2.2 Data preparation for observer studies

To evaluate the impact from radiation dose, lesion size, lesion contrast, and reconstruction methods, 16 conditions were selected from 50 possible experimental conditions (5 diameters, 5 dose levels, and 2 reconstruction algorithms) for the observer studies, which are listed in Table 1. These 16 conditions can be categorized into 4 groups: (a) varying dose levels for FBP reconstruction (Conditions 1–5); (b) varying dose levels for IR reconstruction (Conditions 6–10); (c) varying lesion contrast (Conditions 11, 2, and 12); (4) varying lesion sizes (Conditions 13, 2, 14, 15, and 16).

Configurations for different effective mAs, reconstruction algorithm, object contrast and size in the 16 2AFC study.

A 3D volume of interest (VOI) surrounding each sphere was extracted (Figure 2). In the axial slice, the dimension of the VOI was approximately 3 times the diameter of each sphere. Along the longitudinal direction, the dimension of the VOI was equal to the diameter of each sphere. For each sphere, another 3D VOI was extracted from exactly the same in-plane location, but 2 cm away along the longitudinal direction. These VOI served as the signal-absent background.

Extraction of multi-slice images surrounding low-contrast objects with signal present and signal absent. The signal present images after ensemble average over 100 scans are also shown.

Signal present and absent VOIs were randomly placed side-by-side to create a 2-alternative forced choice (2AFC) trial. For each 2AFC trial, observers were challenged to decide which VOI contained the signal. Sixteen 2AFC studies were generated (Table 1), each with 100 trials. In total, 1600 trials were presented to both model and human observers.