Atsuto Watanabe BS, Abstract Co-Author: Nothing to Disclose

Xiangrong Zhou PhD, Presenter: Nothing to Disclose

Takeshi Hara PhD, Abstract Co-Author: Nothing to Disclose

Ryujiro Yokoyama, Abstract Co-Author: Nothing to Disclose

Masayuki Kanematsu MD, Abstract Co-Author: Consultant, DAIICHI SANKYO Group

Hiroshi Fujita PhD, Abstract Co-Author: Nothing to Disclose

We presents a fast and robust segmentation scheme that automatically identifies and extracts a massive-organ region on CT images. In contrast to the conventional algorithms that are designed empirically for segmenting a specific organ based on traditional image processing techniques, the proposed scheme uses a fully data-driven approach to accomplish a universal solution for segmenting the different massive-organ regions on CT images by using three processing steps: machine-learning-based organ localization, content-based image (reference) retrieval, and atlas-based organ segmentation techniques.

We proposed a universal scheme for segmentation of different massive-organ regions automatically in 3D CT cases. This scheme was applied to the segmentations of heart, liver, spleen, left and right kidneys in 100 cases of CT cases, and its efficiency and accuracy were showed in experimetal results.

The segmentation results of five kinds of organs are compared with the ground truth that manually identified by a medical expert. The Jaccard similarity coefficient between the ground truth and automated segmentation result centered on 67% for heart, 81% for liver, 78% for spleen, 75% for left kidney, and 77% for right kidney.The experimental results showed that proposed apprach accomplished the segmentations of the different organs by using one algorithm within 16 seconds. We also confirmed that our method was very robust for both normal and abnormal CT cases.

A database that includes 100 cases of 3D volumetric CT cases was used in the experiment. These CT cases were generated by CT scanners of LightSpeed Ultra16 of GE Healthcare and Brilliance 64 of Philips Medical Systems. Each CT case used a common protocol (120 kV/Auto mA) and covered the entire human torso region. Each 3D CT case has approximately 800-1200 axial CT slices with an isotropic spatial resolution of approximately 0.625 mm and a density (CT number) resolution of 12 bits. All of these CT images were taken from patients with certain real or suspicious abnormalities.The heart, liver, spleen, left kidney, and right kidney are selected as the segmentation targets for performance evaluations.

Watanabe, A, Zhou, X, Hara, T, Yokoyama, R, Kanematsu, M, Fujita, H, Automated Organ Segmentation on CT Images by Using Similar Image Retrieval Based on Machine-Learning.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12031813.html