Desktop version

Home arrow Health arrow Computational Anatomy Based on Whole Body Imaging: Basic Principles of Computer-Assisted Diagnosis and Therapy

Source

Experimental Results

We demonstrate the effectiveness of our image registration algorithms on 31 datasets of chest MDCT images with early and later scans of each subject. The difference in time between the previous images and current images was in the range of 38 months.

Fig. 4.9 Search for local matching

illustrates the results of the use of our temporal subtraction method

Figure 4.10 illustrates the results of the use of our temporal subtraction method. Figure 4.10a, b show the previous and current CT images with the lung nodule, respectively. Figure 4.10c-e show the subtraction images based on the global matching technique (using cross correlation value), local image matching (using 3D elastic matching), and with the voxel matching technique as shown in Fig. 4.8. The temporal subtraction enhances the differences between the previous and the current examination, caused by new or changed abnormalities. It is obvious that the majority of the subtraction artifacts in Fig. 4.10c, d are removed in Fig.4.10e, with a clean and smooth background.

Discussion and Conclusion

In this subsection, we proposed a nonrigid image registration algorithm for temporal subtraction of chest MDCT images using the voxel matching technique. Satisfactory generation of a temporal subtraction image was achieved. To evaluate the usefulness of the voxel matching technique for removal of subtraction artifacts, we performed our new technique on clinical examinations without and with voxel matching. With our new method, subtraction artifacts due to normal structures such as blood vessels were substantially removed on temporal subtraction images. This computerized method can enhance lung nodules on chest MDCT images without the disturbance of misregistration artifacts. Figure 4.11 illustrates an example of generation of a temporal subtraction image. Figure 4.11a shows the results without the voxel matching technique. Figure 4.11b, c show registered images using commercial software and our voxel matching technique, respectively. A brain tumor (arrow) is enhanced on all images. However, subtraction artifacts still remain in (a) and (b). We think this CAD system may be a useful tool in screening examinations.

Experimental results [124]

Fig. 4.10 Experimental results [124]: (a) and (b) show the original previous and current MDCT images, respectively. (c) illustrates the temporal subtraction image based on global image matching based on the 2D cross correlation technique. (d) shows the temporal subtraction image based on local image matching using 3D elastic matching. (e) shows the temporal subtraction image based on 3D nonlinear image warping using 3D voxel matching. In figure (e), most normal structures such as airway and vessels are cleanly removed. The circle in each image shows the lung nodule (Previous; 6.7 mm diameter, Current; 10.8 mm diameter)

Temporal subtraction images from contrast medium to non-contrast medium head MR image

Fig. 4.11 Temporal subtraction images from contrast medium to non-contrast medium head MR image. A brain tumor is enhanced on the subtraction image (arrow area). (a) Without VM. (b) Commercial version. (c) Voxel matching

Perspective

CA helps a computer to analyze detailed human anatomy, with the aim of supplementing the work of the radiologist with the advantage of a second opinion via automated diagnostic processes. Critical decision support will someday be implemented by utilizing the power of CA.

 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >

Related topics