Soft-tissue detectability in cone-beam CT: evaluation by 2AFC tests in relation to physical performance metrics

Area of Science:

• Medical Imaging
• Radiology
• Image Analysis

Background:

• Cone-beam computed tomography (CBCT) is increasingly used in medical imaging.
• Assessing soft-tissue detectability is crucial for optimizing CBCT imaging parameters.
• Understanding the impact of noise and prior knowledge is essential for accurate diagnosis.

Purpose of the Study:

• To evaluate soft-tissue detectability in CBCT using two-alternative forced-choice (2AFC) tests.
• To investigate the influence of radiation dose, noise-power spectrum (NPS), and prior knowledge on detectability.
• To establish detectability limits and their implications for clinical implementation.

Main Methods:

• Custom phantoms with soft-tissue-simulating spheres were imaged using an experimental CBCT system.
• 2AFC tests were conducted to assess the proportion of correct responses (Pcorr).
• Parameters varied included sphere size, contrast, radiation dose, visualization plane, and observer prior knowledge (SKE to SUK).

Main Results:

• Detectability was dose-dependent, with higher doses generally improving accuracy.
• Axial planes showed better detectability than sagittal planes, influenced by 3D NPS asymmetry and apodization filters.
• Prior knowledge significantly enhanced detectability; for example, a 6 mm sphere was detected with 70%-85% accuracy under stimulus known exactly (SKE) conditions versus 55%-65% under stimulus unknown (SUK) conditions.

Conclusions:

• Detectability limits can guide the development of minimum-dose CBCT techniques for specific tasks.
• Spatial-frequency characteristics of the 3D NPS affect detectability across different imaging planes.
• Lower radiation doses are sufficient for tasks with prior knowledge (e.g., image guidance) compared to diagnostic tasks without prior knowledge.