Premium
TH‐C‐330A‐08: Soft‐Tissue Detectability Limits in Cone‐Beam CT: 2AFC Tests of Human Observer Performance in Relation to Contrast, Spatial Resolution, and the 3D Noise‐Power Spectrum
Author(s) -
Siewerdsen J,
Tward D,
Daly M,
Paul N
Publication year - 2006
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2241850
Subject(s) - cone beam computed tomography , imaging phantom , sagittal plane , observer (physics) , apodization , optics , mathematics , medical imaging , physics , medicine , anatomy , radiology , computed tomography , quantum mechanics
Purpose: To quantify the contrast‐detail detectability limits of soft‐tissue structures in cone‐beam CT (CBCT) and to investigate the influence of the plane of visualization (axial/sagittal) and level of prior knowledge on observer performance. Method and Materials: Custom‐built cylindrical phantoms containing spherical lesions of varying size and contrast were imaged on a CBCT bench across a broad range of dose. Two‐alternative forced choice (2AFC) tests were conducted under controlled conditions using 7 observers (physicists and radiation therapists). For each 2AFC test, the proportion of correct responses, Pcorr, was analyzed as a function of lesion size (1.6 – 12.7mm) and contrast (20 – 165HU), dose (2.1 – 6.4mGy), plane of visualization (axial/sagittal), apodization filter (smooth Hanning to sharp Ram‐Lak), and degree of prior knowledge provided to the observer (ranging from Signal‐Known‐Exactly (SKE) to Signal‐Unknown (SUK)). Results: 2AFC analysis provided valuable quantitation of contrast‐detail detectability limits. For example, the lowest contrast lesion (20 HU) was detected at Pcorr>70% for diameters down to ∼6mm at doses >2mGy, but smaller 20 HU lesions (<3.2mm) were barely detectable (Pcorr<60%) at any dose. Detectability was significantly improved in axial versus sagittal planes, and the effect was amplified by sharper apodization filters in a manner consistent with 3D noise‐power spectrum asymmetry. Prior knowledge had a marked influence on detectability — e.g., a ∼6mm (20 HU) sphere was detected at Pcorr∼70–85% for SKE conditions, compared to Pcorr∼55–65% under SUK conditions across the same range of dose. Conclusion: Comprehensive human observer tests provide valuable quantitation of soft‐tissue detectability limits in CBCT and help to define low‐dose techniques for specific imaging tasks. Two factors in particular — plane of visualization and prior knowledge — hold significant practical implications: axial planes typically offer improved detectability, and performance is maintained at significantly lower dose under SKE conditions (e.g., lesion‐known image guidance) than in SUK conditions (lesion‐unknown diagnostic imaging).