Ryujiro Yokoyama MD, Presenter: Nothing to Disclose

Atsushi Matsui, Abstract Co-Author: Nothing to Disclose

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

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

Xiangrong Zhou PhD, Abstract Co-Author: Nothing to Disclose

Masayuki Kanematsu MD, Abstract Co-Author: Nothing to Disclose

Takahiko Asano, Abstract Co-Author: Nothing to Disclose

Satoshi Goshima MD, Abstract Co-Author: Nothing to Disclose

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

Toru Iwama, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

The purpose of our study was to develop an algorithm enabling automatic detection of lacunar infarct on T2-weighted MR images as a part of computer-aided detection (CAD) system of the brain, which can assist physicians to detect lacunar infarct and prevent the occurrence of cerebral apoplexy in high-risk patients.

Lacunar infarct regions were detected by multiple-phase binarization, using threshold values determined by the pixel values of cerebral ventricles. Thereafter, true-positive candidates were selected by three features (area, circularity, and gravity center) determined in accordance with the criterion of radiologic diagnosis. Lacunar infarct regions adjacent to hyperintense structures such as cerebral ventricles were discriminated, using opening processing and subtraction between images　reconstructed with two kinds of configuration elements. Tow methods were applied to eliminate false positive (FP) lesions from the detected candidates. One was to eliminate along the periphery of the brain by using region growing technique. And the other one was so-called multi-circular regions difference method (MCRDM), which was based on comparing the mean pixel values in lacunar infarct regions and peripheral regions in T1 weighted image. From the detected candidates of lacunar infarct on T2 weighted image, the peripheral radius R was determined by the area of a candidate, and the total pixel values within the circle was calculated as A. This procedure was repeated to generate a serial of total pixel values As by changing circles radius r one by one pixel within the region of R/2≦r＜Ｒ,and the FPs were eliminated if the maximum of A-As lower than a threshold value.

The sensitivity of the detection of lacunar infarct was 90% on 828 MR images from 100 patients with 1.37 false positives per image.

We found that this algorithm had a potential to automatically detect lacunar infarct on T2-weighted MR images.

Yokoyama, R, Matsui, A, Fujita, H, Hara, T, Zhou, X, Kanematsu, M, Asano, T, Goshima, S, Hoshi, H, Iwama, T, et al, , Development of an Automated Method for Detecting Lacunar Infarct Regions on Brain MR Images.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4407611.html