AI Scribe for Radiologists
Structured CT, MRI, mammography, and IR reports generated from a single dictation pass — with comparison logic and BI-RADS, LI-RADS, and Lung-RADS scoring built in. Built as a credible alternative to Dragon Medical One and M*Modal Fluency for the volumes that don't need custom macros.
Documentation for every modality
From overnight head CT and chest x-ray volume to complex IR cases and screening mammography, the same report engine handles your full read list.
CT Imaging
Structured reports for head, chest, abdomen/pelvis, CTA, and trauma protocols. Captures contrast type and volume, slice thickness, and reformations.
MRI Imaging
Brain, spine, MSK, MRCP, prostate, and breast MRI with sequence-by-sequence findings, gadolinium documentation, and PI-RADS where applicable.
Mammography
Screening and diagnostic mammography with automated BI-RADS 1-6 categorization, breast density (a-d), and prior comparison.
Plain Film Reading
High-volume chest x-ray, abdominal series, skeletal survey, and orthopedic post-op films with concise impression and comparison.
Interventional Radiology
Biopsies, drainages, ablations, embolizations, and vascular access. Captures fluoroscopy time, contrast dose, sedation, and complications.
Pediatric Imaging
Age-appropriate dose documentation, ALARA notes, and parental presence captured for plain film, ultrasound, and limited CT/MRI.
Built for radiology reporting workflows
Designed around the parts of dictation that drag the day out — comparison, scoring, and the IR paperwork that gets dinged on chart audit.
Comparison-aware findings
Pulls measurements and impressions from prior reports so growth rates, RECIST response, and "stable from prior" language are generated automatically with the correct exam date reference.
Structured scoring built in
BI-RADS for breast, LI-RADS for liver, Lung-RADS for incidental pulmonary nodules, TI-RADS for thyroid, and PI-RADS for prostate are applied based on the findings you describe — not from a separate dropdown.
Critical findings audit trail
Time-stamped logs for critical and non-routine communications, with prompts that satisfy ACR and Joint Commission requirements for closed-loop notification of referring providers.
IR procedure documentation
Pre-procedure consent, time-out, fluoroscopy time, contrast dose, sedation drugs and doses, complications, and post-procedure orders captured in a single pass — the parts most often dinged on chart audit.
House-style templates
Import your existing PowerScribe or Fluency templates so reports come out in your group's preferred phrasing. Section order, default negatives, and signature blocks all customize per radiologist.
Speed-to-final-report
Dictate or upload a 60-90 second voice memo and get a structured draft back in under 30 seconds. Final sign-off is typically within 2-3 minutes for routine plain film and 5-7 minutes for cross-sectional studies.
A typical body-imaging shift, before and after
Radiology is the one specialty where dictation has been mature for two decades. The marginal gain from ambient AI is not transcription accuracy — it is structure.
On a 50-study body-imaging list, the bottleneck is rarely how fast Dragon transcribes the words you say. It is the lookup-and-paste cycle for the prior comparison, the manual selection of LI-RADS or Lung-RADS from a dropdown, and the time-out and contrast-dose paperwork that has to be perfect on every IR case. Those three patterns account for roughly 30-40% of report time on a routine list.
PatientNotes treats those patterns as the actual product. When you describe a 3.2 cm arterially enhancing segment 7 mass with portal venous washout in a known cirrhotic, the report comes back with LI-RADS 5 already applied, the prior measurement ("previously 2.4 cm on 11/04/2025") pulled forward, and a structured impression that lists the lesion, the cirrhosis context, and the recommendation. You read, edit if needed, and sign.
The same logic applies to lung nodule follow-up (Lung-RADS 4B versus 3 versus 2 based on size, density, and growth), thyroid (TI-RADS), prostate MRI (PI-RADS), and screening mammography (BI-RADS plus density category). Last updated April 2026, the structured-output coverage spans all five major scoring systems plus standard plain film and ultrasound templates. The result is not "AI does your job" — it is "the boring 30% of every report is already done when you start reading."
Sample AI-generated CT chest report
A real-world Lung-RADS follow-up report — comparison-aware, scored, with critical-findings audit trail.
EXAM: CT CHEST WITH IV CONTRAST (CPT 71260) PATIENT: 64-year-old female DATE OF SERVICE: 04/12/2026 CLINICAL HISTORY: Former smoker (35 pack-years, quit 2019) with 8 mm right upper lobe pulmonary nodule on screening low-dose CT 6 months ago. Lung cancer screening follow-up. No new respiratory symptoms. TECHNIQUE: Helical CT of the chest performed from the thoracic inlet through the upper abdomen following administration of 100 mL Omnipaque 350 IV contrast. 1.25 mm slice thickness with axial, coronal, and sagittal reformations. Lung and mediastinal windows reviewed. COMPARISON: Low-dose screening CT chest 10/15/2025. CT chest with contrast 03/04/2024. FINDINGS: Lungs and Airways: - Right upper lobe, posterior segment: 11 mm subpleural solid nodule (previously 8 mm on 10/15/2025, 6 mm on 03/04/2024). Volumetric doubling time approximately 280 days. - Left upper lobe apicoposterior segment: 3 mm subpleural nodule, stable from prior, Lung-RADS 2. - No new pulmonary nodules. - No pleural effusion or pneumothorax. - Mild centrilobular emphysema, predominantly upper lobes, unchanged. - Trachea and central airways are patent. Mediastinum and Hila: - No mediastinal or hilar lymphadenopathy by size criteria. - Right paratracheal lymph node measures 8 mm in short axis, stable. - Heart size is normal. No pericardial effusion. - Mild coronary artery calcification, multivessel. - Thoracic aorta is normal in caliber. No dissection. Pleura and Chest Wall: - No pleural thickening or effusion. - No suspicious chest wall mass. - Mild degenerative changes of the thoracic spine. Upper Abdomen (limited evaluation): - Liver, spleen, and adrenals are unremarkable. - No suspicious upper abdominal lesion. Bones: - No suspicious lytic or blastic lesion. - Old healed left posterior 7th rib fracture. IMPRESSION: 1. Growing right upper lobe subpleural solid nodule, now 11 mm (previously 8 mm, 6 mm). Lung-RADS 4B - suspicious for primary lung malignancy. PET-CT and tissue sampling recommended. 2. Stable 3 mm left upper lobe nodule, Lung-RADS 2. 3. Mild upper-lobe predominant centrilobular emphysema, unchanged. 4. Multivessel coronary artery calcification - clinical correlation as indicated. RECOMMENDATIONS: - PET-CT for staging - CT-guided or bronchoscopic biopsy of right upper lobe nodule - Multidisciplinary thoracic oncology discussion - Pulmonology referral CRITICAL FINDINGS COMMUNICATION: Lung-RADS 4B finding communicated to Dr. Helen Chen (referring PCP) at 14:22 on 04/12/2026 via secure message; read receipt confirmed at 14:31. ICD-10: R91.1 (Solitary pulmonary nodule), Z87.891 (Personal history of nicotine dependence) Electronically signed by: PatientNotes Clinical Team Date/Time: 04/12/2026 14:18
CPT codes that get suggested automatically
Radiology billing pivots on technique documentation. PatientNotes flags missing contrast volume, fluoroscopy time, or sedation details before they cost reimbursement.
71250CT chest without contrast74177CT abdomen and pelvis with contrast70551MRI brain without contrast70553MRI brain with and without contrast77067Screening mammography, bilateral76700Ultrasound, abdominal, complete77002Fluoroscopic guidance for needle placement19083Breast biopsy with imaging guidanceCommon ICD-10 pairings include R91.1 (solitary pulmonary nodule), Z03.89 (encounter for observation), R93.X (abnormal imaging findings), and modality-specific Z-codes. AI suggests both CPT and ICD-10 from the report content.
How radiologists actually use it
Three working radiologists, three different deployment patterns. None of them dropped Dragon overnight.
Dr. M. Patel, MD
Solo teleradiologist, night coverage for three rural hospitals
Reading 80-120 studies an overnight shift on Dragon Medical One. Switched routine plain film and head CT to PatientNotes after a one-week trial; cut average time-per-chest-x-ray report from 90 seconds to 35 seconds because the structured impression and comparison line populate from the dictation. Kept Dragon for IR follow-ups with custom macros. Net cost change: dropped a $1,800/year Dragon seat, added a $600/year PatientNotes seat.
Northwest Imaging Group, 14 radiologists
Outpatient subspecialty group with mammography focus
Mammographers were spending 6-8 minutes per diagnostic mammogram report because BI-RADS and density had to be selected manually in PowerScribe. PatientNotes pulls density and BI-RADS straight from the dictation. Group-wide pilot reduced average mammography report turnaround from 14 hours to under 4 hours and cut after-hours catch-up reading by roughly half.
Dr. R. Okafor, MD
Hospital-employed body-imaging radiologist, 600-bed academic center
Took on increased CT volume during a colleague's parental leave. Used PatientNotes for routine abdomen/pelvis CT and HCC follow-ups while staying on Fluency for the institutional PowerScribe templates. Liked the LI-RADS automation and the comparison logic that pulled forward 'increased from 2.1 cm to 3.2 cm' without manual lookup. Estimated time savings: about 45 minutes per 50-study list.
Coming from Dragon Medical One?
Dragon has been the radiology dictation default since the early 2000s. It is mature, deeply EHR-integrated through PowerScribe and Fluency, and most senior radiologists have trained voice profiles on it. None of that goes away the day you sign up for PatientNotes.
The honest case for trying PatientNotes alongside Dragon is not "Dragon is bad." It is that the dictation-as-transcription model — where you say every word and the software writes it down — is now competitive with ambient-AI structured generation for most routine modalities. A modern radiologist reading 80 chest x-rays, 30 head CTs, and 15 follow-up body cases on a shift spends a lot of time on report scaffolding that AI now generates from a 30-second dictation.
The cost gap is also real. A six-radiologist outpatient group on Dragon is paying somewhere between $7,200 and $14,400 per year more than the equivalent PatientNotes deployment. For a single solo teleradiologist, the difference is roughly $600 to $1,800 per year — not life-changing, but a credible line item.
How to switch (or run side-by-side)
- Week one:Create an account and import your three or four most-used PowerScribe or Fluency templates. PatientNotes will match the section order and default negatives so the output looks like your group's house style.
- Weeks two and three: Pick one routine modality — usually chest x-ray, head CT, or screening mammography — and run both systems in parallel. Dictate the study through PatientNotes, review the structured output, and only fall back to Dragon if something is missing. Most radiologists trust PatientNotes for that modality within ten studies.
- Week four onward: Move that modality fully to PatientNotes. Keep Dragon for IR cases and any subspecialty work that depends on heavy custom voice macros until those macros are recreated. A full Dragon-replacement timeline is typically two to three months.
Detailed feature-by-feature comparison lives at /compare/dragon-medical-one and the broader ambient-vs-dictation discussion at /compare/nuance-dax.
PatientNotes vs M*Modal Fluency and Nuance DAX
The other entrenched radiology dictation stack is Nuance / Microsoft's M*Modal Fluency Direct, often bundled with PowerScribe at hospital systems. Here is the honest side-by-side.
For large hospital deployments where PowerScribe and Fluency are already a contract line item, ripping them out is rarely worth it. PatientNotes makes most sense for solo teleradiologists, outpatient imaging groups under 25 readers, and as a cost-justified second tool for routine modalities at larger groups.
Frequently asked questions
Last updated April 2026 by the PatientNotes Clinical Team.
Does PatientNotes work for radiology?
Yes. It generates structured reports across CT, MRI, plain film, ultrasound, mammography, and interventional procedures, with BI-RADS, LI-RADS, Lung-RADS, TI-RADS, and PI-RADS scoring applied from the findings you describe. Output is paste-ready text for PowerScribe, Fluency, Epic Radiant, or any RIS/PACS reporting field.
How much does an AI scribe cost for radiology?
PatientNotes is $50 per radiologist per month, or $600 per year. Dragon Medical One typically runs $99 to $200 per month per user ($1,200 to $2,400 per year) and M*Modal Fluency Direct is sold via per-seat enterprise contracts in a similar range. A six-radiologist group on Dragon pays roughly $7,200 to $14,400 per year more than the equivalent PatientNotes deployment.
Is Dragon Medical One better than PatientNotes for radiology?
Dragon is more entrenched in radiology than in any other specialty — it has been the dictation standard for two decades and integrates deeply with PowerScribe and Fluency. PatientNotes does not replace custom voice macros directly. What it does well is turn a freeform dictation into a structured report with comparison logic and scoring already filled in. Many radiologists run both during a transition; the cost gap (about $50 vs $99-200/month) makes the parallel pilot easy.
Can radiology-specific reports be auto-generated?
Yes. Modality-specific templates exist for CT (head, chest, abdomen/pelvis, CTA), MRI (brain, spine, MSK, MRCP, prostate), plain film, ultrasound, mammography (with BI-RADS), DEXA, fluoroscopy, and interventional procedures. The AI populates technique, comparison, findings by region, and impression from your dictation. Custom house-style templates can be added in the dashboard.
Does it integrate with Epic Radiant, PowerScribe, or Fluency?
Not yet via native plugin. Reports are produced as fully formatted text that pastes directly into Epic Radiant, PowerScribe 360, M*Modal Fluency, or any RIS/PACS reporting field with headings and structure preserved. HL7 and FHIR feeds are on the 2026 roadmap for hospital deployments.
Will it help me bill radiology CPT codes correctly?
PatientNotes suggests CPT codes (71250 chest CT non-contrast, 70551 brain MRI non-contrast, 77067 screening mammography, 76700 complete abdominal ultrasound, 77002 fluoroscopic needle guidance, etc.) directly from your report. It flags missing technique details — contrast type and volume, fluoroscopy time, sedation drug and dose — which is most useful on IR cases where documentation gaps cost reimbursement.
How do I switch from Dragon Medical One to PatientNotes?
Most radiologists pilot both for two to four weeks. Week one: import your top PowerScribe or Fluency templates. Weeks two-three: dictate one routine modality (chest x-ray, head CT, screening mammo) through PatientNotes and compare to your Dragon report. Week four onward: move that modality fully, keep Dragon for IR or subspecialty work with custom macros until you recreate them. A complete Dragon-off transition typically takes two to three months for solo readers.
Is it HIPAA compliant for radiology?
Yes. PatientNotes is HIPAA-compliant, signs a Business Associate Agreement (BAA), encrypts data in transit and at rest, and stores PHI in US-based data centers. Audio is processed in-region, never used to train external models, and can auto-purge after report generation. Hospital and academic groups can request the SOC 2 Type II report and security questionnaire.
Related specialty pages
Read more studies. Spend less time on scaffolding.
Pilot PatientNotes alongside Dragon for a week. Most radiologists move at least one routine modality over within ten studies and never look back.
No credit card required. $50/month after trial. Cancel any time.