CT Pulmonary Angiography Pulmonary Embolism Detection Oversight

AI-assisted workflow for detecting pulmonary embolism on high-volume CT pulmonary angiography exams, with real-time monitoring of model safety and performance in radiology operations.

Business Blueprint

GROUNDED

AI flags pulmonary embolism on high-volume CT pulmonary angiography exams while radiologists and thoracic adjudicators oversee discordant findings.

The Problem

Radiology operations need to detect pulmonary embolism on CT pulmonary angiography at large scale and quantify how an FDA-cleared AI tool agrees or disagrees with AI-informed radiologists after implementation.

AI-informed radiologists

They remain responsible for pulmonary embolism detection on CTPA while working alongside AI outputs that must be checked for concordance.

Thoracic radiologists

They are needed to adjudicate radiologist-AI disagreements and establish the reference standard for discordant cases.

Process Fit

Diagnostic & decision support

As-Is

Adult CTPA exams are interpreted by radiologists, with quality review focused on whether pulmonary embolism is detected correctly across a high-volume imaging workload.

To-Be

Each adult CTPA receives real-time AI analysis alongside radiologist interpretation; when the AI and radiologist disagree, thoracic radiologists adjudicate through the AI Quality Oversight Process and the adjudicator diagnosis becomes the reference standard for those discordant cases.

Human Checkpoints

  • Radiologist interpretation of each CTPA remains part of the diagnostic workflow. — AI-informed radiologist
  • Radiologist-AI disagreements are routed for expert adjudication. — Thoracic radiologist
  • The adjudicator diagnosis is used as the reference standard for discordant cases. — Thoracic radiologist adjudicator

Systems Touched

CT pulmonary angiography imaging workflowPE AI tool (AIDOC)Radiologist interpretation workflowAI Quality Oversight Process

Business Cycle

Upstream

  • Adult CTPA exams must be available for real-time AI analysis and radiologist interpretation.
  • An FDA-cleared PE AI tool must be implemented across the clinical imaging network.
  • A defined oversight path must exist so radiologist-AI disagreements trigger thoracic radiologist adjudication.

Downstream

  • Radiology leadership can monitor concordance between the AI tool and radiologists after implementation.
  • Discordant AI and radiologist findings receive expert adjudication rather than being left unresolved.
  • The operation can identify complementary roles for AI and radiologists in PE detection.

Value Evidence

  • CTPA exam volume evaluatedIMPROVED

    32,501 CTPAs obtained from 29,492 patients (mean age, 62.4 years ± 18.6 [SD], 17,424 female) were evaluated.

  • Overall AI-radiologist concordanceIMPROVED

    Overall concordance was 97.79% (95% confidence interval [CI]: 97.62-97.94%), higher for AI-negative than for AI-positive examinations

  • AI-negative versus AI-positive concordance visibilityIMPROVED

    higher for AI-negative than for AI-positive examinations (98.18% vs 93.75%; P < .001).

  • Expert resolution of discordant casesIMPROVED

    Expert adjudication favored the radiologist in 88.73% of discordances.

  • Unique PE diagnoses contributed by radiologist versus AIIMPROVED

    The rate of unique diagnosis by the interpreting radiologist (483/3,226, 14.97%) was approximately 19 times that of the AI tool alone (26/3,226, 0.81%).

  • Concordance by PE featureIMPROVED

    Concordance varied by PE features: acute versus chronic (87.34% vs 60.12%; P < .001)

  • Concordance by PE locationIMPROVED

    location (central 95.79%, lobar/segmental 83.81%, subsegmental 58.62%; all P < .001).

Adoption Journey

  1. LEVEL 1 — QUICK WIN

    Gate: Prove value on a limited set of adult CTPA exams by showing that real-time AI analysis can run alongside radiologist interpretation without removing the radiologist from the diagnostic workflow.

    Outcome: The department gets an initial assistive read for PE detection while preserving radiologist accountability.

  2. LEVEL 2 — STANDARD

    Gate: Prove production readiness by measuring concordance between AI and AI-informed radiologists and defining how disagreements are handled.

    Outcome: Radiology operations can use the AI in routine PE detection with a measurable safety and quality oversight loop.

  3. LEVEL 3 — ADVANCED

    Gate: Prove scalability across an integrated network and high-volume CTPA workload while maintaining expert adjudication capacity.

    Outcome: The organization can monitor PE AI performance across many exams and sites, with thoracic radiologist review focused on discordant cases.

Detailed per-level builds in the solution spectrum below

Risk & Governance

  • AI and radiologists may disagree on whether PE is present.

    Posture: Route radiologist-AI disagreements to thoracic radiologists through the AI Quality Oversight Process, using the adjudicator diagnosis as the reference standard for discordant cases.

  • AI performance may not be uniform across PE types or locations.

    Posture: Monitor concordance by PE features and location, including acute versus chronic PE and central, lobar/segmental, and subsegmental locations.

  • The organization may over-rely on the AI tool and underweight radiologist judgment.

    Posture: Keep radiologists in the interpretation workflow and use expert oversight to confirm complementary roles between AI and radiologists.

Operating Intelligence

How it works

AI runs the first three steps autonomously.

Humans own every decision.

The system gets smarter each cycle.

Confidence89%
ArchetypeRecommend & Decide
Shape6-step converge
Human gates1
Autonomy
67%AI controls 4 of 6 steps

Who is in control at each step

Each column marks the operating owner for that step. AI-led actions sit above the divider, human decisions and feedback loops sit below it.

Loop shapeconverge

Step 1

Assemble Context

Step 2

Analyze

Step 3

Recommend

Step 4

Human Decision

Step 5

Execute

Step 6

Feedback

AI lead

Autonomous execution

1AI
2AI
3AI
5AI
gate

Human lead

Approval, override, feedback

4Human
6↺ Loop
AI-led step
Human-controlled step
Feedback loop
TL;DR

AI handles assembly, analysis, and execution. The human gate sits at the decision point. Every cycle refines future recommendations.

The Loop

6 steps

1 operating angles mapped

Operational Depth

Technologies

Technologies commonly used in CT Pulmonary Angiography Pulmonary Embolism Detection Oversight implementations:

AIDOC PE AI toolOther
4 mentions
CT pulmonary angiography imaging dataOther
4 mentions
Radiologist interpretation workflowOther
4 mentions
AI Quality Oversight ProcessOther
3 mentions
Thoracic radiologist adjudicationOther
3 mentions

Key Players

Companies actively working on CT Pulmonary Angiography Pulmonary Embolism Detection Oversight solutions:

Qure.aiRapidAIViz.ai

Real-World Use Cases

AI-assisted pulmonary embolism detection on CT pulmonary angiography with real-time quality oversight

An AI system checks chest CT scans for blood clots in the lungs and compares its findings with the radiologist’s report; when they disagree, specialist radiologists review the case.

Computer vision detection and classification of pulmonary embolism on CT pulmonary angiography, followed by human-AI concordance review.deployed at scale in clinical practice; evaluated retrospectively across 32,501 ctpa examinations with real-time ai analysis and adjudication workflow.
10.0

Free access to this report