Audit & data mapping
Mapped the existing PACS worklist, queue data and radiologist review patterns to find where urgency signals were being lost to FIFO ordering.
HEALTHCARE · CASE STUDY
Meridian Health cut chart review time −58%
Imaging-triage AI that flags urgent studies first, cutting turnaround on critical cases from a day to under an hour. Runs inside the existing PACS workflow, with clinicians retaining full override control.
- 12 weeks to production
- PACS-native, zero new tools
- 100% clinician override retained
MedAI Live−58%chart review time
Capabilities
Urgency scoring
Live queue re-ordering
PACS-native
Clinician override
The urgency-sorted worklist
The worklist radiologists see the moment they sign in — sorted by urgency, not arrival time.
LIVE WORKLIST — RE-SORTED BY URGENCYAUTO-REFRESH
MRN-8841Chest CT, contrastSuspected acute finding · flagged 00:02 agoURGENT< 1 min
MRN-7723Head MRI, non-contrastBorderline signal · needs reviewREVIEW4 min
MRN-9012Abdomen CTRoutine follow-upROUTINE22 min
MRN-6650Knee MRIRoutine, pre-opROUTINE31 min
Every flag is a prioritization signal — a radiologist reviews and can override each one.
Why Meridian called us
A routine clinical-governance review flagged that critical-finding turnaround was being driven by queue position, not urgency — a same-day fix wasn't optional once it was logged as a patient-safety risk. That set a hard deadline: something live before the next quarterly review, not a multi-year platform overhaul.
Cut urgent-case turnaround by re-prioritizing the imaging review queue on clinical urgency, not arrival order.
12 weeks to productionTimeline
AI engineers + clinical workflow leadTeam
From audit to production
Why 12 weeks was realistic, not a marketing number — the scope was one model and one integration point, not a platform rebuild. Here's the breakdown:
Model build & validation
Trained the urgency-scoring model on imaging metadata and validated it against radiologist-labeled priority calls before it touched a live queue.
PACS integration
Wired the scoring layer directly into the existing PACS worklist — no separate login, no new interface for radiologists to learn.
Clinician rollout
Rolled out with full override control from day one, then tuned the model on real override activity during the first weeks live.
What we delivered
The stack behind Meridian's real-time urgency-scoring layer.
Python
PyTorch
Hugging Face
ONNX Runtime
FastAPI
Redis
PostgreSQL
Docker
Plus PACS/DICOM and HL7 integration into Meridian's existing clinical systems — the connective work most off-the-shelf AI vendors skip.
The challenge
Radiology and clinical teams were reviewing imaging studies in strict FIFO order, so studies with urgent findings sat in the same queue as routine follow-ups. Turnaround on genuinely critical cases stretched to a full day, driven by queue position rather than clinical urgency.
What we built
We built an imaging-triage layer that scores incoming studies for urgency signals and re-orders the review queue in real time, surfacing likely-urgent studies to the top without removing a clinician from the loop — every flag is a prioritization signal, not an autonomous diagnosis. The system integrates with the existing PACS workflow so radiologists work from the same interface, just a smarter queue.
Core capabilities
URG
Urgency scoring
Every incoming study is scored for likely-urgent findings the moment it lands.
SORT
Live queue re-ordering
The review queue re-sorts in real time as new studies arrive — no manual triage.
PACS
PACS-native
Runs inside the existing PACS interface radiologists already use, zero new tools to learn.
MD
Clinician override
Every AI flag is a suggestion a clinician can accept, deprioritize, or override outright.
Inside MedAI
The urgency-scoring worklist radiologists use every shift, from sign-in to audit trail.
Main feature — urgency-sorted worklist
Dashboard — turnaround & volume analytics
Login — clinician sign-in
Detail — explainable urgency score
Audit log — alerts & overrides
The outcome
Median turnaround on urgent-flagged studies dropped from roughly a day to under an hour, a 58% reduction in chart review time for the highest-priority cases. Clinical staff retained full override control over every triage decision throughout.
−58%chart review time
12weeks to production
3AI systems shipped
What the clinical team is saying
Straight from the team running triage on this system every shift.
“We almost didn't bother piloting this — we'd seen imaging AI vendors before and it's usually a black-box score nobody on the floor trusts. This wasn't that. It sits inside the PACS we already use so there was no new tool to train anyone on and every flag comes with the actual reason not just a number. Studies that used to sit for hours now surface the moment they matter and our radiologists still have full override on every one. It's the first AI tool our clinical staff asked us to keep.”
Amira KhalilClinical Operations Lead, HealthcareFrequently Asked Questions
No. Every urgency score is a prioritization signal, not a diagnosis — a radiologist reviews and can accept, deprioritize, or override any flag before it affects the read order.
The triage layer runs inside your existing PACS environment and processes studies under the same access controls and data-residency rules already governing that system — no data leaves your clinical network to a third party.
12 weeks from kickoff to production for Meridian Health, covering the scoring model, PACS integration, and clinician rollout — not just a proof of concept.
A wrong score costs a clinician a few seconds of re-reading a queue position — it never removes a study from review or auto-closes a case. Override activity is logged and reviewed to keep the model tuned.
No — it's PACS-native. Radiologists keep working in the same worklist interface they already use; the queue just arrives pre-sorted by urgency.
◆ Let's build
Ready to put AI to work in your industry?
Tell us your challenge. We'll come back with a concrete, no-obligation plan and a live demo of what's possible for your team.
- Free AI auditWe map the highest-ROI AI opportunities across your workflows.
- Prototype in weeksA working proof-of-concept on your real data before you commit.
- One accountable teamStrategy, models, data and deployment — end to end.
120+ teams shipped across 6 industries
Book a free demo
Reply within 1 business day · No obligation.