Katsumi Abe MD, PhD, Presenter: Nothing to Disclose

Yasuo Sasaki MD, Abstract Co-Author: Nothing to Disclose

Yoshiyuki Moriya, Abstract Co-Author: Nothing to Disclose

Ikue Tanaka, Abstract Co-Author: Nothing to Disclose

Motoichiro Takahashi MD, Abstract Co-Author: Nothing to Disclose

Mitsuhiro Narata, Abstract Co-Author: Nothing to Disclose

Yoshiaki Tanaka MD, Abstract Co-Author: Nothing to Disclose

Tsutomu Saito MD, Abstract Co-Author: Nothing to Disclose

Yoshitaka Okuhata, Abstract Co-Author: Nothing to Disclose

Akiko Takemoto MD, Abstract Co-Author: Nothing to Disclose

Motoaki Fujii, Abstract Co-Author: Nothing to Disclose

Tomoya Saito, Abstract Co-Author: Nothing to Disclose

Toshiya Maebayashi, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

Although accurate information on the thoracic-lumber bone structure is essential when CT images are observed using a diagnostic workstation, there is no automated method of bone labeling on a CT scan. We are developing a computer-aided diagnosis system that labels ribs and thoracic and lumbar vertebrae automatically, and have evaluated its diagnostic usefulness.

First, a candidate bone is extracted from the CT image volume data (slice thickness, 1mm) by pixel thresholding and connectivity analysis. Then, all non-bony anatomic structures are removed using a linear discriminate of distribution of CT values and anatomic characteristics. Thirdly, the vertebrae are separated from the ribs on the basis of their distances from the image center of the axial CT projection images. Finally, the thoracic cage and lumbar vertebrae are clearly extracted and each vertebral body is labeled with its own anatomical number by histogram analysis along the craniocaudal midline. The ribs are labeled in a similar way, based on location data. We used this technique in our diagnostic workstation to evaluate its clinical usefulness. Twenty-three CT cases (6 dynamic contrast-enhanced, 7 contrast-enhanced and 10 non-enhanced studies) were used to compare the accuracy and productivity of our method with those of the radiologist.

The automated labeling of the thoracic and lumbar vertebrae was concordant with the judgments of the radiologist in all cases, and all but the 1st and 2nd ribs were labeled correctly. The first two ribs were frequently misidentified, presumably because of pericostal anatomic clutter or high densities of contrast material in the injected veins.

Almost all ribs, and thoracic and lumbar vertebrae can be readily labeled automatically on axial CT images by our method. We strongly believe that this system can improve the workflow of radiologists, though further technical improvement is required for detection of the upper ribs.

Our newly developed fully automated bone labeling system for CT diagnosis offers valuable radiological information.

Abe, K, Sasaki, Y, Moriya, Y, Tanaka, I, Takahashi, M, Narata, M, Tanaka, Y, Saito, T, Okuhata, Y, Takemoto, A, Fujii, M, Saito, T, Maebayashi, T, et al, , et al, , Intelligent Assistant System for Diagnostic Workstations: Fully Automated Bone Labeling Useful for CT Diagnosis.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5001395.html