RadioView.AI™ · SaveLife.AI™
An AI report draft radiologists trust enough to sign.
An AI co-pilot that turns a dictated read into a structured report, used by physicians across the US, Qatar, and UAE.


TL;DR
- Problem
- Radiologists juggle disconnected tools, and they won't sign AI text they can't check.
- What I did
- I designed a panel beside their PACS, the system they read scans in, turning AI output into a draft they review, edit, and sign.
- Result
- Radiologists sign the drafts. It is live at a US hospital partner and with Hearts Medical Solutions in Qatar, and no site had to replace its PACS.
Outcomes
- LiveAt a US hospital partner and Hearts Medical Solutions in Qatar
- 1M+Reports generated, SaveLife.AI™ platform total
- 510(k)NeuroICH, FDA cleared in 2024. I owned its reporting UI
Project details
- Role
- Senior Product Designer, sole designer
- Team
- Ahmed Daud (me)
- Dr. Junaid Kalia, CEO and practicing physician
- Three engineers
- Timeline
- May 2024 to Oct 2026
- Platforms
- Web, Chrome extension, iOS, Android
- Scope
- Reporting workflow, AI review patterns, NeuroICH reporting UI
- Status
- Live with physicians across the US, Qatar, and UAE
Problem: The Radiologist Is the Integration Layer

The scan sits in the PACS, the report in a dictation tool, the history in the EHR. The radiologist holds it all together, dozens of times a day.
Hours of every shift lost to dictation and cleanup
Shifts that often end with two more hours at home
The urgent case buried in a worklist of eighty
Research: Radiologists Asked for Trust, Not Speed
The same three worries came up in almost every interview.
- Method
- Interviews
- Participants
- 8+ radiologists
- Where
- Our early sites
- When
- Before picking a direction
Nobody would replace the PACS they already own
Led to Decision 01
Nobody would sign a report they couldn't check
Led to Decision 02
Most had been burned by confident, wrong AI
Led to Decision 03
Decisions: Three Decisions That Earned a Signature
Each one answers one of those worries.
Decision 01
Sit Beside the PACS
- Why
- The PACS is bought, integrated, and trusted. So Companion Mode runs beside it.
- Not this
- Reading in our viewer to get the AI. Nobody migrates a PACS for an AI feature.
- Result
- Hospitals could say yes without changing the PACS they trust.
- Status and priority flags on every study
- Edit the dictated read and regenerate the report from it
- Send stays disabled while the report is in progress
Decision 02
A Draft to Review and Sign
- Why
- The radiologist's name and license go on every report. So the AI drafts from their dictation, and they refine and sign.
- Not this
- A finished AI answer: rubber-stamp it and own any miss, or redo the work.
Speak the read and let the template handle the formatting.
- Upfront about latency instead of faking an instant result
01Speak
Speak the read and let the template handle the formatting.
Callouts on this screen: - Upfront about latency instead of faking an instant result
02Refine
Clean up the transcript in Scribe Mode before the report is generated.
Callouts on this screen: - Every AI edit announces itself and offers Undo
03Sign
Sign, then dispatch the report from the same screen.
Callouts on this screen: - A plain-language summary written for the patient
Decision 03
Show the Work on Every Measurement
- Why
- A number with no visible basis gets measured again by hand. So the segmentation stays on the image, visible and editable.
- AI or Manual: the radiologist chooses who draws
- The region sits on the scan, checkable at a glance
- Each segment can be hidden, edited, or deleted
- Values enter the report only when the radiologist adds them
NeuroICH: An Alert Built Around the Handoff
NeuroICH flags suspected intracranial hemorrhage on a non-contrast head CT and alerts a specialist's phone. The clinician makes the call. I owned its reporting UI, including the handoff to a radiologist.
< 1 min
To notify a specialist of a suspected bleed
Mean time to notification in the 510(k) study: 0.37 minutes, about 22 seconds.
94.8%
Sensitivity for intracranial hemorrhage
With 92.5% specificity, across 376 head CT studies.
510(k)
FDA cleared in 2024
K241719. Every figure here comes from its public summary.
Insight
Where the nearest radiologist can be hours away, SaveLife.AI™ made NeuroICH free across low- and middle-income countries.
Impact: Radiologists Sign the Drafts
Where it runs
- Named sites
- A US hospital partner; Hearts Medical Solutions, Qatar
- Readers
- Physicians across the US, Qatar, and UAE
- Platforms
- Web, Chrome extension, iOS, Android
- PACS replaced
- None. Companion Mode runs beside it
Reflection: What Trust Taught Me
Trust is the outcome that matters
A draft gets signed when it is easy to check and never overconfident.
Fit the workflow before adding to it
Changing nothing hospitals trusted mattered more than any feature.
Reversible beats perfect
The model will be wrong sometimes, so every AI edit can be undone.
Next Case Study
SaveLife.AI™ · Clinical AI
AizaMD™
An AI scribe designed to give physicians the visit back.
Shipped on 4 platforms
View Case Study

Building clinical AI? I'd like to hear about it.
I'm open to senior product design roles in healthcare and clinical AI. Email is the quickest way to reach me.
Or write to theedaud@gmail.com