AI use case

AI prioritization of radiology and imaging worklists

An AI system that analyzes a medical image immediately after a scan, flags time sensitive findings such as a brain bleed, a stroke causing large vessel occlusion or a pulmonary embolism, and reorders the radiologist's worklist and notifies the care team so the most urgent cases are read and acted on first, while a radiologist confirms every finding before it changes a patient's treatment.

By Len Debets · Last verified 28 September 2026 · 2 public deployments

About 44%
Reported cycle time reduction
Adventist Health + Rideout, vendor claim.
USD 400,000 to USD 7.2 million
Indicative value per year
A 400 bed hospital reading 40,000 CT and MRI studies a year for time sensitive pathologies. Worked example, see how it is calculated.

What problem does it solve?

Within the same priority class, a radiology worklist is normally read in roughly the order scans arrive. A scan that shows a brain bleed or a blood clot blocking a major vessel can sit behind several routine studies at the same priority level before a radiologist opens it, and for time sensitive conditions every extra minute has a cost: in acute ischemic stroke, treatment delay is directly linked to worse outcomes. The problem compounds at regional or community hospitals, where a patient needing specialist treatment must first be identified, then transferred to a comprehensive center, a handoff that traditionally depends on a radiologist's read, a phone call to a specialist, and a manual transfer process.

How does it work?

  1. Scan and analyze. As soon as a CT or MRI is acquired, an AI model, cleared for that specific use, analyzes it for the patterns it is trained to detect, such as intracranial hemorrhage, large vessel occlusion or pulmonary embolism.
  2. Flag and notify. A positive finding pushes the case to the top of the radiologist's worklist and sends a mobile alert to the on call specialist and care team, often within seconds of the scan completing.
  3. Confirm and act. The radiologist reviews the flagged images and confirms or rules out the finding; the specialist team begins the treatment pathway, such as a thrombectomy, transfer or surgery, based on the confirmed read, not the AI flag alone.
  4. Coordinate transfer. At a regional hospital without full stroke or trauma capability, the same alert can trigger a standardized transfer protocol and direct communication with a comprehensive center.
  5. Audit and monitor. Every flagged and missed case feeds back into ongoing monitoring of sensitivity, specificity and turnaround time by pathology, site and shift.
Audience
Employee facing
Autonomy
Assist
Adoption
Mainstream
Channels
Internal tools

What is it worth?

Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.

Value benchmarks for AI prioritization of radiology and imaging worklists
KPIMedianReported rangeData pointsClaimed by
Cycle time reductionToo few to pool
about 44%
11 vendor

Value drivers: Speed and cycle time, Risk and loss reduction, Employee productivity.

Indicative value

A 400 bed hospital reading 40,000 CT and MRI studies a year for time sensitive pathologies

USD 400,000 to USD 7.2 million

Annual value of faster time sensitive treatment per year

How this is calculated

Formula: studies * positiveShare * minutesSaved * valuePerMinute. The low scenario uses every low input, the high scenario every high input.

InputLowHighBasis
Time sensitive CT and MRI studies read per year studies, studies per year40,00040,000Editorial assumption for a 400 bed hospital's time sensitive imaging volume; replace with your own case mix.
Share of studies with a time sensitive positive finding positiveShare, fraction of studies0.020.05Editorial assumption across intracranial hemorrhage, large vessel occlusion and pulmonary embolism screening; replace with your own case mix.
Minutes of care team activation time saved per positive case minutesSaved, minutes per case1030Below the figure in Viz.ai's release (describing the Adventist Health + Rideout deployment, reporting care team notification time falling from 45 minutes to 7 minutes, a 38 minute reduction, for large vessel occlusion stroke at one hospital), because this input averages across intracranial hemorrhage, large vessel occlusion and pulmonary embolism, and the 38 minute figure covers only large vessel occlusion. This figure is not a recorded metric or a computed benchmark; the evidence record does not report it as a metric because it measures a narrower step, care team notification, than the end to end transfer time the taxonomy's Cycle time reduction KPI defines, and no KPI in this taxonomy covers that step on its own.
Value of a minute of faster time sensitive treatment valuePerMinute, USD per minute50120Editorial assumption combining avoided length of stay, disability and readmission cost for time sensitive conditions; replace with your own health economic estimate.

What it leaves out: A rough proxy for the value of speed only. It leaves out the cost of the software and its integration, the value of pathologies not modeled here, and the fact that faster notification does not guarantee a faster or better clinical outcome for every patient.

Who already uses it?

2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.

Sheba Medical Center

Israel · Healthcare · 2025

ScaledGrade B

Sheba Medical Center, where Aidoc originated, has embedded Aidoc's AI platform across its emergency and radiology workflows to flag urgent findings, including intracerebral hemorrhage, large vessel occlusion stroke and pulmonary embolism, directly on images in real time, and alert the treating physician on desktop and mobile. Sheba's own account of the deployment cites a peer reviewed clinical study that found the integration of Aidoc into its emergency workflow was associated with a 30% reduction in mortality for patients with intracerebral hemorrhage, earlier treatment initiation, improved discharge outcomes and fewer unnecessary ICU stays. Sheba's own site frames Aidoc as one part of its wider Smart Hospital program, and lists a separate initiative, Project K, an AI powered emergency room, as a related case study; the page does not describe Project K as an extension of Aidoc's triage.

No outcome disclosed.

Adventist Health + Rideout

United States · Healthcare · 2026

ProductionGrade C

Adventist Health + Rideout, a regional primary stroke center in a hub and spoke network, used the Viz.ai platform's real time imaging analysis and automated care coordination as part of a quality improvement initiative to speed up the transfer of large vessel occlusion stroke patients to a comprehensive stroke center. The program combined the AI platform with a partnership with a comprehensive stroke center and standardized transfer protocols. Data presented at the American Heart Association's 2026 International Stroke Conference, led by the hospital's stroke program manager Caezar G. Jara, showed the changes cut average door in door out transfer time to 113 minutes, below the Joint Commission's 120 minute national benchmark.

  • Cycle time reduction: about 44%, quality improvement initiative combining Viz.ai platform deployment, a partnership with a comprehensive stroke center, and standardized transfer protocols, presented at ISC 2026
    "Viz.ai, the leader in AI-powered disease detection and intelligent care coordination, today announced the presentation of new clinical data at the American Heart Association's International Stroke Conference (ISC) 2026 demonstrating a 44% reduction in door-in-door-out (DIDO) time — the time required to evaluate, coordinate, and transfer a patient to a comprehensive stroke center — for patients with large vessel occlusion (LVO) stroke in regional care settings."
    Claimed by: vendor

How do you implement it?

A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.

Data you need

  • A regulatory clearance (FDA clearance or CE mark) that covers the pathology, patient population and scanner protocols in use
  • PACS and imaging archive integration for the modalities in scope
  • A defined care team activation and paging workflow to wire the alert into

Systems to integrate

  • Picture archiving and communication system (PACS)
  • Radiology information system and worklist
  • Clinician paging and care team activation system
  • Electronic health record, for the confirmed finding and downstream care pathway

Complexity: High

Analyzing the image is a small part of the work. The device must be cleared or CE marked for the exact indication, scanner types and patient population in use, integrated with PACS and the hospital's paging and care team activation system, and validated on local data, all before it changes a single worklist.

  1. 1

    Start with one time critical pathology with a clear clinical owner

    Pick a pathology such as large vessel occlusion stroke or intracranial hemorrhage where a stroke or trauma lead can own the rollout, rather than deploying every available module at once.

  2. 2

    Confirm the clearance matches your population before go live

    Check that the FDA clearance or CE mark covers your scanner models, contrast protocols and patient population; performance on a mismatched population can be unreliable and go unnoticed.

  3. 3

    Wire the alert into the paging system specialists already use

    Route the notification through the existing on call and care team activation workflow, not a new inbox nobody checks at 3am.

  4. 4

    Set a turnaround time target per pathology and measure it before and after

    Track time from scan completion to notification and to radiologist confirmation, not only sensitivity and specificity.

  5. 5

    Add pathologies and sites one at a time, each with its own validation

    Treat every new pathology or site as a new rollout with its own local validation, not an automatic extension of the first one.

Guardrails

  • Every flagged finding is confirmed by a radiologist before it changes a treatment plan; the AI reorders the queue, it does not diagnose
  • The system only runs within its cleared indications, scanner types and patient population
  • A defined fallback (acuity based or FIFO ordering) applies when the AI is unavailable or a study fails triage

KPIs to instrument

  • Time from scan completion to critical finding notification, by pathology and shift
  • Radiologist report turnaround time for flagged versus unflagged cases
  • False positive and false negative rate against a sampled radiologist read

Human in the loop

Radiologists confirm every AI flagged finding before it drives a clinical decision, and review a sample of unflagged cases to catch missed findings. A clinical safety lead owns the pathology's performance against its cleared claims and approves any expansion to a new site or population.

Common failure modes

Alert fatigue from false positives
Too many low value alerts and clinicians start deprioritizing all of them, including true positives. Track and act on the false positive rate per pathology and site, not only overall sensitivity.
Silent underperformance outside the cleared population
The model performs unreliably on a scanner protocol or patient group it was not validated on, and nobody notices because the model gives no signal that it is out of its depth. Validate on local data before go live and monitor for performance drift by site.

What are the risks and rules?

EU AI Act

High risk

Article 6(1) and Annex I: software that analyzes a medical image to detect or prioritize a disease finding is itself, or is a safety component of, a device in scope of the EU Medical Device Regulation, and typically needs a notified body conformity assessment as software as a medical device (the FDA's AI Enabled Medical Device List shows US market authorization for devices in this category, listing authorized stroke triage devices from Viz.ai and Aidoc's BriefCase triage devices), which makes it high risk under the EU AI Act regardless of Annex III. The radiologist's own diagnostic read stays a human decision; the AI narrows and reorders the queue. Annex I high risk classification under Article 6(1) applies from 2 August 2028 (Article 113(c)); until then, Article 4 (AI literacy obligations) and Article 5 (prohibited practices), which bind the hospital as a deployer, already apply.

Guidance

  • Regulation (EU) 2017/745 on medical devices (European Union, Europe). Does not mention AI by name. Software that provides information used to take decisions with diagnostic or therapeutic purposes, such as a finding used to prioritize or route a patient, is classified under Annex VIII Rule 11, usually as class IIa or higher, which requires a notified body conformity assessment before CE marking.
  • Article 6: Classification Rules for High-Risk AI Systems (European Union, Europe). A safety component of, or a product that is itself, a CE marked medical device under EU harmonisation legislation requiring third party conformity assessment is high risk under the EU AI Act. Per Article 113(c), this Annex I route applies from 2 August 2028, later than the 2 December 2027 date for the Annex III use cases.
  • List of Artificial Intelligence-Enabled Medical Devices (US Food and Drug Administration, North America). FDA's list of AI enabled medical devices authorized for marketing in the United States. The list names each device and its company, for example "Viz.ai, Inc.", along with a decision date and submission number; the cleared or granted indications are in the linked 510(k) or De Novo records. It lists authorized stroke triage devices from Viz.ai and Aidoc's BriefCase triage devices.

Controls to put in place

  • Maintain the regulatory clearance and intended use statement for each detection module, and never enable a pathology it is not cleared for
  • Radiologist confirms every flagged finding before it changes a treatment plan
  • Track sensitivity, specificity and turnaround time by pathology, site and shift against the cleared performance claims
  • Defined fallback ordering when the AI is unavailable or a study fails triage, so the worklist never silently reverts to an unmanaged queue

Frequently asked questions

What does AI radiology triage actually do?
It analyzes an image right after the scan, flags time sensitive findings it is cleared to detect, and reorders the radiologist's worklist and alerts the care team so urgent cases are read first. It does not replace the radiologist's diagnostic read.
Does the AI diagnose the patient?
No. The AI flags a likely finding and reprioritizes the queue; a radiologist confirms or rules out the finding before any treatment decision is made.
Is AI radiology triage regulated as a medical device?
Generally yes. In the United States these tools typically hold FDA clearance as software as a medical device, and in the EU they are usually CE marked medical devices under Annex VIII Rule 11 of the Medical Device Regulation. Because that classification requires a notified body conformity assessment, they are high risk under the EU AI Act's Annex I route, regardless of whether the specific use appears in Annex III, though that Annex I classification only takes effect on 2 August 2028.
How much time does it actually save?
It depends heavily on the pathology, the baseline workflow and the hospital. According to Viz.ai, a study led by Adventist Health + Rideout's stroke program manager and presented at the 2026 International Stroke Conference found that average door in door out transfer time for large vessel occlusion stroke patients fell by 44%, from 202 to 113 minutes, after a quality improvement program that included the Viz.ai platform, a partnership with a comprehensive stroke center and standardized transfer protocols.

How to cite this page

Blits.ai AI Use Case Library, "AI prioritization of radiology and imaging worklists", last verified 28 September 2026, https://www.blits.ai/ai-use-cases/radiology-worklist-triage. Licensed under CC BY 4.0. Method: how we verify use cases.

Changelog
  • 28 September 2026: First published

Related use cases

Healthcare

AI command center for hospital bed and staff capacity planning

An AI powered operations center that predicts patient admissions, discharges and transfers across a hospital or health system, and helps a team of coordinators sitting in one room sequence real time bed assignments, staffing levels and patient moves, so patients get into the right bed faster and existing capacity is used fully without adding beds.

Deployments
2 public, best grade B
Reported cycle time reduction
38%
Johns Hopkins Medicine, organization claim
Cross industryBanking

AI agent for IT service desk resolution

An AI agent in Microsoft Teams, Slack or the intranet that takes the high volume IT support queue, such as password and MFA resets, account unlocks, VPN, device and software requests, and resolves common requests by acting in the identity and IT service management systems, handing the rest to the right resolver group with the context attached.

Deployments
6 public, best grade B
Reported employee adoption
94%
Mercari US, vendor claim
HealthcarePharma and life sciences

AI clinical trial patient matching and prescreening

AI that reads structured data and clinical notes in the health record, compares each patient with the inclusion and exclusion criteria of open clinical trials, and gives research staff and treating clinicians a ranked list of likely eligible patients with the evidence for each criterion, so that people confirm eligibility and invite the patient.

Deployments
3 public, best grade B
Reported accuracy
100%
Cleveland Clinic, organization claim
Telecommunications

AI copilot for network operations centre fault triage

AI in the network operations centre (NOC) that correlates alarms and performance data from radio, transport, core and fixed networks into a small number of probable faults, ranks them by customer impact, proposes the likely root cause and fix from runbooks, vendor documentation and past tickets, and routes the ticket to the right team, while an engineer decides what to change.

Deployments
6 public, best grade B
Reported cycle time reduction
at least 95%
Deutsche Telekom, organization claim