Classify as: Primary, Metastatic, Unknown
4. Link metastatic lesions to primary cancer
Acceptance Criteria: 1. [ ] Differentiates primary vs metastatic correctly 2. [ ] Links metastases to primary site 3. [ ] Tracks metastatic burden
UC-ONC-005: Extract IHC Panel¶
Purpose: Extract immunohistochemistry panel results.
| Property | Value |
|---|---|
| Actor | IHC Extraction Service |
| Trigger | IHC report ingested |
| Priority | P0 |
Main Success Scenario:
1. Extract IHC markers and results (e.g., "ER: Positive", "HER2: 3+")
2. Parse percentage positivity where applicable
3. Store structured IHC results in oncology bundle
Alternative Flows:
Alt-1: Incomplete IHC Panel
- Some key markers missing from report - Extract available markers - Flag incomplete panel for follow-upObservability:
- Metric: onc_ihc_marker_extracted, onc_ihc_panel_complete
- Log: {"event": "ihc_extracted", "ER": "Positive 90%", "PR": "Positive 70%", "HER2": "2+"}
Acceptance Criteria: 1. [ ] Captures all common IHC markers 2. [ ] Handles percentage and intensity scoring 3. [ ] Supports breast, GI, lung cancer panels
UC-ONC-006: Interpret Biomarkers¶
Purpose: Interpret biomarker results (ER/PR, HER2, KRAS, EGFR).
| Property | Value |
|---|---|
| Actor | Biomarker Interpretation Service |
| Trigger | IHC/molecular test result available |
| Priority | P0 |
Main Success Scenario:
1. Read biomarker results from structured data
2. Apply interpretation rules (e.g., HER2 3+ = Positive)
3. Suggest treatment implications if configured
4. Flag actionable mutations
Alternative Flows:
Alt-1: Conflicting Results
- IHC vs molecular test conflict (e.g., HER2 2+ vs FISH negative) - Flag discrepancy - Defer to molecular/FISH resultObservability:
- Metric: onc_biomarker_interpreted, onc_actionable_mutation_flagged
- Log: {"event": "biomarker_interpreted", "HER2": "Positive", "treatment_impact": "Herceptin-eligible"}
Acceptance Criteria: 1. [ ] Supports ER/PR, HER2, KRAS, EGFR, PD-L1 2. [ ] Rule-based interpretation accurate 3. [ ] Links biomarkers to treatment options