radiology-radiomics

Hand-crafted Radiomics Study Design

Use this skill to build (or audit) a hand-crafted radiomics study that is reproducible and leakage-free, from preprocessing through validation. Radiomics papers are desk-rejected for the same recurring reasons: non-standardised features, segmentation not characterised, and data leakage in selection/normalisation. This skill encodes the IBSI/CLEAR pipeline and the partition hygiene reviewers enforce.

Core stance

• IBSI defines; evidence verifies. Use IBSI to define and report image processing and feature calculations (resampling, discretisation, filters, aggregation, software+version). Conformance improves semantic and implementation comparability, but does not by itself establish empirical repeatability across segmentations, scans, software builds, scanners or protocols. Pair it with implementation checks and task-matched phantom/test-retest/sensitivity evidence.
• Discretisation is a decision, not a default. Fixed bin width vs fixed bin count changes every texture feature; state which, the value, and why; keep it consistent.
• Segmentation error propagates. Use reproducible masks and filter unstable features (ICC) before modelling (→ radiology-annotation).
• Selection lives inside training only. Feature selection, normalisation, imputation, and harmonisation are fit on training folds, never on the whole cohort — the classic leak.
• Match effective complexity to information. Thousands of features versus few independent patients/events is unstable, but no universal EPV threshold grants adequacy. Let radiology-stats define a project-specific information/precision brief from the estimand, event/non-event support, effective degrees of freedom, shrinkage, calibration and resampling stability.
• Report calibration + utility, not just AUC, for a clinical signature (→ radiology-stats).
• Integrity. Never invent feature counts, ICCs, or performance; mark what must be computed.

When to use

• "Design / review my radiomics pipeline (PyRadiomics, IBSI)." / "影像组学流程设计或审查。"
• "Bin width or bin count? what resampling/normalisation/filters?"
• "How do I select features without leakage?" / "LASSO/mRMR feature selection 怎么做才不泄漏？"
• "Build a radiomics signature/score and validate it."
• "Audit this radiomics Methods for leakage and IBSI compliance."

When to open extra files

File	Open when
references/preprocessing-ibsi.md	Resampling, intensity normalisation, gray-level discretisation/bin width, filters, IBSI reporting
references/feature-extraction.md	Feature families, PyRadiomics settings, aggregation, software/version, parameter file, delta/longitudinal radiomics, test–retest/phantom repeatability
references/selection-modelling.md	Leakage-safe selection (variance/ICC/correlation/LASSO/mRMR), modelling, signature/score, EPV
references/leakage-audit.md	The radiomics-specific leakage checklist reviewers weaponise
references/reproducibility-artifacts.md	Producing a final analysis package, feature dictionary, run/split manifest, signature equation, or auditable feature-count trail

Workflow

1. Confirm the design (reuse radiology-design) — endpoint, independent unit/hierarchy, cohorts, validation type and project-specific information/precision constraints.
2. Segmentation & stability — masks and reproducibility from radiology-annotation, which owns the ICC stability metric and threshold provenance; this skill freezes and applies that filter inside training.
3. Preprocessing (preprocessing-ibsi.md) — resample, normalise, discretise (state bin width/count), filters; record everything for IBSI.
4. Extraction (feature-extraction.md) — feature families, PyRadiomics (or equivalent) + version, parameter file; produce a documented, versioned feature matrix.
5. Selection & modelling (selection-modelling.md) — selection inside CV/training only; model complexity matched to the statistical brief; build the signature/score; pre-specify the primary analysis.
6. Validate — internal (nested CV/bootstrap) + external/temporal/geographic; report discrimination, calibration, DCA (→ radiology-stats).
7. Freeze reproducibility artifacts (reproducibility-artifacts.md) — split/run manifest, parameter file, feature dictionary/count trail, fitted pipeline, coefficients/signature equation, environment, and deviations.
8. Audit leakage (leakage-audit.md) and write Methods to CLEAR/IBSI.

Output contract

1. Pipeline spec — preprocessing → extraction → selection → model → validation, each step with its parameters and the leakage control marked.
2. Parameters — resampling, normalisation, discretisation (bin width/count), filters, feature families, software+version (a PyRadiomics parameter file where applicable).
3. Selection/modelling plan — method, where it sits relative to the split, effective complexity and the project-specific information/precision check.
4. Validation plan — internal + external; metrics incl. calibration/DCA.
5. Leakage audit — pass/fail per item with the fix.
6. Reproducibility bundle — run/split manifest, parameter file, feature dictionary and count trail, fitted preprocessing/selection/model objects, coefficient/signature mapping, environment, and deviations actually available.
7. Table handoff — feature-stage counts, selected-feature dictionary, coefficients/weights, and model-performance value keys for radiology-table.
8. Methods paragraph — CLEAR/IBSI-aligned prose (+ 待确认 for Chinese authors).

Quality bar

A good radiomics spec lets another lab identify and execute the same feature definition and then test whether the implementation and measurements agree within declared tolerances. It quantifies segmentation/acquisition sensitivity, keeps selection inside the split, and reports performance with calibration and CIs, never AUC alone.

Handoffs

• Series/phase/sequence qualification, acquisition/reconstruction, quantitative transforms, artifacts, dose, phantom/test-retest and protocol drift → radiology-acquisition-qc; consume its frozen measurement passport before radiomics preprocessing.
• Mask SOP & feature-stability → radiology-annotation.
• IBSI/CLEAR/METRICS/RQS audit → radiology-reporting.
• Selection/CV statistics, calibration, DCA, multiplicity, sample size → radiology-stats.
• Parameter provenance, bin-width/spacing/interpolation/segmentation sensitivity, fair method comparison, ablation and claim-level robustness → radiology-method-evaluation; this skill still owns radiomics configuration and execution.
• Deep features / deep-learning comparison → radiology-deep-learning.
• Biological interpretation of the signature → radiology-radiogenomics.
• Figures (feature heatmap, ROC, calibration, nomogram) → radiology-figure.
• Tables (feature audit, coefficient/signature, performance, subgroup/sensitivity) → radiology-table.
• Reframing this pipeline as a funding proposal → radiology-grant.
• Full-project state, simulation labels, and cross-artifact consistency → radiology-pipeline.