I N F O R M A T I O N

Analysis Guide

Step-by-step instructions for every analysis available in QAF — from image uniformity and artefact assessment through spectral and material analysis, to scheduling and tracking a full quality control programme.

Software Analyses.

Core image quality assessment modules

01 Image Uniformity −
Terminology

Image Uniformity verifies that CT numbers stay consistent across the scan field of view using a uniform phantom. Five regions are measured — one at the phantom centre and four at the periphery (12, 3, 6 and 9 o'clock) — and the largest difference between any periphery region and the centre is compared against a tolerance.

How to Perform
  1. Select Analysis: Navigate to Analysis → Image Uniformity and upload a uniform phantom series.
  2. Image type: Choose the image type being tested. This sets the appropriate tolerance; photon-counting energy-bin images have no published limit, so you set your own.
  3. Hounsfield Units: Tick Convert images to HU — the criteria are defined in HU. If no calibration exists you are guided through a quick water and air ROI calibration inside the module.
  4. Tolerances: Optionally override the default globally, or set a separate tolerance for each energy bin.
  5. Place ROIs — automatic: Draw one circle tracing the entire phantom body, then click Auto-place. The five regions are positioned and sized (300–400 mm²) for you, keeping clear of the phantom edge.
  6. Place ROIs — manual: Alternatively place each region yourself in uniform areas, avoiding inserts, in the order Centre, 12, 3, 6, 9 o'clock.
  7. Measure: All images and energy bins are analysed, with PASS/FAIL per bin and an overall verdict.
  8. Review and save: Expand the per-ROI diagnostics to see individual values, then export to Excel or save to a QC record.
Tip: If the centre reads far from 0 HU in a water phantom, re-check the water and air calibration. If all four periphery values differ from the centre in the same direction, the pattern indicates beam hardening rather than a placement problem.
02 Artefact Assessment +
Terminology

Artefact Assessment is a structured visual check of uniform phantom images for ring artefacts, streaks, shading or cupping, and blooming. The criterion is that no artefacts are visible; the recommended frequency is daily.

How to Perform
  1. Select Analysis: Navigate to Analysis → Artefact Assessment and load a uniform phantom series.
  2. Review each image type at an optimised window and level, stepping through the slices.
  3. Record findings: Name the image type or energy bin reviewed, tick any artefacts observed, add notes, and save the record. Reviewer and timestamp are stored automatically.
  4. Result: A record with no artefacts passes; any artefact ticked fails. The overall result appears with the record list.
  5. Export or save the assessment log to Excel or to a QC record.
Tip: Artefacts are usually most visible with the widest collimation and thinnest slice setting. Ring artefacts in particular are the classic photon-counting failure mode, caused by detector threshold drift.
03 Contrast-to-Noise Ratio (CNR) +
Terminology

CNR expresses how well a feature stands out from its surroundings relative to the background noise. It underpins low-contrast detectability: the higher the CNR, the easier a lesion of that contrast is to see.

How to Perform
  1. Select Analysis: Navigate to Analysis → CNR and upload your images.
  2. Acceptance criteria (optional): Enter a baseline CNR with a tolerance percentage and/or a minimum CNR. Entering 1 as the minimum applies the common low-contrast criterion.
  3. Background ROI: Draw and set a background region first — this provides the reference level and the noise.
  4. Signal ROIs: Add one or more ROIs on the targets you want to evaluate.
  5. Measure: CNR is calculated per energy bin across all images, with PASS/FAIL against your criteria.
  6. Plot, export, or save the result to a QC record.
Tip: Comparisons are only meaningful when dose and reconstruction settings match between runs. If the background region is perfectly uniform, CNR is reported as undefined rather than zero.
04 Spectral Linearity +
Terminology

Spectral Linearity measures how linearly a material's response changes across different energy bins or concentrations. It evaluates the consistency of material properties (HU or linear attenuation coefficients) across energy levels in spectral CT imaging.

How to Perform
  1. Select Analysis Type: Navigate to Software → Spectral Linearity. Choose HU or Linear Attenuation (cm⁻¹) scale.
  2. Upload Images: Upload DICOM images containing your material samples.
  3. Material Setup: Select the material (e.g., Iodine, Gadolinium) and set up concentrations.
  4. ROI Selection: For HU: measure air/water references first. Draw ROIs on material samples at different concentrations.
  5. Analysis Scope: Choose Across Concentration or Across Energy Bin.
  6. Generate Plot: Create linearity plots to visualize the relationship.
Tip: Ensure consistent ROI placement and sufficient sample sizes for accurate linearity assessment.
05 Signal-to-Noise Ratio (SNR) +
Terminology

SNR quantifies the ratio of signal strength to noise level in an image. Higher SNR indicates better image quality with less noise relative to signal.

How to Perform
  1. Select Analysis: Navigate to Software → Signal-to-Noise Ratio.
  2. Upload Images: Upload DICOM images with the regions you want to analyze.
  3. SNR Setup: Select energy bin and analysis mode (single image or all-images averaged).
  4. Acceptance criteria (optional): Enter a baseline SNR with a tolerance percentage, and/or a minimum SNR. Leave blank to report values only. You can also load a baseline automatically from a scanner's saved QC records.
  5. ROI Selection: Draw ROIs on homogeneous regions (signal) and background areas (noise). Ensure ROIs > 100 pixels.
  6. Calculate: The system calculates SNR for each ROI and energy bin, with PASS/FAIL against your criteria.
  7. Save: Store the result permanently against a scanner via Save to QC record.
Tip: Use homogeneous regions for signal measurement. Avoid edges or artifacts that could affect noise calculations. When the ROI is perfectly uniform (standard deviation zero), SNR is reported as undefined rather than zero.
06 Noise Power Spectrum (NPS) +
Terminology

NPS characterizes the spatial frequency content of noise, describing how noise power is distributed across spatial frequencies using FFT of noise data from uniform ROIs.

How to Perform
  1. Select Type: Choose 1D Radial, 2D Planar, or 3D volumetric NPS.
  2. Upload Images: Upload DICOM images with uniform phantom regions (water phantoms).
  3. Set Voxel Size: Enter physical voxel size in mm — critical for accurate frequency scaling.
  4. Add ROIs: Draw rectangular ROIs on uniform regions (at least 64×64 pixels). Multiple ROIs improve statistics.
  5. Acceptance criteria (optional): Enter baseline peak and average noise frequencies with a tolerance percentage, or load them from a scanner's saved records.
  6. Calculate: Run NPS for all ROIs and images.
  7. View Results: 1D plots, 2D heatmaps, or 3D surface plots per energy bin, plus a summary table of peak and average noise frequency with PASS/FAIL.
  8. Save: Store the result against a scanner via Save to QC record.
Tip: Use uniform water phantoms. Avoid artifacts and non-uniformities. Correct voxel size is essential for frequency accuracy. A shift in the summary frequencies usually indicates that reconstruction or post-processing settings have changed.
07 Modulation Transfer Function (MTF) +
Terminology

MTF measures spatial resolution by quantifying contrast preservation at different spatial frequencies. Derived from PSF (point/wire sources) or LSF (edge spread function differentiated, then FFT).

How to Perform
  1. Upload Images: Upload DICOM images with high-contrast edges, thin wires, or point structures.
  2. Set Voxel Size: Enter physical voxel size for spatial frequency calculations.
  3. Signal ROI: Draw a circle (wire/point) or rectangle/line (edges) on the high-contrast feature.
  4. Background ROI (optional): Draw on a uniform background for subtraction.
  5. Acceptance criteria (optional): Enter baseline resolution values with a tolerance percentage, or load them from a scanner's saved records.
  6. Calculate: System calculates FWHM and the spatial frequencies at 50%, 20% and 10% modulation per energy bin, with PASS/FAIL against your baseline.
  7. Save: Store the result against a scanner via Save to QC record.
Tip: Circle ROI is best for wire/point sources; rectangle or line for edges. Ensure the feature is well-centered. Resolution criteria are baseline-relative — establish reference values at acceptance and monitor drift against them.
08 Electron Density & Effective Atomic Number +
Terminology

Relative electron density and effective atomic number describe how many electrons a material contains relative to water and how strongly it interacts with X-rays. They matter for particle-therapy and brachytherapy dose calculation, for characterising kidney stones, and for improving Monte Carlo dose accuracy — so their uncertainty should be understood before clinical use.

How to Perform
  1. Select Analysis: Navigate to Analysis → ρe / Zeff and load the corresponding image maps.
  2. Image type: Choose whether you are measuring relative electron density or effective atomic number.
  3. Acceptance criteria (optional): Set a tolerance as a percentage of the reference value and/or an absolute tolerance. No universal limits are published — your physicist establishes them at acceptance from the clinical application or the manufacturer's specification.
  4. For each tissue-surrogate insert: enter its name and known reference value, and its diameter in millimetres.
  5. Size the region: use the automatic sizing button — the region is set to about 60% of the insert diameter so it samples only insert material and never the boundary.
  6. Place and add: click the centre of the insert, then add it. Repeat for every insert.
  7. Measure All: each insert is averaged across all loaded images and compared with its reference value, reporting the error and PASS/FAIL.
  8. Export or save to Excel or to a QC record.
Tip: Keeping the region well inside the insert is the single most important step — including any of the boundary blends in surrounding material and biases the result. Testing frequency should follow the clinical application or the manufacturer's guidance.

QC Programme.

Scheduling, records, baselines and trending

09 QC Scheduling & Records +
Terminology

A quality control programme is more than measurement: tests must be performed at defined intervals, judged against criteria, and recorded permanently. The QC Schedule page turns individual analyses into a tracked programme per scanner.

How to Use
  1. Add a scanner: record its name, model, serial number and location. Scanners can be edited or removed later; saved records are always retained.
  2. Schedule tests: choose a scanner, a test, and a frequency. Sensible defaults are offered — artefact and noise checks daily, most other tests annually and after relevant service.
  3. Status board: every scheduled test shows its last run, last result, next due date, and a status of Done, Due, Overdue or Pending. A Run button takes you straight to the right analysis module.
  4. Save results: after any analysis, choose the scanner and click Save to QC record. The result, criteria used, operator and timestamp are stored permanently and the board updates.
  5. Mark a baseline: star the record measured at acceptance. Analysis modules can then load it automatically as their reference.
  6. Review history: recent records are listed with their results and can be removed by the person who created them, or by an administrator.
Tip: Establish baselines only once the scanner is properly configured. Recording a fixable defect as normal will hide it from every future comparison.
10 Baselines, Tolerances & Trending +
Terminology

Some tests have published absolute limits, but for noise, resolution and noise texture the criteria are baseline-relative: you record the performance of a known-good scanner at acceptance and flag later deviation from it. A tolerance is the allowed deviation before a test is judged to fail.

How to Use
  1. First run: leave the criteria fields blank. Results are reported without a verdict and marked as having no criteria set.
  2. Establish the baseline: save that run as a QC record and star it as the acceptance baseline.
  3. Later runs: select the scanner in the Load baseline box and the criteria fields fill in automatically, showing where the values came from.
  4. Tolerance types: a percentage deviation from the baseline, an absolute minimum or maximum, or both together. Uniformity additionally supports a separate tolerance for each energy bin.
  5. Trending: on the QC Schedule page, choose a scanner and test to plot the metric over time, one line per energy bin, with the tolerance band shaded and failures marked in red.
Tip: Baseline comparisons are only valid when acquisition and reconstruction settings match between runs. A change in dose, slice thickness or kernel will shift the metric for reasons unrelated to scanner performance.
11 Automatic Phantom Data Reception +
Terminology

Instead of exporting and uploading files by hand, QAF can receive phantom images directly from the scanner console over the network, matching each study to the correct scanner automatically.

How to Use
  1. Start the receiver on the server; your administrator configures the port and name it listens under.
  2. Configure the console: add the receiver as a secondary destination on the scanner, then send the phantom series after each QC scan.
  3. Automatic matching: studies are grouped and matched to a registered scanner by serial number or station name.
  4. Review: received datasets appear on the QC Schedule page with their study description and image count.
  5. Load: click Load to place a dataset into your session, then open any analysis module — the images are already there.
Safety: studies that do not identify themselves as phantom or QC data are rejected and never written to disk, so patient examinations sent by mistake cannot be stored. Rejections are logged for review.

MIQ Analyses.

Material identification and quantification (Advanced tier)

12 Material Identification +

Material Identification identifies unknown materials by comparing density map predictions with ground truth energy images. Calculates accuracy, precision, recall, and F1-score.

  1. Load your energy images and the corresponding density maps
  2. Navigate to MIQ → Material Identification
  3. Select energy bin and target material
  4. Draw target and background ROIs
  5. Set density threshold and perform identification
Note: Requires density maps for the materials of interest and an Advanced tier subscription.
13 Material Quantification +

Material Quantification determines concentration of specific materials using calibration data from known samples. Builds a calibration matrix from ROI measurements.

  1. Navigate to MIQ → Material Quantification
  2. Upload DICOM images and define materials
  3. Measure calibration ROIs for each material/concentration
  4. System builds calibration matrix automatically
  5. Draw ROIs on unknown regions to quantify
Note: Accurate quantification depends on proper calibration with known material samples.

Tools Analyses.

Image processing utilities

14 ROI Analysis +

ROI Analysis calculates statistical metrics (mean, std, min, max) for user-selected regions across multiple energy bins.

  1. Navigate to Tools → ROI Analysis
  2. Upload DICOM images and select slice/energy bin
  3. Toggle HU conversion if desired
  4. Draw circular or rectangular ROI on the image
  5. Click Calculate to get statistics for all energy bins
Tip: Larger ROIs provide better statistical significance. HU conversion requires air/water calibration.
15 Histogram Analysis +

Histogram Analysis generates frequency distributions of pixel values in selected ROIs — single image or averaged across all slices.

  1. Navigate to Tools → Histogram Analysis
  2. Upload DICOM images and draw ROIs
  3. Set scope (selected image or all images) and bin count
  4. Generate histogram and review interactive plot
  5. Export to Excel or CSV
Tip: Multiple ROIs provide better statistical representation. "All Images" scope averages across slices.
16 Line Profile Analysis +

Line Profile extracts intensity values along drawn lines, creating 1D profiles for analyzing edges, boundaries, and spatial gradients.

  1. Navigate to Tools → Line Profile Analysis
  2. Upload DICOM images and select slice/energy bin
  3. Draw lines on the image (multiple lines to compare)
  4. Adjust line width for smoother profiles
  5. View interactive intensity vs. distance plot
Tip: Wider line widths provide smoother profiles by averaging across perpendicular pixels.
17 HU Conversion +

HU Conversion converts raw attenuation values to the standardized Hounsfield Unit scale (water = 0 HU, air = −1000 HU) using air/water calibration.

  1. Navigate to Tools → HU Conversion
  2. Upload DICOM images
  3. Draw ROIs on air and water regions, measure mean values
  4. System calculates conversion parameters automatically
  5. Toggle "Use HU" to view images in HU scale
  6. Download converted DICOM or PNG files
Note: HU values are energy-dependent in spectral CT. Calibrate per energy bin for best accuracy.

Ready to Learn More?

Create a free account to access all analysis modules and start working with your spectral CT data.

Get Started ↗
© 2025 QAF — Khalifa University of Science & Technology. All Rights Reserved.