Speaker
Description
Quantitative imaging performance analysis has recently been the focus in medical imaging, which provides objective information and it could aid a patient diagnosis by giving optimized system parameters for various imaging tasks [1]. In recent years, task-driven approach has been adopted in which NEQ is combined with the imaging task and model observers to form the detectability index (d’) [2]. In this study, we investigated mass detectability performance with different angular range settings between human and model observers. The purpose of this study is to validate our model observer performance on the newly developed prototype digital breast tomosynthesis (DBT).
The images of mass with different sizes of 3.9, 4.7 and 6.3 mm in the target slab of CIRS breast phantom were acquired with four different total angular range of ±10.5°, ±14°, ±21°, ±24.5°, with equal 15 projection views. Each DBT acquisition was repeated as twenty times in a same condition in order to yield the local ensemble of noise power spectrum (NPS). The human and the model observers were four-alternative forced choice (4AFC) and non-prewhitening with an eye filter (NPWE) in Fourier domain. For NPWE model observer, the ideal local modulation transfer function (MTF) was estimated by injecting a point signal in a simulation of the system. All human observers had a training session for each angular range dataset prior to 4AFC test. A percentage of correct responses (PC) was measured at the end of each human observer test [3].
As results, the local MTFs were same for each angular ranges, whereas the local NPS provided reasonable predication of increased noise with the increase of angular range distribution. The performance of the theoretical model observer values resulted in similar trend to the human observers’ PC results. The d’ results with different sizes of masses decreased with increasing angular distribution from 0.55 to 0.18, 1.01 to 0.45 and 1.51 to 0.74 with mass sizes of 3.9, 4.7 and 6.3 mm, respectively. In the human observer study, the average PC from seven observers were 0.87, ranging PC values from 0.71 to 0.92. The resulted patterns of PC decreased with increasing the angular ranges from ±10.5° to ±24.5° with different size of tasks.
Our results showed that the NPWE model could reasonably predict mass detectability from small to large sizes for different angular ranges. The correlation between theoretical and measured performance is necessary for better description of task-based model observer performance for future study.
