Assessing CT acquisition parameters with visual-search model observers.

Area of Science:

• Medical Imaging
• Computational Imaging
• Observer Performance Modeling

Background:

• Channelized Hotelling (CH) and visual-search (VS) model observers exhibit distinct responses to image noise texture.
• Understanding these differences is crucial for optimizing imaging parameters in lesion detection tasks.

Purpose of the Study:

• To compare CH and VS model observers in lesion-detection studies.
• To evaluate the impact of linear and angular sampling parameters on computed tomography (CT) imaging.
• To assess the effectiveness of a modified VS observer for evaluating noise texture effects.

Main Methods:

• Simulated lung images with circular lesions were generated using a 2D torso phantom.
• Projection datasets were created with varying detector pixel sizes and projection numbers (15-128) at different photon counts.
• Filtered backprojection reconstructions were performed, and a localization receiver operating characteristic study was conducted with human and various VS observers.

Main Results:

• The effects of sampling parameters on lesion detection performance were consistent across human observers and different VS observer types.
• The CH observer, in location-known studies, showed no sensitivity to variations in angular sampling.
• The two-stage VS framework proved effective in modifying the CH observer for assessing noise texture's impact on human performance.

Conclusions:

• The study highlights the differing sensitivities of CH and VS observers to image noise texture.
• VS model observers, particularly the modified two-stage framework, are valuable tools for evaluating the influence of noise texture on human performance in CT lesion detection.
• Findings inform the optimization of sampling parameters in CT imaging for improved lesion detection.