Abstract:
HTML
Purpose: While mammography is considered the gold standard in breast
screening, the use of ultrasound as an adjunct especially for the dense breast
is under the clinical trials. The aim of this study was to develop computerized
automatic detection method for the localization of lesions on three dimensional
breast ultrasound images.
Methods and Materials: Our algorithm is composed of preprocessing,
binarization, labeling, and postprocessing. And all the process is based on
three dimensional algorithm. In the preprocessing stage, order statistic
filtering is used for image enhancement. After that, automatic binarization is
applied. For the grouping and indexing, labeling is used consecutively. At
last, postprocessing such as reduction of labeled region with low possibility
of mass is used. We applied our computerized method to a database of 240 three
dimensional volume images composed of 80 malignant cases and 160 benign cases.
The tumor sizes ranged from 0.5 to 3.4 cm (mean, 1.3cm) in diameter. 3D volume
data were obtained by one breast radiologist using 10- or 12-MHz dedicated
volume transducer (Voluson 530D or 730, Kretz-GE). Image resolution of three
dimensional volume data was 184 x 184 x 184 in cases and 256 x 256 x 256 in
cases.
Results: The overall detection performance was a sensitivity of 92.1% (221/240)
for volume data at a false-positive rate of 6.2 (1480/240) per one volume data.
There was a different performance between malignant and benign cases. For
malignant lesions, a sensitivity was 96.3% (77/80) with false-positive rate of
5.3% (427/80). For benign lesions, a sensitivity was 90.0% (144/160) with
false-positive rate of 6.6 (1053/160). The required general processing time on
1.6GHz, Pentium PC was 15 seconds.
Conclusion: We have developed a novel lesion detection algorithm on three
dimensional breast ultrasound images. The sensitivity for detection of
malignant lesion by computer was 96.3%. (H.L., W.M., K.O. are stockholders in
Image Computing Lab., Co., Ltd.)
Questions about this event email: g1991177@inhavision.inha.ac.kr
Lee MS, H,
Computerized Automatic Detection of Breast Tumors on Three Dimensional Ultrasound Images. Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL.
http://archive.rsna.org/2003/3108415.html