CT Protocol Optimization

Your Protocols Are Costing You on Three Fronts Simultaneously

Dose compliance. Image quality. AI performance. Most CT protocol reviews address one. A GammaMetric protocol audit addresses all three — and shows you exactly where the trade-offs actually are.

Compliance Risk
Leapfrog Benchmarks Are a Moving Target
Leapfrog CT dose benchmarks are derived from the ACR Dose Index Registry's 25th percentile — meaning they tighten as the field improves. A protocol that passed last year may not pass this year.
Many hospital imaging programs fall above ACR benchmark thresholds for at least one standard exam type.
Non-compliance flags in public Leapfrog scores.
Dose & Quality
Dose Reduction and Image Quality Are Not the Trade-Off You Think
The common assumption — lower dose means lower quality — is often wrong. Reconstruction kernel and slice thickness have a larger impact on diagnostic performance than modest mAs reductions, and they're frequently misconfigured.
Protocol parameters like kernel and thickness account for the majority of image quality variance across a scanner fleet.
Dose can often be meaningfully reduced with no diagnostic penalty when the right parameters are targeted.
AI Performance
The Protocol Your Physicist Set May Be Degrading Your AI
AI detection algorithms are sensitive to acquisition parameters in ways that aren't obvious from image quality alone. A protocol that looks clinically acceptable can silently cut your algorithm's sensitivity by double digits.
GammaMetric validation data: soft kernel → −10.5pp sensitivity. 5mm slice → −13.2pp sensitivity.
Dose reduction alone: only ~−4pp.
01 The Core Finding

You're Optimizing the Wrong Parameter

When sites try to improve dose compliance, they typically target mAs. That's the right instinct — but the data shows it's the least consequential parameter for both image quality and AI performance. The parameters that actually matter are being left unexamined.

Key Insight

A 25% mAs reduction costs approximately 4 percentage points of AI sensitivity. Switching to a soft reconstruction kernel costs 10.5 points. Moving to 5mm slice thickness costs 13.2 points. The parameters driving Leapfrog non-compliance are often not the parameters destroying your AI's performance — and vice versa. You need to see both simultaneously.

Protocol Parameter Risk Map
Dose savings potential vs. AI sensitivity impact. Top-right quadrant is the target zone: meaningful dose reduction with minimal AI cost.
TARGET ZONE DANGER ZONE 0pp −4pp −8pp −12pp −16pp AI Sensitivity Impact 0% +10% +15% +20% +30% Dose Reduction Achieved → Soft Kernel −10.5pp sensitivity 5mm Slice −13.2pp, ~0% dose savings mAs Reduction −4pp sensitivity Optimized Protocol −20% dose, AI preserved
CTDIvol: Typical Site vs. ACR Benchmark
Approximate values from ACR Dose Index Registry public data. Leapfrog uses the ACR DIR 25th percentile as its threshold. Site medians vary; these figures are illustrative of typical exposure.
0 5 10 15 20 mGy CTDIvol (mGy) Chest CT 7.0 benchmark 11.0 typical Abd. CT 12.0 benchmark 17.0 typical CAP CT 14.0 benchmark 21.0 typical ACR/Leapfrog Benchmark (25th pctile) Approximate Median Site Dose
Protocol Parameter Dose Impact AI Sensitivity Impact Leapfrog Relevance Optimization Priority
mAs / tube current High (direct) Low (~−4pp) Primary metric MEDIUM
Reconstruction kernel Minimal High (−10.5pp) Indirect via noise HIGH
Slice thickness Minimal (~0%) Very high (−13.2pp) Indirect HIGH
kVp selection High (size-adjusted) Moderate SSDE component MEDIUM
Iterative reconstruction level Enables dose reduction Variable by vendor Permits lower mAs EVALUATE
What the data suggests most sites will find
≥1
kernel selection in the high-risk zone for AI performance
≥1
exam type with slice thickness above the AI-safe threshold
≥1
exam type above the Leapfrog CTDIvol benchmark
Most
sites have never reviewed kernel or slice thickness with AI use in mind
02 Entry Point

Start With a Protocol Risk Scan

A full fleet audit is the right long-term move. For sites that want to understand their exposure quickly — or validate that a problem exists before committing to a larger engagement — the Protocol Risk Scan is a defined, bounded first step.

Starter Engagement

Protocol Risk Scan

A targeted audit of 1–2 CT scanners covering dose compliance, parameter risk profiling, and AI interaction assessment. Delivered as a structured report with prioritized findings and specific protocol modification recommendations.

  • 1–2 scanner configurations audited
  • DLP/CTDIvol vs. Leapfrog benchmark comparison for up to 5 exam types
  • Kernel and slice thickness risk flag per protocol
  • AI sensitivity impact estimate where AI is deployed
  • Top 3 protocol modifications ranked by combined dose + AI benefit
  • No PHI required — audit works from protocol sheets and scanner configuration exports
Turnaround 2 weeks
On-site visit required Optional
PHI exposure None
Deliverable Written report + findings call
Extends to Full fleet audit or AI validation
What this surfaces
Most sites discover at least one high-priority parameter misconfiguration within the first scanner reviewed. The Risk Scan makes that visible in two weeks instead of never.
03 Full Engagement Deliverables

Complete Protocol Optimization Program

For sites ready to address their full scanner fleet — or those where the Risk Scan has surfaced findings worth acting on.

01
Fleet-Wide Protocol Benchmarking
Systematic DLP and CTDIvol assessment across all CT scanners and exam types, benchmarked against current ACR Dose Index Registry and Leapfrog thresholds.
  • CTDIvol and DLP per exam type per scanner
  • SSDE-based size-adjusted dose assessment
  • Leapfrog compliance status per scanner
  • Year-over-year trend analysis if prior data available
02
Parameter Risk Profile
Per-scanner, per-protocol breakdown of acquisition parameters flagged for dose, quality, or AI impact — with specific attention to kernel and thickness settings often overlooked in standard audits.
  • Kernel classification (sharp / standard / soft)
  • Slice thickness flag by clinical application
  • kVp and mAs optimization opportunity assessment
  • Iterative reconstruction utilization review
03
AI Interaction Analysis
For sites with deployed AI tools: quantitative mapping of which protocol configurations are in the high-risk zone for algorithm sensitivity degradation, assessed against known model behavior patterns from published validation research.
  • Protocol-AI sensitivity impact estimates
  • High-risk parameter flags per deployed algorithm
  • Recommended parameter guard rails
  • Trigger criteria for formal AI re-validation
04
Optimization Recommendations
Specific, ranked protocol modifications with projected impact on dose, image quality, and AI performance. Designed to be actionable — formatted for direct submission to scanner application specialists.
  • Ranked modification list by combined benefit
  • Projected CTDIvol reduction per change
  • AI sensitivity impact estimate per change
  • Implementation risk notes per recommendation
05
Compliance Documentation Package
Formatted documentation suitable for Leapfrog submission, ACR accreditation response, or internal QA committee review. Includes physicist attestation.
  • Pre/post optimization dose comparison
  • Leapfrog-formatted compliance summary
  • ABR board-eligible physicist signature
  • Methodology documentation for audit trail
06
Annual Monitoring Protocol
Structured plan for ongoing protocol surveillance — ensuring that scanner software updates, hardware changes, or protocol drift don't silently erode your compliance or AI performance baseline.
  • Annual re-audit framework
  • Protocol change tracking procedure
  • Scanner update impact checklist
  • Escalation criteria for urgent re-review
04 Engagement Path

From Risk Scan to Full Program

1
Protocol Risk Scan
1–2 scanners. Rapid exposure assessment. Identifies whether a broader problem exists before committing to a full program.
2 weeks
2
Fleet Benchmarking
Full scanner fleet dose and parameter audit. CTDIvol, SSDE, Leapfrog gap analysis across all active exam protocols.
3–5 weeks
3
Optimization & Implementation
Ranked protocol modifications implemented with application specialist coordination. Pre/post dose verification and documentation.
2–4 weeks
4
AI Validation (Optional)
For sites with deployed AI: post-optimization re-validation to confirm algorithm performance is maintained or improved at the new protocol settings.
3–4 weeks
Identify your highest-risk CT protocols in two weeks.
Protocol Risk Scan — no PHI required, no on-site visit required.
dan@gammametric.com
gammametric.com