Pulmonary Nodule Size Evaluation With Chest Tomosynthesis.
Pulmonary nodule size evaluation with chest tomosynthesis
The 80 blinded DT datasets without true-positive marks and clinical information were divided into four groups of 20 cases each for the image review, and they were randomly mixed and presented in the same reading order for the six readers. The four groups were evaluated independently by the six readers during four 1-day sessions performed within 3 weeks. The readers were permitted to adjust window width and level. The readers were requested to mark nodules using an arrow and record the confidence level. Confidence for the presence of each marked nodule was graded on a 4-point scale, in which a score of 4 represented "a definite nodule", and a score of 1 represented "probably not a nodule" (). Each nodule was marked on only one representative section image. Readers were instructed to ignore obvious calcified granulomas, postoperative scars, atelectases, and extrapulmonary lesions. The readers were also instructed to record their reading time for each session. At the end of each session, readers were unblinded to the true-positive markings on the DT images and instructed to self-review and analyze their own misreads for each case. Coronal reconstructed chest CT images were also provided so that readers could check the false-positive or false-negative findings, if necessary. After being unblinded and reanalyzing their images, the readers were allowed to advance to the next session until all 80 cases were done. Finally, an unblinded researcher gathered all marks and confidence levels and recorded all data on Excel spreadsheets for analysis. If any false-positive or false-negative results were observed in any reader's interpretation, reasons for the misinterpretation were also assessed through a review of the CT and DT images.
pulmonary nodule size measurements on chest tomosynthesis images
We analyzed nodule visibility on DT based on a sophisticated review of thin-section chest CT images. James et al. () reported that 70% of nodules ≥ 3 mm in diameter are visible on DT, and Vikgren et al. () concluded that 92% of all nodules are visible on DT. However, neither study investigated the reason for the inability to visualize nodules on DT and did not correlate their findings with pathological data. Therefore, we also examined reasons for invisibility in addition to assessing a visibility score, and matched each nodule with a pathological result. According to our data, 53.3% (221/414) of nodules were visualized on DT. We identified all visible nodules on 1 or 1.25 mm section thickness chest CT images, with the aid of CAD, as a CT nodule reference standard. Therefore, a larger number and smaller size of nodules may have been detected on CT in our study (, , ). This result explains the lower proportion of visible nodules on DT in our data, when compared with previous studies. As most nodules > 5 mm were visible on DT and a large proportion of these nodules were malignant, DT could be a suitable alternative imaging method for detecting and following pulmonary nodules.