SESSION TITLE: Imaging
SESSION TYPE: Original Investigation Slide
PRESENTED ON: Sunday, October 27, 2013 at 10:45 AM - 11:45 AM
PURPOSE: To compare CAD as concurrent reader (cCAD) and CAD as second reader (sCAD) for lung nodules on CT in terms of reading time and observer performance when using prototype CAD software.
METHODS: Thirteen observers consisting of 7 radiologists (referred to as group A) and 6 pulmonologists (group B) independently read 60 standard-dose 1-mm-thick CT datasets three times to detect and rate non-calcified nodules (≥ 5 mm). In the 1st (then 4 weeks later, 2nd) reading session, 30 datasets and the remainder were read with cCAD (then sCAD) and with sCAD (then cCAD), respectively. In the third session held 2 months afterwards, all datasets were read without CAD. Later 2 board certified chest radiologists determined a gold standard of nodules by consensus. In each group, the reading times with cCAD and sCAD were compared using analysis of variance, and the observer performances with and without CAD were compared using jackknife free-response receiver operating characteristic analysis.
RESULTS: The difference between the reading time with cCAD (mean, 172 s and 94 s in groups A and B) and that with sCAD (mean, 221 s and 146 s in groups A and B) was highly significant in both groups (p<0.0001). The mean figures of merit (FOMs) without CAD, with cCAD, and with sCAD were: in group A, 0.718, 0.762, and 0.757; in group B, 0.683, 0.749, and 0.735. The FOMs with cCAD as well as those with sCAD were significantly higher than those without CAD in both groups (p<0.01). There was no significant difference between the FOMs with cCAD and those with sCAD (p=0.66 in group A; p=0.22 in group B).
CONCLUSIONS: In both radiologists and pulmonologists, CAD as concurrent reader and CAD as second reader were about equally effective in improving the observer performance for lung nodule detection on CT, with the former requiring much shorter reading time.
CLINICAL IMPLICATIONS: CAD can assist pulmonologists as well as radiologists to detect lung nodules. cCAD enables both of them to reduce their reading time.
DISCLOSURE: Sumiaki Matsumoto: Grant monies (from industry related sources): Dr. Matsumoto has a research grant from Toshiba Medical Systems Corporation. Takatoshi Aoki: Grant monies (from industry related sources): Dr. Aoki has a research grant from Toshiba Medical Systems Corporation. Tae Iwasawa: Grant monies (from industry related sources): Dr.Iwasawa have a research grant from Toshiba Medical Systems Corporation. Hitoshi Yamagata: Employee: Dr. Yamagata is an employee of Toshiba Medical Systems Corporation. Yoshiharu Ohno: Grant monies (from industry related sources): Dr. Ohno has a research grant from Toshiba Medical Systems Corporation. The following authors have nothing to disclose: Shuntaro Tokunaga, Nobuko Hazeki, Daisuke Tamura, Tatsuya Nagano, Kazuyuki Kobayashi, Yoshihiro Nishimura
No Product/Research Disclosure Information