RadVector Reporting
Clinical intelligence, directly in radiology workflow.
RadVector Reporting combines report construction, clinical inference, QA, evidence and model orchestration while keeping the radiologist in control. It is the complete radiologist-facing implementation of the iTech intelligence stack.
The pipeline
The radiologist at both ends.
Radiologist input
Dictate, type, paste from any reporting system, or drop images and worksheets.
Study and template
Exam and CPT detected; the matching CPT-coded template selected.
Findings
The report is built through FINDINGS from the radiologist's own words.
Clinical inference
Structured classifications and criteria applied to what was observed.
Impression and referencesthe model
Impression, recommendations and references drafted around the findings.
QA
Deterministic and AI checks, optionally by a different model than the one that drafted.
Radiologist approval
The radiologist reviews, edits and signs. Always.
What it demonstrates
Proof that the intelligence layer works in real workflow.
Findings authority
The interpretive half of the report cannot rewrite what the radiologist observed. This is enforced mechanically, not by prompt.
Decision support
Fleischner, BI-RADS and DEXA from validated criteria packs; Lung-RADS, TI-RADS, LI-RADS, O-RADS, PI-RADS and Bosniak from draft packs that answer as advisory. See the reporting standards guides.
Quality assurance
Laterality, anatomy-concordance, date and contrast checks, plus a validation queue for flagged content.
References
Claims checked against a registry of published sources, with the guidance applied cited in the report.
Governance
A per-report record of every standard applied and the pack version behind it, with CSV export; a workflow value panel from measured data; isolated groups, permissions and peer review.
Model orchestration
Many vendors with fallback, and a per-report record of which model ran, its tokens and cost.
Reporting standards guides: Lung-RADS, BI-RADS, TI-RADS, LI-RADS, O-RADS, PI-RADS, Bosniak, Fleischner, DEXA. RadCloudAI is now RadVector. Same platform. New name. TumorBoard, an oncology case-preparation surface, is part of the same platform.
Quality measures
MIPS-Aware Radiology Reporting
RadVector Reporting supports clinical documentation relevant to selected MIPS radiology quality measures. It is guideline-informed reporting with integrated clinical QA: documentation relevant to those measures is drafted into the report and checked before the radiologist signs.
Live: In production in a shipping product today.Representative clinical areas
Incidental pulmonary nodules
Fleischner 2017 follow-up applied from the radiologist's stated size and type, from a validated criteria pack, with the recommendation written into the report and limited to whole-chest studies.
Coronary artery calcium scoring
Per-artery Agatston, mass and volume scores carried from the scoring workstation printout, one labeled line per artery. Numbers are written only when they add up to the printed totals; otherwise the line is marked for the radiologist.
Incidental coronary calcification
Chest CT templates carry a coronary artery calcification line, documented as present or not significant and never graded, with the management sentence the template defines.
Thyroid nodules on ultrasound
ACR TI-RADS points and category from the described features, from a draft pack that answers as advisory. The radiologist confirms the category.
Incidental thyroid nodules on CT, MRI and PET
A thyroid nodule described on CT, MRI or PET brings ACR incidental thyroid nodule guidance into report drafting where the practice has clinical guidance enabled, so the recommendation is drafted from the guidance and confirmed by the radiologist.
FDG PET/CT
PET/CT templates carry structured SUVmax fields, including mediastinal and liver reference values, and current and prior SUVmax are compared when they are provided. For patients with a history of lymphoma, the Deauville score is carried exactly as the radiologist dictates it; RadVector never assigns one, and QA flags a lymphoma PET with no dictated score.
How the documentation is produced
- Versioned criteria
- Criteria decisions come from versioned packs that return the same answer for the same input, with the pack and version recorded for each report.
- Deterministic report logic
- Template structure, worksheet values and required report lines are filled by rules, not by a language model.
- Constrained language model
- A language model drafts the impression from the findings; deterministic checks flag any change to what the radiologist observed.
- Integrated clinical QA
- Deterministic and AI checks flag missing or inconsistent documentation before the radiologist signs.
MIPS-aware reporting concerns what is documented in the report. RadVector does not submit quality measures to CMS, is not a CMS-qualified registry, and does not guarantee measure performance, scores or reimbursement. Measure selection and submission remain with the practice and its reporting vendor. The radiologist reviews and signs every report.
RadVector Reporting questions
What is the RadVector pre-read service?
POST /v1/context/prepare assembles what a radiologist needs before reading: from the current order and prior report texts it returns the comparable priors ranked same-type first with their intervals, prior measurements bound to their lesions, the reporting standards that apply with their required elements, recommendations from earlier reports that are still open, and missing-information warnings. It calls no language model.
Does RadVector Reporting keep an audit trail of the standards applied to each report?
Yes. RadVector Reporting writes a governance record for each report: every standard applied, the category reached, whether it was authoritative, advisory or deferred, and the criteria-pack version, validation status, validator and date behind it, plus the knowledge release used. The record holds no report text. Group administrators see it in the Governance section of Analytics and can export it as CSV, beside a workflow value panel built only from measured data.
What is RadVector Reporting?
RadVector Reporting is a radiologist-facing reporting and decision-support platform at https://radcloud.net. The radiologist dictates, types or pastes findings; RadVector builds the report through FINDINGS from the radiologist's own words, then drafts the impression, recommendations and references, runs deterministic and AI quality checks, and cites the guidance applied. The radiologist reviews, edits and signs every report. RadVector Reporting was formerly called RadCloudAI.
How is MIPS-aware radiology reporting different from MIPS submission?
MIPS-aware radiology reporting is about what the report says; MIPS submission is about sending measure data to CMS. RadVector Reporting supports clinical documentation relevant to selected MIPS radiology quality measures, for example Fleischner follow-up for incidental pulmonary nodules, ACR guidance for incidental thyroid nodules on CT, MRI and PET, per-artery coronary calcium scores and incidental coronary calcification on chest CT, and runs QA that flags missing documentation before the radiologist signs. RadVector does not submit quality measures to CMS, is not a CMS-qualified registry or certified submission vendor, and does not guarantee measure performance, MIPS scores or reimbursement. Measure selection and submission remain with the practice and its registry or reporting vendor.
Is RadVector FDA-cleared?
No. RadVector Reporting and RadVector API are clinical decision support for qualified professionals. They have not received FDA clearance or approval as a medical device, are not a diagnostic device, and must not be used for autonomous diagnosis, triage or unsupervised reporting. The radiologist remains the final clinical authority.
Which AI models does iTech use?
RadVector is model-neutral. It routes work across more than ten commercial model and speech vendors with cross-vendor fallback, and records on every report which model actually ran, with its tokens and cost. Enterprise groups can choose their vendors and models and bring their own keys. Deterministic criteria decisions never depend on a language model.
See it on a real case.
Clinical decision support for qualified professionals. Not for autonomous diagnosis or unsupervised reporting.