Radiology dictation that doesn't make you remember Bosniak at 4pm.
For every liver mass, manually recall LI-RADS criteria. For every lung nodule, Fleischner. For every renal cyst, Bosniak. For every adrenal nodule, washout calculation. Get them wrong or omit them and you've created liability and inconsistency across your practice.
GigHz Precision AI structures dictation into Technique → Findings → Impression and embeds the relevant clinical scoring system in the right place — not as a pop-up lookup, but as part of how the report is generated. LI-RADS, Lung-RADS v2022, Fleischner, Bosniak v2019, adrenal washout, TI-RADS, BI-RADS, PI-RADS, and more — applied automatically as the radiologist talks.
See it work
40-sec live demo — PE study + Fleischner nodule popup
What gets missed in a thousand reads a year
Radiologists dictate thousands of reports a year. Three failure modes that aren't about clinical knowledge — they're about cognitive load and consistency:
Numeric scoring under fatigue
LI-RADS, Bosniak, Lung-RADS — these are categorical assignments with numeric thresholds. They aren't easy to recall by 4pm on a 100-RVU day. Wrong category creates downstream care-pathway errors.
Inconsistency across the practice
Same finding reported four different ways across four radiologists. Referring physicians can't trend the patient. Quality reviews can't assess. The group looks unstandardized — because it is.
Generic AI dictation misses the clinical layer
Most 'AI radiology' tools target image analysis or worklist triage. The dictation tools don't know LI-RADS exists. Radiologists end up doing the scoring as a separate step — or skipping it.
What Precision AI embeds in the dictation flow
The system recognizes from context which scoring framework applies. The radiologist describes the finding; the classification surfaces inline.
LI-RADS — automatic
Describe a liver lesion with size, arterial enhancement, and washout — Precision AI surfaces the LI-RADS category in the impression, with rationale traceable to the criteria. Major and ancillary features mapped automatically.
Fleischner + Lung-RADS — embedded
Pulmonary nodule size, density, and growth context produce Fleischner follow-up recommendations or Lung-RADS scoring directly in the impression. Defensible documentation, no separate worksheet.
Bosniak v2019 — applied correctly
Cyst features → Bosniak category, with the right thresholds for septations, wall thickness, and enhancement. The version that referring urologists actually use, not the 2005 version that's still in some templates.
Washout — calculated
Pre-contrast, portal-venous, and delayed HU values produce absolute and relative washout percentages and the appropriate adenoma-vs-not call — built into the report instead of a calculator window.
What it looks like in the workflow
GigHz Precision AI listens to your natural dictation and detects when classification criteria apply. It surfaces the relevant popup, pre-fills what it can, and lets you confirm or adjust. Semi-automatic — the AI proposes, you verify.
Why Precision AI is different from your current dictation tool
Most AI radiology tools work on the image. Precision AI works on the report — the part of radiology that touches every read, every day.
Embedded, not bolted-on
Scoring surfaces inline, in the impression — not as a pop-up calculator the radiologist closes without reading. The workflow stays speech-first.
Multiple frameworks in one report
A chest CT can hit Lung-RADS, Fleischner, and adrenal washout in the same read. Precision AI handles all four scoring layers in a single dictation pass.
Defensible documentation
Every category assignment is traceable to the criteria. Quality reviews and medico-legal audits get a structured trail instead of a free-text impression.
Who's using it
Diagnostic + IR — full-volume practices
Body imaging, abdominal, thoracic, IR — the radiologists doing the highest scoring-framework volume. Validation Partner program is currently selecting practices.
Apply →Free access in exchange for feedback
Residency and fellowship trainees get 3+ months of free access. Apply with PGY year and program. Designed to teach structured reporting alongside clinical scoring frameworks.
Apply (residents) →Standardize departmental reporting
Practice directors using Precision AI to bring consistency across radiologists — same finding reported the same way, every time, with traceable scoring.
Talk to us →Enforce quality across distributed reads
Teleradiology groups whose radiologists work across many client systems benefit most from baked-in scoring frameworks. Same report quality regardless of site template.
Contact →Two ways to get access
🩻 Validation Partner (practicing rads)
Three months of full Precision AI access in exchange for structured feedback from real reads. Selecting practices now.
Apply →🎓 Residents & Fellows — Free
3+ months free in exchange for feedback. ACR + SIR templates, guideline popups, full scoring layer. Apply with PGY year + program.
Apply (trainees) →Common questions
Is this another image-analysis AI?
No. Precision AI works on the report layer — dictation, structuring, and clinical scoring. It does not analyze pixels. The radiologist still reads the study; Precision AI handles how that read becomes a report.
Which scoring frameworks are covered?
LI-RADS (current version), Lung-RADS, Fleischner Society follow-up criteria, Bosniak v2019, adrenal washout calculation, and a growing list of additional frameworks. New frameworks are added based on Validation Partner feedback.
Does it handle PHI?
Precision AI processes report text. PHI handling is part of the deployment configuration — most Validation Partner deployments are designed so PHI doesn't leave the practice's environment unnecessarily.
How do residents and fellows get free access?
Apply at /radiology-report-assistant/residents/ with PGY year and program. Free 3+ months in exchange for structured feedback. The trainee program is specifically about teaching structured reporting alongside the scoring layer.
What's the medico-legal posture?
Precision AI surfaces classification with traceable rationale. The radiologist remains the responsible interpreter and finalizes the report. The structured trail of how each category was assigned is more defensible than free-text impressions, not less.
Who built it?
Pouyan Golshani, MD — practicing interventional radiologist — and the GigHz team. Built because the dictation layer is where every read touches and where consistency tooling barely exists.
Stop recalling Bosniak from memory.
Apply as a Validation Partner — three months of full Precision AI access in exchange for structured feedback. Or apply for the trainee program if you're a resident or fellow.
GigHz Precision AI surfaces clinical scoring with traceable rationale to support — not replace — the interpreting radiologist. The radiologist remains responsible for the final report. A GigHz product. Built by Pouyan Golshani, MD.
Written and reviewed by Pouyan Golshani, MD, Interventional Radiologist — Last updated April 7, 2026