Authors:
N. Takahashi1, Y. Lee2, D.-Y. Tsai2, S. Kamio1; 1Sendai/JP, 2Niigata/JP
DOI:
10.1594/ECR05/C-0372
Results
The effect of threshold value determination on SD of the composite images is calculated and the results are shown in Fig.8 . The SD value decreases with the increase of threshold value and maintains constant when the threshold value is higher than 3.0. Since low SD value means low noise level, the noise reduction can be effectively achieved if the threshold value is set at T=3.0. In this case, the SD value of the original image is 2.5 and that of the processed image is 0.6. As a result, a 76% of the noise reduction can be obtained.
Fig. 9 illustrates the variation of the SR with the threshold value used in the APSF for LN-contained composite image. The SR is at approximately 60% for the threshold values ranging from 0 to 3.0 and gradually declines when the threshold value is greater than 3.0. Recall that low SR value shows high edge blurring. Therefore, the edge blurring of the LN-contained composite image can be suppressed if the threshold value is set at 3.0 or less. From the two experimental results, it is reasonable to conclude that T=3.0 is the optimal parameter used in the APSF for image improvement in the noise.
The APSF with the pre-determined optimal condition of thresholding T=3.0 was applied to clinical CT images with the MCA infarction. Fig.10 and Fig. 11 show two examples illustrating the original and processed images. From the results, obviously, the visibility and detectability of the lentiform nucleus were much improved by our proposed the APSF, moreover, another early sign (loss of the cortical ribbon)[3] was also enhanced.