Technical note: A proposal of a novel figure-of-merit defined from task-based detectability index in abdominal...

Figures of Merit (FOM) based on signal-difference-to-noise ratio (FOM_SDNR) are commonly used to determine optimal technical parameters in digital radiography.However, this does not evaluate image quality for a specific clinical task nor account for spatial frequency response.

To evaluate a new FOM based on the detectability index (FOM_d') using the International Atomic Energy Agency (IAEA) methodology. To compare the optimum parameters obtained from FOM_d' and FOM_SDNR on an anthropomorphic test-object.

Several settings were evaluated involving six additional filtrations from none to 1 mm Cu, two tube voltages (80 and 110 kV), and four incident air kerma at the detector(1, 2, 3, 4µGy). FOM_d' and FOM_SDNR were determined on a Clisis Exel DRF unit for the IAEA test-object. The maximum FOM values determined the optimum settings. They were used to image an anthropomorphic test-object modified with the presence of a lesion, either in homogeneous or heterogeneous backgrounds. Three radiologists scored the images with a 5-point Likert scale. A Mann-Whitney test was used to compare overall image scoring for both backgrounds.

The highest FOM values were at: 80 kV, 2µGy, 1 mm Cu for FOM_SDNR and 80 kV, 1µGy, and 0.3 mm Cu+2 mm Al for FOM_d'. Overall image scorings were statistically higher for FOM_d' with homogenous background: 4.6±0.5 vs 3.7±1.0 (p = 0.013) and statistically similar for FOMs with heterogeneous background: 3.5±1.3 vs 3.9±0.8 (p = 0.540).

A novel FOM based on d' was introduced and compared with FOM_SDNR, giving better and similar results on homogeneous and heterogeneous backgrounds, respectively. Since the d' has the advantage of selecting a size-specific lesion for detection, the FOM_d' opens the perspective for task-based optimization in digital radiography using the IAEA methodology.