radiology-skills

Design and document ROI/VOI/mask annotation that survives Radiology (RSNA) review — lesion-selection strategy (2D vs 3D, whole-tumour vs largest-slice vs peritumoral vs habitat vs multi-lesion), reader protocol (number, seniority, blinding, independent vs consensus, third-party adjudication), reproducibility (repeat annotation, ICC, Dice, Hausdorff, feature-stability filtering), and geometric integrity of masks against DICOM/NIfTI/DICOM-SEG/RTSTRUCT (spacing, origin, direction, slice order). Use when the user mentions ROI, VOI, mask, segmentation protocol, contouring, "标注", "勾画", "分割规范", inter-/intra-observer agreement, or needs the Methods paragraph for annotation. Produces an annotation SOP, a QC plan, and submission-ready Methods text. Never invents agreement values or reader details.

Turn manuscript text, claims, figure/table statements, abstracts, slides, or novelty/comparison assertions into verified, imaging-journal-scoped citation candidates and export one reference-manager-ready file (RIS, EndNote ENW, or BibTeX). Use when the user needs references for an imaging paper, wants supporting citations, wants a two-pass claim/citation/numerical verification gate, needs to check whether a cited source actually says the claimed thing, scope citations to radiology/imaging journals (Radiology, Radiology: AI, RadioGraphics, AJR, European Radiology, JACR, etc.), verify a DOI/PMID, or export a bibliography. Verifies identifiers before formatting and never fabricates DOIs, pages, volumes, journal metadata, or source support.

Prepare and audit Data/Code Availability statements, DICOM de-identification plans, repository selection, dataset citations, and FAIR/sharing checks for Radiology (RSNA) and Nature-portfolio imaging+omics submissions. Use when the user needs a data availability statement, must de-identify DICOM imaging, choose a repository (TCIA, Zenodo, GEO, dbGaP, EGA), share code/models, write dataset citations, check FAIR compliance, or needs Extended Data / Supplementary Information / Source Data planning for a Nature-family journal — including controlled-access genomics for radiogenomics. Bilingual-aware (中文作者备注 → submission-ready English). Never overstates availability or invents accessions.

Design and audit imaging deep-learning studies to Radiology (RSNA) / CLAIM 2024 standard, or to Nature-portfolio / FUTURE-AI trustworthy-AI standard — architecture choice (2D/2.5D/3D CNN, Transformer/ViT, segmentation/detection nets, prognostic models), transfer learning vs self-supervised pretraining vs training from scratch, how images/masks/clinical/text/molecular inputs enter the model, data splitting and augmentation, class imbalance, hyperparameter search, baselines, external validation, interpretability/explainability (Grad-CAM, SHAP, attention), uncertainty quantification (MC dropout, ensembles, conformal prediction), and robustness/OOD testing — with patient-level partition hygiene throughout. Use when the user plans or reviews a CNN/Transformer/3D/segmentation/detection/foundation/multimodal imaging model, mentions transfer learning, self-supervised, nnU-Net, ViT, data augmentation, class imbalance, explainability, uncertainty, robustness, or "影像深度学习/深度学习模型". Produces a model+training+validation des

Assess whether an imaging dataset can support a study and turn it into a complete, submittable design — from feasibility triage to clinical question, target population, endpoint/estimand, minimum-viable vs stronger methods, and a validation strategy (internal resampling, temporal, geographic, fully external, multi-center, federated). Use when the user has CT/MRI/PET/US/mammography/multimodal data but is unsure what to do, asks "can this topic be done?" / "能不能做" / "帮我设计课题" / "study design" / "what can I study with this data", or needs a multi-center / external-validation plan ("多中心", "external validation", "generalisability", "center effect", "scanner effect"). Produces a study blueprint with feasibility verdict, design options, validation plan, and the limiting constraint surfaced. Never invents cohort numbers or overstates what the data can support.

Write compliant, non-overstated Ethics, Informed-Consent, and privacy/governance text for imaging and imaging+omics studies — study-type framing (retrospective, prospective, registered, multi-center), IRB/ethics-committee approval and numbers, consent vs documented waiver, re-identification risk (DICOM metadata, rare-disease small cohorts, genomic data), and the human-subjects governance behind data sharing (HIPAA/GDPR/PIPL, data-use agreements, controlled access). Use when the user mentions ethics, IRB, ethics committee, informed consent, consent waiver, "伦理", "知情同意", "隐私", HIPAA, GDPR, PIPL, data-use agreement, or needs the Ethics/Consent statement for a manuscript or grant. Produces submission-ready statements and a re-identification risk read. Never fabricates approval numbers, dates, or consent status.

Produce publication-quality white-background academic figures for Radiology (RSNA), Nature-portfolio/npj, European Radiology, NEJM, Science, or Lancet-family venues with Python (matplotlib): ROC curves, calibration plots, decision-curve analysis, forest/SROC plots, Kaplan-Meier curves with numbers-at-risk, Bland-Altman, heatmaps, radiogenomics plots, graphical/visual abstracts, and annotated imaging panels. Uses The Lancet Digital Health guide as the default Lancet-series proxy. Use when the user wants figures, plot cleanup, figure-set planning, journal-specific figure formatting, or overlap/crowding QA. Outputs editable vector (.svg/.pdf) plus 300+ dpi raster, enforces de-identification and journal typography, and never invents data points.

Find publishable frontier directions and innovation points for imaging-AI / radiomics / radiogenomics research, grounded in the publication patterns of high-impact journals (Radiology, Radiology: AI, Lancet Digital Health, Lancet Oncology, Nature Medicine, Nature Communications, npj Digital Medicine, npj Precision Oncology, eClinicalMedicine, Cell Reports Medicine). Use when the user asks for frontier directions, innovation points, hot vs suitable topics, "近三年前沿方向", "创新点", "what's novel in imaging AI", or wants to know the evidence/publication-pattern basis behind a recommendation ("有什么文献依据", "证据"). Translates trends (foundation models, self-supervised, vision-language, multimodal fusion, longitudinal, weak/federated learning, radiogenomics) into executable research questions matched to the user's actual data. Encodes publication-pattern heuristics, not a fabricated citation list — routes live verification to radiology-search and never invents PMIDs/DOIs.

Reframe and polish imaging-AI / radiomics / radiogenomics research as a research grant proposal — convert paper-style "we built a model" into grant logic (clinical need → scientific question → hypothesis → specific aims/研究内容 → technical route/技术路线 → innovation/创新点 → feasibility/可行性 → expected outcomes), and strengthen title, abstract, background/立项依据, aims, key scientific question/关键科学问题, and significance. Primary and most-developed track is NSFC (国家自然科学基金: 青年/面上/地区) and provincial funds (省自然); also covers reframing the same proposal for international funders (NIH R01, ERC Starting/Consolidator/Advanced, Wellcome Trust) including realistic eligibility for China-based applicants and cross-border routes (NSFC international collaboration line, RGC, MSCA, foundations). Use when the user mentions 国自然/省自然/基金申请/标书/立项依据/科学问题/技术路线/创新点/可行性, NIH/R01/ERC/Wellcome/国际基金, or wants to turn a study into a fundable proposal for any of these. Flags weak innovation, thin preliminary data, open-loop technical routes, over-promisi

Match a finished or near-finished imaging-AI / radiomics / radiogenomics manuscript to the right target journals and build a submission tier list (reach / target / safety) grounded in each venue's publication patterns, author-guide style profiles, and the paper's real strengths and weaknesses. Use when the user asks where to submit, "选刊/投哪个期刊", journal selection, "can this go to Nature Medicine / Science / NEJM / Lancet / Lancet Oncology / Lancet Digital Health / Radiology", fit assessment, or a submission ladder. Uses The Lancet Digital Health guide as the default Lancet-series proxy. Grades external validation, prospectivity, reader study, calibration, clinical utility, sample size, centers, novelty, and reporting compliance; returns fit, risk, strengthening priorities, and venue-style requirements. Verifies current journal scope via live search; never selects on impact factor alone.

Turn an imaging-research paper, preprint, PDF, abstract, or reading notes into a concise Chinese .pptx deck for journal club / 读片会 / 组会 / paper sharing. Use when the user wants a paper PPT, journal-club slides, 读片会/文献汇报 slides, or paper-to-slides for an imaging study. Identifies the paper type and evidence chain, selects only the figures/tables that carry the argument (imaging panels, ROC/KM/calibration), writes Chinese slide text + speaker notes, builds a real .pptx, and runs package QA. Never fabricates results, metrics, or figure details.

Polish finished imaging-research prose to Radiology (RSNA), Nature-portfolio/npj, European Radiology, NEJM, Science, or Lancet-family house style: concise sentences, correct tense, precise statistical reporting, venue-specific p-value/number/reference style, AI/data-sharing declarations, abbreviation control, and overclaim/causal-language detection. Uses The Lancet Digital Health guide as the default Lancet-series proxy. Use when the user wants to tighten, copyedit, or align existing paragraphs to a target journal style, not draft new content. Returns clean copy-paste prose plus a short change log; never alters reported numbers or invents content.

Run a rigorous pre-submission mock peer review of an imaging-AI / radiomics / radiogenomics manuscript — simulate the methods, statistics, reporting-guideline, figure, citation/claim-verification, and data-sharing reviewer a top journal would assign, and surface the issues that cause desk-reject or major revision before submission. Use when the user wants a mock review, pre-submission audit, "投稿前预审/模拟审稿", "find the holes before a reviewer does", a two-pass abstract/figure/table claim audit, or a readiness check. Returns a reviewer-style report with Blocker / Major / Minor issues, each tied to the manuscript location and the reporting-guideline or methodological risk, plus an editor-style recommendation and a prioritised fix order. Never fabricates compliance or papers over a real weakness.

Design, analyse, report, and submit imaging-multi-omics radiogenomics studies that link radiomic/deep imaging phenotypes to genomic, transcriptomic, single-cell, and spatial-omics data. Use when the user mentions radiogenomics, imaging genomics, imaging-transcriptomics, TCIA/TCGA, GEO, dbGaP, EGA, cBioPortal, multi-omics integration, MOFA, iCluster, SNF, DIABLO, scRNA-seq deconvolution, CIBERSORTx, BayesPrism, spatial transcriptomics, Visium, Xenium, imaging habitats, gene expression, mutations, pathways, biological validation, omics QC, sample-to-image mapping, radiogenomics submission packages, or reviewer comments about batch/leakage/spatial mismatch. Covers the full chain from matched-cohort feasibility and protocol/SAP to omics QC, imaging pipeline, integration, validation, reporting, submission, and rebuttal support. Never fabricates associations, cohort counts, accessions, approvals, metrics, or reviewer actions.

Design and audit a hand-crafted radiomics study end-to-end to Radiology (RSNA) / CLEAR / IBSI standard — image preprocessing (resampling, intensity normalisation, gray-level discretisation/bin width, filters), IBSI-compliant feature extraction (PyRadiomics or equivalent), reproducibility/stability filtering, leakage-safe feature selection, modelling, and internal/external validation. Use when the user plans or reviews a radiomics pipeline, mentions PyRadiomics, IBSI, feature extraction, bin width, gray-level discretisation, LASSO feature selection, radiomics signature/score, or "影像组学/放射组学". Produces a reproducible pipeline spec, runnable parameter settings, a leakage audit, and Methods text. Never fabricates feature counts or performance.

Build full-paper Chinese-English side-by-side (中英对照), figure/table-aware, source-grounded Markdown readers for imaging-research papers from PDF, DOI, arXiv, publisher HTML, or pasted text. Use whenever the user asks to translate or read an imaging paper, make 中英文对照/原文对照/全文翻译解读, extract figures/tables (including imaging panels and result charts) into the right positions, or keep exact source anchors. Especially attentive to radiology specifics: imaging figures, ROC/forest/KM charts, radiomics/AI methods detail, scanner/protocol tables, and statistical reporting. Must not degrade into a summary-only output unless the user explicitly asks for a summary.

Route an imaging-research manuscript or protocol to the correct reporting/quality guideline and audit it item-by-item for Radiology (RSNA) or Nature-portfolio submission. Use when the user mentions CLAIM, TRIPOD+AI, STARD, PRISMA-DTA, QUADAS-2, CLEAR, METRICS, RQS, IBSI, PROBAST, CONSORT-AI, FUTURE-AI, TRIPOD-LLM, the Nature Portfolio Reporting Summary / Editorial Policy Checklist, a "reporting checklist", "what's required for submission", radiomics quality, or wants to know what a reviewer will check. Produces a filled checklist with PRESENT / PARTIAL / MISSING per item, manuscript location, and concrete fixes. Do not fabricate compliance — flag missing items honestly.

Draft, audit, and revise point-by-point reviewer response letters for Radiology (RSNA) and imaging-journal revisions. Use when the user has reviewer comments / a major or minor revision / 审稿意见 to answer for an imaging-AI/radiomics/radiogenomics manuscript. Assigns each comment a stable ID, classifies it, maps it to a concrete manuscript action, and ties every claimed change to a specific location — without fabricating experiments, analyses, citations, line numbers, or results. Bilingual-aware (中文作者备注 → English response + Chinese confirmation items).

Multi-source literature, dataset, and local-corpus search for imaging research, with deduplication, identifier verification, reference management, and survey/RAG workflow support. Searches PubMed, arXiv, Crossref, and imaging dataset registries (TCIA, GEO, cBioPortal, OpenNeuro, Grand Challenge, Medical Segmentation Decathlon), and can structure local paper/RAG workflows when a paper database, PaperQA2, NotebookLM, or paperpipe-style backend exists. Use when the user wants to find papers/datasets, run systematic or survey-level search, verify DOI/PMID/arXiv IDs, map a field, build a gap matrix, or synthesize literature for a manuscript. Verifies identifiers and exposes incomplete metadata; never fabricates bibliographic fields.

Plan, run, and report imaging biostatistics the way Radiology (RSNA) reviewers expect — diagnostic accuracy with CIs, ROC/AUC comparison (DeLong/bootstrap), reader agreement (Cohen/Fleiss kappa, ICC, Bland-Altman), multi-reader multi-case (MRMC) studies, calibration and decision-curve analysis, multiplicity control for high-dimensional radiomics/omics (Bonferroni vs FDR), survival/prognostic modelling, and sample-size/EPV planning. Use when the user mentions AUC, DeLong, sensitivity/specificity, kappa, ICC, MRMC, calibration, decision curve, p-value, FDR, multiple comparisons, sample size, C-index, or asks how to report a statistic for Radiology. Provides runnable Python/R and a results sentence; never fabricates numbers.

Move a validated imaging model from retrospective performance toward clinical use — design the clinical-use scenario, model outputs, reader (multi-reader multi-case) studies of radiologist-vs-AI-vs-radiologist+AI, decision thresholds tied to actions, net-benefit/decision-curve framing, prospective and real-world validation, and PACS/RIS workflow integration. Use when the user wants a reader study, clinician+AI gain, prospective/real-world validation, deployment plan, "临床转化/读者研究/前瞻性验证", or to know what it takes to claim clinical utility. Produces a translation plan with the use scenario, reader-study design, threshold-to-action mapping, and prospective/workflow design. Never lets a retrospective AUC be reported as clinical readiness.

Draft or rebuild imaging-AI / radiomics / radiogenomics manuscript sections for Radiology (RSNA), Nature-portfolio/npj, European Radiology, NEJM, Science, or Lancet-family style from author-provided results, figures, notes, or Chinese drafts. Enforces the target venue's manuscript shape: Radiology Summary/Key Results, Nature-style broad scientific narrative, European Radiology key points/clinical relevance, NEJM four-part clinical abstract and SAP rigor, Science compact display-item story, or Lancet-series Research in context and AI/data transparency using The Lancet Digital Health guide as the default proxy. Use when the user wants to write or restructure imaging-research prose, not just polish it. Never invents data, metrics, or citations.

Use when planning, auditing, writing, or revising radiomics, medical imaging AI, and radiology deep learning studies for Radiology/RSNA, Nature-portfolio, Lancet, Cell, npj, European Radiology, or similar venues. Trigger for research frontiers, literature, CT/MRI/PET/ultrasound datasets, ROI/masks/segmentation annotation, radiomics features, CNN/Transformer/foundation models, trustworthy AI, FUTURE-AI, uncertainty/OOD/interpretability, radiogenomics and multi-omics mechanisms, statistics, figures, pre-submission review, reproducibility, multicenter validation, public datasets, ethics/privacy, clinical translation, validation/leakage, CLAIM/CLEAR/RQS/IBSI/TRIPOD+AI/PROBAST+AI/STARD-AI, manuscript writing, journal selection, NSFC/provincial/international grants, reviewer response, and Chinese-English support.