lev-os Creating Teaching Files
Curates and annotates imaging cases for educational purposes with clinical correlation.
Why This Skill Exists
Teaching files are the primary educational resource in radiology residency programs and continuing education. The ACR and ABR require that residency programs maintain teaching-file collections that cover the core curriculum, and the ABR Core and Certifying examinations draw heavily on case-based image interpretation. Well-curated teaching files accelerate pattern recognition, reinforce diagnostic criteria, and provide a searchable reference for rare or classic findings. The MIRC (Medical Imaging Resource Center) teaching-file standard, developed by RSNA, provides the metadata framework for creating searchable, shareable cases.
Beyond training, teaching files support research, quality improvement, and medicolegal documentation of unusual cases. Poorly constructed teaching files — lacking clinical context, pathologic confirmation, or proper de-identification — are educationally useless and may violate HIPAA. This skill provides the systematic framework for case selection, image annotation, clinical correlation, and HIPAA-compliant publication of teaching-file cases.
Checkpoint A: Pre-Draft Intake (Mandatory)
- What is the educational purpose of this case? (Default: Classic presentation, diagnostic pitfall, rare entity, or technique demonstration)
- What modality and body region? (Default: Identify — CT chest, MRI brain, US thyroid, etc.)
- Is there pathologic or surgical confirmation? (Default: Strongly preferred — document source)
- Has the case been de-identified per HIPAA? (Default: Required before any sharing — verify removal of all 18 identifiers)
- What is the target audience? (Default: Radiology residents — specify if fellows, technologists, referring clinicians, or CME)
- What teaching-file system will this be published to? (Default: Institutional PACS teaching folder, MIRC, Radiopaedia, institutional LMS)
- Is patient consent required for publication? (Default: De-identified cases generally exempt per HIPAA; check institutional IRB policy)
Documents to Request
- Imaging study with key images identified
- Complete radiology report
- Clinical history, presentation, and outcome
- Pathology/surgical report for confirmation
- Relevant laboratory data
- Follow-up imaging (if demonstrates disease evolution)
- Institutional HIPAA de-identification policy
- MIRC metadata template (if publishing to MIRC-compliant system)
Step 1: Case Selection Criteria
What Makes a Good Teaching Case?
| Criterion | Excellent | Adequate | Poor |
|---|---|---|---|
| Diagnostic certainty | Pathology/surgical confirmation | Strong clinical-imaging correlation | Presumptive diagnosis only |
| Image quality | Optimal technique; clear demonstration | Adequate but not ideal | Poor quality; finding barely visible |
| Educational value | Classic presentation of important entity | Reasonable example | Obscure variant with limited teaching impact |
| Clinical completeness | Full history + labs + imaging + path + outcome | History + imaging + path | Imaging only with minimal history |
| Uniqueness | Fills a gap in existing collection | Useful even if similar cases exist | Duplicate of existing case with no added value |
Case Categories to Build a Comprehensive Collection
| Category | Purpose | Examples |
|---|---|---|
| Can't-Miss Cases | Life-threatening findings every radiologist must recognize | Aortic dissection, tension PTX, saddle PE, cord compression, bowel ischemia |
| Classic Presentations | Textbook appearances for board preparation | Pancoast tumor, horseshoe kidney, Crohn terminal ileum, meningioma with dural tail |
| Diagnostic Pitfalls | Commonly missed or misinterpreted findings | Scaphoid fracture, posterior malleolus, retrocardiac pneumonia, pseudopneumothorax |
| Mimics and Look-Alikes | Entities that resemble each other | Round pneumonia vs. lung mass, thymic rebound vs. lymphoma, lipoma vs. liposarcoma |
| Rare Entities | Uncommon conditions for exposure | Erdheim-Chester disease, extramedullary hematopoiesis, pneumatosis cystoides intestinalis |
| Technique Demonstrations | Protocol or technical teaching | Adrenal washout, LI-RADS lesion on multiphase MRI, PI-RADS scoring on mpMRI |
| Complications | Post-procedural or treatment-related | Post-biopsy pneumothorax, radiation pneumonitis, immunotherapy colitis |
Step 2: Image Selection and Annotation
Key Image Selection Rules
- Include the finding — at least one image clearly demonstrating the primary finding
- Include context — surrounding anatomy to localize the finding
- Include multiple planes — axial, coronal, sagittal when cross-sectional
- Include comparison — normal side for paired structures; prior study for interval change
- Include all relevant sequences/phases — DWI + ADC for stroke; arterial + venous for liver mass
- Include pathology correlation — gross specimen or histology when available
Annotation Standards
| Element | Guideline |
|---|---|
| Arrows/circles | Point to the finding; use consistently colored annotations |
| Labels | Anatomic labels on key structures; use standard abbreviations |
| Measurements | Include caliper measurements where relevant |
| Window/level | Show appropriate window settings (lung window for nodules, bone window for fractures) |
| Comparative images | Side-by-side layout for before/after or normal/abnormal |
| De-identification | Remove all text overlays containing PHI before annotating |
Annotation Tools
- PACS annotation tools (arrows, text, measurements)
- Dedicated teaching-file software (MIRC, Pacsbin)
- Presentation software (PowerPoint, Keynote) for formatted case layout
- Open-source tools (3D Slicer, OHIF viewer) for advanced annotations
Step 3: Clinical Correlation and Case Write-Up
Teaching File Case Structure
1. Clinical Presentation
- Patient age range and sex (not exact age — de-identification)
- Chief complaint and relevant history
- Relevant physical exam findings
- Pertinent laboratory values
- Clinical question prompting the imaging study
2. Imaging Findings
- Modality, technique, and relevant protocol
- Systematic description of findings (as if writing the report)
- Annotated key images with findings highlighted
- Differential diagnosis (ranked by likelihood)
3. Diagnosis
- Final diagnosis with supporting evidence
- Pathology/surgical correlation (if available)
- How the diagnosis was confirmed
4. Discussion
- Disease background: etiology, epidemiology, pathophysiology
- Imaging features: classic and variant appearances across modalities
- Differential diagnosis discussion: how to distinguish from mimics
- Classification systems applied (BI-RADS, LI-RADS, etc.)
- Management implications based on imaging findings
- Relevant literature citations (key references, not exhaustive)
5. Teaching Points
- 3–5 key takeaways in bullet-point format
- Common pitfalls and how to avoid them
- Board-relevant pearls
- Links to related cases in the collection
Step 4: HIPAA De-Identification
18 HIPAA Identifiers That Must Be Removed
| # | Identifier | Where to Check |
|---|---|---|
| 1 | Name | Image headers, annotations, overlays |
| 2 | Geographic data (smaller than state) | Image headers |
| 3 | Dates (except year) for dates related to patient | Exam dates, DOB, admission dates |
| 4 | Telephone numbers | Report headers |
| 5 | Fax numbers | Report headers |
| 6 | Email addresses | Report headers |
| 7 | Social Security numbers | Should never appear in imaging |
| 8 | Medical record numbers | Image headers, overlays |
| 9 | Health plan beneficiary numbers | Report metadata |
| 10 | Account numbers | Accession numbers, billing |
| 11 | Certificate/license numbers | Provider data in headers |
| 12 | Vehicle identifiers | N/A for most radiology |
| 13 | Device identifiers/serial numbers | Implant data in headers |
| 14 | Web URLs | N/A for most radiology |
| 15 | IP addresses | DICOM metadata |
| 16 | Biometric identifiers | N/A for most radiology |
| 17 | Full-face photographs | Scout images, 3D reconstructions |
| 18 | Any other unique identifying number | Accession numbers in overlays |
De-Identification Methods
- DICOM anonymization tools: CTP (Clinical Trial Processor), DicomCleaner, PACS built-in tools
- Burned-in pixel data: Must be removed or obscured; DICOM header cleaning alone is insufficient
- 3D facial reconstructions: Must be defaced or excluded — facial CT/MRI 3D reconstructions can identify patients
- Verification: Review every exported image for residual identifiers before publishing
Step 5: Publication and Metadata
MIRC/RSNA Teaching File Metadata
| Field | Description | Example |
|---|---|---|
| Title | Descriptive case title | "Classic LI-RADS 5 Hepatocellular Carcinoma on Multiphase MRI" |
| Author | Creating radiologist | "Dr. [Name], Department of Radiology" |
| Abstract | 2–3 sentence case summary | Brief clinical context + key finding + diagnosis |
| Keywords | Searchable terms (RadLex preferred) | "hepatocellular carcinoma, LI-RADS, arterial phase hyperenhancement, washout" |
| Modality | DICOM modality codes | CT, MR, US, NM, XR |
| Anatomy | RadLex anatomic terms | "liver, hepatic segment VI" |
| Pathology | RadLex pathology terms | "hepatocellular carcinoma" |
| Category | Educational category | "Can't-miss case," "Classic presentation" |
| Difficulty | Target trainee level | "PGY-2," "Fellow," "Board review" |
| Confirmation | Method of diagnostic confirmation | "Pathology-confirmed," "Surgical," "Clinical-imaging correlation" |
Checkpoint B: Post-Draft Alignment (Mandatory)
- Is the diagnosis confirmed by pathology, surgery, or strong clinical-imaging correlation?
- Is the case fully de-identified per HIPAA (all 18 identifiers removed)?
- Are images annotated with clear, consistent labels and arrows?
- Does the discussion provide educational value with differential diagnosis and teaching points?
- Is the case metadata complete for searchability?
Quality Audit
- Case meets selection criteria (diagnostic certainty, image quality, educational value)
- All 18 HIPAA identifiers are removed from images and metadata
- Burned-in pixel PHI is removed or obscured
- 3D facial reconstructions are defaced or excluded
- Key images clearly demonstrate the primary finding
- Images include appropriate window/level settings
- Annotations are consistent, clear, and correctly placed
- Clinical history is included (de-identified)
- Pathology or surgical confirmation is documented
- Differential diagnosis is discussed with distinguishing features
- Teaching points include 3–5 actionable takeaways
- Metadata is complete (title, keywords, anatomy, pathology, modality)
- Case is categorized within the teaching-file taxonomy
- Institutional IRB/compliance approval is obtained if required
Guidelines
- Prioritize pathology-confirmed cases — a beautiful teaching case without diagnostic confirmation is unreliable for education.
- De-identify thoroughly before any sharing — a single residual identifier is a HIPAA violation regardless of the educational value of the case.
- Write teaching points as actionable clinical rules, not abstract facts — "Always check the posterior malleolus on ankle radiographs" is more educational than "Posterior malleolus fractures can occur."
- Use RadLex terminology for keywords and metadata to ensure interoperability across teaching-file systems and maximize searchability.
- Include the differential diagnosis in every teaching file — board exams and clinical practice require distinguishing entities from their mimics.
- Build the teaching-file collection systematically to cover the ABR Core Exam curriculum — identify gaps in the collection and actively seek cases to fill them.
- Review and update teaching files periodically — outdated classification systems or management recommendations reduce educational value.
- Consider multi-case series for complex topics (e.g., LI-RADS scoring across all categories, Bosniak classification with examples of each class) rather than isolated single cases.