Takaya Suzuki, Presenter: Nothing to Disclose

Yoshikazu Uchiyama PhD, Abstract Co-Author: Nothing to Disclose

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

Hiroaki Hoshi MD, Abstract Co-Author: Nothing to Disclose

Min Suk Chung MD, Abstract Co-Author: Nothing to Disclose

Hiroshi Fujita PhD, Abstract Co-Author: Research grant, Hitachi, Ltd

Toru Iwama, Abstract Co-Author: Nothing to Disclose

The new viewing technique would be useful in assisting radiologists in detecting unruptured aneurysms in MRA images.

The incidence of subarachnoid hemorrhage (SAH) is increasing every year in Japan. Therefore, a screening system for early detection of cerebrovascular diseases is widely performed. In the screening system, the detection of unruptured aneurysm is important because aneurysm rupture is the main cause of SAH. However, it is often difficult for radiologists to detect small aneurysms in MRA images because of the overlap between aneurysms and the adjacent arteries in MIP image. In facilitating the radiologists in detecting small aneurysms, we developed SelMIP image as a new viewing technique in our CAD scheme. The technique takes on an approach in making a new type of MIP image with the interested vessel only manually selecting a desired cerebral artery from a list. By using SelMIP image, the selected arteries can be observed from various directions, and small aneurysms would be easy to detect. For making SelMIP images, we developed a new method for automated labeling of eight arteries in MRA studies. Visible Korean Human image was used as a reference image. Image registration was performed on the 3D reference image and an image to be classified.

Our large image database consists of 859 MRA studies. These images were acquired on a 1.5 T MR scanner. To evaluate the classification performance, we employed the subjective rating using a three-point scale as described in the following. (1) Poor : most of the cerebral arteries are not well classified; (2) Adequate : most of the cerebral arteries are well classified but with some minor classification errors; and (3) Good : all cerebral arteries are perfectly classified. The results of the subjective evaluation were that 691 cases were rated ‘good’, 47 cases were rated ‘adequate’, and 121 cases were rated ‘poor’.

Overall, the segmentation and classification of cerebral arteries in 85.9% (738/859) of the MRA studies attained clinically acceptable results. Therefore, our computerized method would be useful for the segmentation and classification of cerebral arteries.

Suzuki, T, Uchiyama, Y, Hara, T, Hoshi, H, Chung, M, Fujita, H, Iwama, T, Automated Classification of Cerebral Arteries Using Visible Korean Human Image and Its Application to CAD Scheme for Detecting Unruptured Aneurysms.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11013706.html