Tomohiro Miyoshi MD, Presenter: Research Grant, Konica Minolta Group

Junji Yoshida MD, PhD, Abstract Co-Author: Research Grant, Konica Minolta Group

Keiju Aokage MD, PhD, Abstract Co-Author: Research Grant, Konica Minolta Group

Tomoyuki Hishida MD, PhD, Abstract Co-Author: Research Grant, Konica Minolta Group

Tsuyoshi Kobayashi MS, Abstract Co-Author: Nothing to Disclose

Shinichi Tsubura MS, Abstract Co-Author: Nothing to Disclose

Shinsuke Katsuhara MS, Abstract Co-Author: Nothing to Disclose

Satoshi Kasai PhD, Abstract Co-Author: Nothing to Disclose

Yasuhiko Sasano, Abstract Co-Author: Nothing to Disclose

Kanji Nagai MD, PhD, Abstract Co-Author: Research Grant, Konica Minolta Group

LROC analysis demonstrated that BSI can improve the detection accuracy and localization of small lung nodules on chest radiographs not only for experts but also observers with limited experience.

Bone suppression imaging (BSI) processes standard chest radiographs to improve the conspicuity of the lung field and the efficacy of small lung nodule detection. In this study, we evaluated the efficacy of a new BSI technique, using conventional ROC and localized ROC (LROC) analyses.

Standard and BSI chest radiographs of 80 patients, of which 40 with a nodular lesion (≤ 30 mm) and 40 without any nodules, were randomly interpreted by 20 observers, including 7 pulmonologists with more than 10-year experiences and 13 pulmonology residents. They recorded the changes in confidence levels regarding the presence or absence of a lung nodule and also the most likely location of the lesion, first without and then with BSI. ROC and LROC analyses were used to evaluate observer performance. The average area under the ROC curve (AUC) for all observers was significantly improved from 0.867 to 0.900 (p=0.004) with BSI. The average AUC for experienced pulmonologist and resident groups were also improved from 0.900 to 0.929 (p=0.035) and from 0.848 to 0.885 (p=0.031), respectively. In LROC analysis, the AUC was also significantly improved in each group. The AUC without BSI were generally greater among experienced pulmonologists compared to residents (0.80±0.03 and 0.72±0.09, respectively), whereas the AUC improvement was reversed (range: -0.03 to 0.18 for residents and -0.01 to 0.08 for experienced pulmonologists, respectively). Among the 13 residents, only one scored a worse AUC with BSI. BSI enabled more than 3 observers to localize nodules correctly in 5 of the 80 patients.

BSI improved the accuracy of small nodule detection on chest radiographs regardless of observers’ experience. LROC analysis showed that BSI improved nodule localization and tended to be more effective in observers with limited experience.

Miyoshi, T, Yoshida, J, Aokage, K, Hishida, T, Kobayashi, T, Tsubura, S, Katsuhara, S, Kasai, S, Sasano, Y, Nagai, K, Novel Bone Suppression Imaging Technique in Small Lung Nodule Detection: Evaluation using Localized ROC Method.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14010468.html