From Phantom to Phantom-Free: "Anatomiq" an Automated CT Quality Assessment without Physical Calibration
Abstract
Purpose
Manual background selection for contrast-to-noise ratio (CNR) calculations in CT image quality assessment is time-consuming, operator-dependent, and introduces >15% measurement variability that compromises reproducibility. Advanced metrics such as Noise Power Spectrum (NPS) and Task Transfer Function (TTF) traditionally require dedicated phantom acquisitions. We developed AnatomIQ, an open-source Python toolkit which automates background detection while performing phantom-free NPS and TTF analysis in vivo, eliminating operator variability and phantom requirements.
Methods
The toolkit implements optimal automated background detection using tissue-specific Hounsfield unit thresholding, dual spatial exclusion zones (5mm inner, 15mm outer), and morphological filtering. A web interface enables click-to-select lesion identification from PACS screenshots with automatic coordinate extraction. Metrics include CNR, signal-to-noise ratio, detectability indices including Rose detectability index (CNR×√Area) for lesion conspicuity assessment, and dose-normalized figures of merit with automated optimization recommendations. NPS analysis quantifies noise texture using uniform tissue regions from automated detection. TTF leverages circular anatomical structures which present actual edge-transitions appropriate for local spatial resolution characterization.
Results
AnatomIQ reduced analysis time to 2-3 minutes per case with <10-second CNR computation. Optional NPS (5-15 seconds) and TTF (10-25 seconds) calculations are available on-demand. Automated detection successfully identified valid regions meeting size (≥100 mm2) and homogeneity (CV < 0.3) criteria across anatomical sites. Phantom-free NPS and TTF analysis successfully utilized in vivo structures (e.g., vessels, trachea), enabling spatial frequency characterization without need for additional radiation or phantom scans.
Conclusion
AnatomIQ streamlines CT image quality analysis by automating background detection and enabling phantom-free spatial frequency analysis from patient scans. By eliminating operator variability and phantom requirements, AnatomIQ enables practical continuous quality monitoring in vivo with evidence-based precision. Open-source availability facilitates academic research and clinical deployment.