跳到正文
FunCoding

radiology-innovation-transfer

Assess disclosure, prior-art handoff, rights and transfer readiness; not patentability, FTO or legal advice.

Star
1,942
最近提交
2026年9月21日

radiology-skills/skills/radiology-innovation-transfer/SKILL.md

在 GitHub 上查看

Radiology Innovation Transfer

Use this skill to organize research assets for responsible transfer from an imaging laboratory to a technology-transfer, development or partnership decision. It owns invention-disclosure preparation, patent-landscape research framing, readiness evidence, value hypotheses, rights/dependency mapping, non-confidential summaries and diligence handoffs. It does not give patentability, freedom-to-operate, ownership, inventorship, regulatory, quality-system, contracting or investment advice.

Non-negotiable boundaries

  • Protect before publicizing. Record prior/public disclosure dates, planned abstracts/papers, presentations, repositories, demonstrations and partner conversations. If potentially protectable confidential information has not been reviewed by the institutional TTO/IP owner, return STOP_PUBLIC_DISCLOSURE; prepare only a non-public disclosure packet.
  • Landscape is not a legal opinion. A patent landscape describes search scope, patent families, classifications, assignees, claims/topics and trends. It is not a patentability, validity, inventorship, ownership or freedom-to-operate (FTO) opinion. Preliminary searches are incomplete.
  • Maturity axes remain separate. TRL, MRL, scientific validity, clinical validity/validation, clinical utility, regulatory status, quality-system readiness, interoperability and commercial adoption are not interchangeable. Use evidence-backed exit criteria for each axis.
  • Market interest is not adoption. Interviews, letters, clicks, citations, pilots and investor interest are different signals. None alone proves willingness to pay, procurement, reimbursement, sustained use or patient benefit.
  • Rights before value claims. Map employment/sponsor obligations, foreground/background IP, software and model licenses, training/validation data rights, annotation rights, consent/governance, open-source obligations, vendor dependencies and collaboration agreements. Do not infer ownership.
  • Current facts are live. Patent status, legal rules, assignees, licensing terms, regulatory classification, reimbursement, competitors, prices and market size require LIVE_VERIFICATION_REQUIRED with jurisdiction, source and access date.
  • No confidential fabrication. Never invent dates, contributors, disclosures, performance, customers, costs, regulatory status, patent families or market evidence.

Input passport

Use innovation-disclosure-passport.md to record:

  • asset/invention version, concise problem/solution and intended use;
  • contributors, affiliations, employment, funding, sponsor and collaboration context;
  • conception/development evidence, dated records and public/planned disclosures;
  • code, algorithms, model weights, data, labels/masks, documentation, hardware and know-how;
  • scientific/technical evidence, comparator, failure modes and unresolved reproduction;
  • TRL/MRL and separate clinical, regulatory, quality, interoperability and deployment states;
  • third-party/open-source/data/model licenses and access/consent/governance constraints;
  • prior-art/landscape objective, concepts, synonyms, CPC/IPC classes, jurisdictions and dates;
  • target users, workflow, payer/procurer, alternatives, benefits, burdens and adoption assumptions;
  • confidentiality tier, authorized recipients and institutional TTO/legal contacts.

Do not send confidential material to an external search, model, partner or public tool unless the user has authority and the channel is approved for that material.

Route by mode

NeedModeMain output
Prepare an institutional invention disclosuredisclosure-triageevidence-backed disclosure passport, contributor/dependency questions and TTO handoff
Map prior art and patent activitypatent-landscapereproducible search protocol, family-level landscape and legal-opinion boundary
Assess technical/manufacturing maturityreadiness-assessmentevidence/exit-criteria matrix across TRL, MRL and separate medical-product axes
Test clinical/customer value assumptionsvalue-propositionuser/workflow/problem/alternative/benefit hypothesis and evidence gaps
Map IP, data, model, partnership or license constraintsrights-and-partnershipdependency/rights matrix and questions for authorized owners
Prepare safe external-facing materialnonconfidential-summaryreviewed non-confidential problem/solution/evidence/status summary
Prepare a diligence or transfer decisiondiligence-handoffclaim-source evidence map, red flags, open questions and TTO/regulatory/quality handoffs

Read transfer-modes-and-evidence-gates.md for readiness/value decisions and patent-readiness-and-diligence.md for patent and diligence controls. Consult source-registry.md and recheck every dynamic source.

STOP gates

  • STOP_PUBLIC_DISCLOSURE: protectable/confidential material could be released before TTO/IP review;
  • STOP_CONFIDENTIAL_CHANNEL: proposed tool, repository, email or partner channel is not approved;
  • STOP_RIGHTS: material data/code/model/contract rights or sponsor obligations are unresolved;
  • STOP_IDENTITY: contributor, inventorship or ownership facts are disputed or missing;
  • STOP_EVIDENCE: claimed performance/readiness/value lacks a source artifact and exit criterion;
  • STOP_LEGAL_OPINION: user requests patentability, validity, FTO, ownership or contract advice;
  • STOP_REGULATED_CLAIM: medical-device, clinical, quality or marketing claims lack qualified review;
  • STOP_MARKET_FACT: current competitor, price, market-size, reimbursement or deal fact is unverified;
  • STOP_PHI: patient/image data, DICOM headers, burned-in text or linked artifacts may expose PHI.

STOP blocks the unsafe release or conclusion, not a bounded internal gap analysis.

Radiology-specific diligence

  • Define intended input/output, modality, anatomy, acquisition/reconstruction, DICOM objects, PACS/RIS/worklist integration, latency, failure/fallback and human oversight.
  • Separate algorithm/model IP from training data, labels/masks, pretrained weights, third-party code, scanner/vendor components, deployment stack and tacit clinical/engineering know-how.
  • Trace dataset permissions across centers, consent/waiver, data-use agreements, de-identification, secondary use, commercial use, cross-border transfer and model/derived-artifact clauses.
  • Preserve measurement validity: protocol/scanner shift, QIBA/phantom/test-retest evidence, reference-standard quality and external/local validation remain scientific gates.
  • Distinguish a research prototype from a maintainable product: cybersecurity, logging, versioning, monitoring, update/change control, service support and rollback require owners and evidence.
  • A radiology AI model with high AUC or a prospective pilot is not automatically patentable, regulatory-authorized, reimbursable, interoperable, manufacturable or adoptable.

Workflow and output contract

  1. Freeze the passport, confidentiality tier and decision being supported.
  2. Build a claim-source ledger inside diligence-evidence-map.md: claim, exact artifact, status, owner, date, restrictions and unresolved verification.
  3. Choose the mode. Search patent literature at family/claim/classification level where authorized; include non-patent literature and record query/date/database/coverage.
  4. Score no maturity level without evidence and explicit exit criteria. Record the chosen framework and any medical-imaging adaptation.
  5. Separate evidence from hypotheses in the value proposition. Test workflow burden, alternative, decision maker, buyer/payer, resource effect and adoption barrier.
  6. Run confidentiality, IP/rights, scientific, clinical, regulatory/quality, interoperability, operational and market gates independently.
  7. For external material use nonconfidential-summary.md and obtain the authorized TTO/legal/communications release decision.

Return Decision and scope, Disclosure passport, Evidence/claim map, Landscape or readiness method, Rights/dependency matrix, Value hypotheses, Red flags and STOP ledger, Qualified-owner handoffs, and Decision status.

Decision status is INTERNAL_TRIAGE, READY_FOR_TTO_REVIEW, READY_FOR_QUALIFIED_DILIGENCE, READY_FOR_AUTHORIZED_NONCONFIDENTIAL_USE, or STOP—never “patentable”, “FTO clear” or “investment ready”.

Handoffs and non-ownership

  • Patentability, FTO, validity, inventorship, ownership, filing and contracts → institutional TTO and qualified patent/legal counsel.
  • Intended use, clinical validation/utility, human factors and deployment → radiology-translation; formal regulatory and QMS decisions → qualified regulatory/quality owners.
  • Imaging acquisition/measurement → radiology-acquisition-qc; disease pathway → radiology-clinical-domain; data/consent/governance → radiology-data / radiology-ethics.
  • Underlying study design/statistics → radiology-design / radiology-stats; grant proposal → radiology-grant; manuscript → radiology-writing.
  • Public adaptation of an authorized non-confidential source → radiology-dissemination; this skill prepares the source but does not publish or market it.

相似的 Skill

blender-motion-state-inspection

AFaffaan-m/ECC278k近 7 天新增 4,046 个 star

官方

slack-gif-creator

ANanthropics/skills181k近 7 天新增 1,119 个 star

精选

algorithmic-art

用 p5.js 生成流场和粒子系统,通过种子随机性创作原创算法艺术。

ANanthropics/skills181k近 7 天新增 1,119 个 star

thinking-out-loud

SHShubhamsaboo/awesome-llm-apps141k近 7 天新增 808 个 star

moneyprinterturbo-video

HAharry0703/MoneyPrinterTurbo130k近 7 天新增 1,722 个 star

complain

PApaperclipai/paperclip101k近 7 天新增 2,948 个 star