Ryo Sakamoto MD,PhD, Presenter: Nothing to Disclose

Masahiro Yakami MD, PhD, Abstract Co-Author: Nothing to Disclose

Koji Fujimoto MD, PhD, Abstract Co-Author: Nothing to Disclose

Susumu Mori PhD, Abstract Co-Author: Research Consultant, AnatomyWorks LLC CEO, AnatomyWorks LLC

Michael I. Miller PhD, Abstract Co-Author: Consultant, Anatomyworks LLC

Kaori Togashi MD, PhD, Abstract Co-Author: Research Grant, Bayer AG Research Grant, DAIICHI SANKYO Group Research Grant, Eisai Co, Ltd Research Grant, FUJIFILM Holdings Corporation Research Grant, Nihon Medi-Physics Co, Ltd Research Grant, Shimadzu Corporation Research Grant, Toshiba Corporation Research Grant, Covidien AG

Keita Nakagomi MSc, Abstract Co-Author: Employee, Canon Inc

Gakuto Aoyama, Abstract Co-Author: Employee, Canon Inc

Can Ceritoglu, Abstract Co-Author: Nothing to Disclose

Takeshi Kubo MD, Abstract Co-Author: Nothing to Disclose

Yutaka Emoto MD, PhD, Abstract Co-Author: Nothing to Disclose

Hiroyuki Sekiguchi, Abstract Co-Author: Nothing to Disclose

Koji Sakai, Abstract Co-Author: Nothing to Disclose

Masami Kawagishi, Abstract Co-Author: Employee, Canon Inc

Yoshio Iizuka, Abstract Co-Author: Employee, Canon Inc

Hiroyuki Yamamoto, Abstract Co-Author: Employee, Canon Inc

To improve observer performance in detecting bone metastases in serial follow-up CT, we developed a temporal subtraction technique based on a non-rigid image registration algorithm called Large Deformation Diffeomorphic Metric Mapping (LDDMM).

With approval of the institutional review board, 60 cancer patients (prostate, 14; breast, 16; lung, 20; liver, 10) were recruited from our clinical database. All patients underwent torso CT scan at least twice between 2007 and 2013. For each patient, an image pair of two time points (previous and current) was used for the observer study. Bone metastases were consensually confirmed in 30 patients in current CT by referring all available clinical information and imaging data. The previous and current CT images were non-linearly registered by LDDMM, and the subtraction image was produced by subtracting previous image from current one. Three board-certified radiologists independently interpreted CT image pairs to identify emerged bone metastases without and with the temporal subtraction images, and marked their locations with confidence level for the diagnosis. The sensitivity and false positive rate for each condition were analysed to evaluate observer performance. Reading time was also recorded, and all observers were asked to rate the usefulness of subtraction image in a five point scale.

The bone skeletons of two time point CT were almost perfectly co-registered except the ventral extremity of ribs and scapulas. The subtraction image clearly visualized all bone metastases as temporal changes. The average sensitivity for detecting bone metastases was improved from 57.4 to 66.2% with temporal subtraction images when 50% confidence level was considered as a threshold, and the false positive rate was slightly increased from 0.17 to 0.21 lesions per case. The reading time was reduced for all readers from 500.0 to 336.1 second in average per case. All the observers recognize the advantage of the subtraction image, and the average usefulness rate was 4.6.

Temporal subtraction image obtained by LDDMM improved the accuracy and efficiency for detecting bone metastases in reading follow-up CT.

http://abstract.rsna.org/uploads/2014/14004098/14004098_8onf.jpg

Sakamoto, R, Yakami, M, Fujimoto, K, Mori, S, Miller, M, Togashi, K, Nakagomi, K, Aoyama, G, Ceritoglu, C, Kubo, T, Emoto, Y, Sekiguchi, H, Sakai, K, Kawagishi, M, Iizuka, Y, Yamamoto, H, Temporal Subtraction Images Derived by Large Deformation Diffeomorphic Metric Mapping Facilitate Identification of Bone Metastases in Follow-up CT.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14004098.html