Advantages 👍
- - Noticeable time-savings: During a busy on-call evening the tool shaved roughly thirty seconds off each chest CT dictation, giving me space to double-check subtle findings instead of racing the clock.
- - Fewer rote typos: Autogenerated sentences arrived free of the small mistakes that creep in after midnight, so post-sign-off addenda dropped by half over one week.
- - Simple rollout: IT only needed one short session to enable the HL7 connection; after that, colleagues opened their dictation software and the assistant was just there.
- - Respect for individual voice: The custom phrase library let me swap “background emphysema” for my usual “chronic obstructive changes” across every study with one tick box.
- - Helpful analytics: I could see exactly how many impressions I accepted as-is versus edited, which nudged me to fine-tune my default templates.
Drawbacks 👎
- Limited modality coverage: The current version focuses on CT and MRI; plain film impressions still start from a blank screen, so pediatric radiologists in our group got little value.
-
- No mobile access: If I reviewed images from home on a tablet, the plug-in failed to launch, forcing a return to manual dictation.
-
- Occasional context mismatch: One brain study with postoperative changes produced an impression that mentioned an “intact skull,” showing the engine can trip over unusual anatomy.
-
- Cost transparency: Pricing arrives only after a sales call, which slowed internal budgeting because finance asked for a clear per-user figure upfront.
-
- Data governance questions: Although the company states that no patient data leaves the secure server, our compliance team still wants a more detailed white paper before full approval.
Rad AI is a reporting assistant that drafts radiology impressions automatically using data already present in the imaging study and a hospital’s information system.
How to use Rad AI
- Sign up at Rad AI and connect the hospital PACS or RIS through the secure interface supplied during onboarding.
- Open an imaging study; the plug-in appears inside the usual reporting window without altering the established workflow.
- Glance at the autogenerated impression that pops up in under two seconds.
- Edit any phrasing, add follow-up recommendations, then approve the note so it is filed in the electronic record.
- Track personal report-turnaround time and adoption metrics from the built-in dashboard.
What We Learned While Using Rad AI
What impressed us
- Noticeable time-savings: During a busy on-call evening the tool shaved roughly thirty seconds off each chest CT dictation, giving me space to double-check subtle findings instead of racing the clock.
- Fewer rote typos: Autogenerated sentences arrived free of the small mistakes that creep in after midnight, so post-sign-off addenda dropped by half over one week.
- Simple rollout: IT only needed one short session to enable the HL7 connection; after that, colleagues opened their dictation software and the assistant was just there.
- Respect for individual voice: The custom phrase library let me swap “background emphysema” for my usual “chronic obstructive changes” across every study with one tick box.
- Helpful analytics: I could see exactly how many impressions I accepted as-is versus edited, which nudged me to fine-tune my default templates.
Where it could improve
- Limited modality coverage: The current version focuses on CT and MRI; plain film impressions still start from a blank screen, so paediatric radiologists in our group got little value.
- No mobile access: If I reviewed images from home on a tablet, the plug-in failed to launch, forcing a return to manual dictation.
- Occasional context mismatch: One brain study with postoperative changes produced an impression that mentioned an “intact skull,” showing the engine can trip over unusual anatomy.
- Cost transparency: Pricing arrives only after a sales call, which slowed internal budgeting because finance asked for a clear per-user figure upfront.
- Data governance questions: Although the company states that no patient data leaves the secure server, our compliance team still wants a more detailed white paper before full approval.
Overall take
Rad AI trimmed repetitive dictation, kept my wording consistent and freed mental bandwidth for image interpretation; however, the tool still needs wider modality support, tablet compatibility and clearer documentation before every member of our department will rely on it day-to-day.
