Image Processing Reference
In-Depth Information
Table 2
Comparison Results for Sample 2
Sample 2
Manual Approach by a Surgeon Proposed
Resected volume(cc)
325
42
The number of cut-points 2
4
Results
Resected region
Resected region
Table 3
Comparison Results for Sample 3
Sample 3
Manual Approach by a Surgeon Proposed
Resected volume (cc)
550
476
The number of cut-points 5
10
Results
Resected region
Resected region
Let us evaluate the computation time. The computing platform that we use is a PC with
a Core-i7 CPU and a 24 GB memory. The operating system and programming language are
Windows 7 and Java 1.6, respectively. The program simply designed for single-thread exe-
cution. The main reason why we use Java as a programming language is that the Java-based
ImageJ has a lot of plugins like a DICOM reader, basic 2D/3D image processing and visualiz-
ation, and can be extended easily by adding user-defined plugins. Moreover, ImageJ runs on
multiplatforms like Microsoft Windows, Linux, and MacOS since it is based on Java.
Table 4
shows the computation times for samples 1, 2, and 3 and their details. Note that ves-
sel extraction, liver extraction, and tumor extraction are done in advance as explained above,
and that their computation times are not included. The time for “computing perfused regions”
is the one for
Step 1
and
Step 2
; the time for “computing resected regions” is the one for
Step
3
. The computation times vary depending on the volumes of the perfused regions and the cut
region. The larger the volumes are, the longer the computation time is. The computation time
for any sample is < 3 min. This computation time is acceptable in real preoperative planning
since the conventional preoperative planning by surgeons takes about 30-60 min for estimat-
ing a resected region. The computation time can be much speeded up by multithreading since
estimating the perfused regions and cut regions has the high degree of pixel-level parallelism.
Search WWH ::
Custom Search