Preliminary assessment of the temporal subtraction of tomosynthesis images for improved detection of pulmonary nodules
Digital tomosynthesis is an imaging technique which reconstructs tomographic planes in an object from a set of projection images taken over a fixed angle [1]. Results from our initial pilot study show that tomosynthesis increases the detectability of lung nodules; while only 50% of CT confirmed nodules were found on typical chest radiographs, 81% were found on tomosynthesis image sets[2]. Temporal subtraction is a method which takes two sequential images and subtracts one from another, emphasizing the appearance of interval change[3-6]. As an addition to conventional chest radiography, it has been shown in several studies to significantly increase observer performance in detecting newly developed abnormalities[7-10]. Thus the combination of temporal subtraction and tomosynthesis may yield improved sensitivity of detection over either method alone. For this preliminary evaluation into the combination of these techniques, images were taken of an anthropomorphic chest phantom in different orientations and subtle lung nodules were simulated in order to emulate temporal discrepancies in anatomy. An automated method of segmentation, registration, and image warping was employed to align corresponding lung regions of each image set. The visibility of temporal change of simulated nodules was more apparent in the subtraction image. By our subjective analysis, tomosynthesis substantially improved the visibility of nodules relative to conventional chest radiography; and tomosynthesis augmented by temporal subtraction even further enhanced the conspicuity of difficultly placed subtle nodules.
