VANCGEN | VisionX V4 |
vancgen - generate attached nodule candidates from a whole lung CT scan.
SYNOPSISvancgen [if=] [ig=] [of=] [c=] [k=] [xy=] [z=] [i=] [x=] [s=] [d=] [th=] [sx=] [sz=] [-d] [-v] [-c] [-s] [-u]
DESCRIPTIONvancgen program generates attached nodule candidates from a whole lung CT scan.
A volume occupancy distribution curve is generated for each suspicious lung boundary voxels [if=] using three-dimensional anisotropic templates. The template is centered at every lung boundary voxel and the ratio of the volume of the lung parenchyma inside the template to the volume of the template is calculated. This is repeated for incrementally increasing size templates and the distribution is plotted. Lung boundary pixels with distinct peak in their distribution curve are marked as potential nodule candidate locations. Boundary pixels belonging to the same nodule candidate are clustered together and updated at every iteration, thereby significantly reducing the computational complexity. In addition, several speed up heuristics have been incorporated in the algorithm to make it more efficient and faster.
The input image in-plane and axial resolutions are automatically read from the image header file. These values can also be set directly by using [xy=] and [z=] options. The program outputs potential attached nodule locations in multiple ways. If a text file is specified in k= option, a text file will be generated with a list of potential nodule XYZ coordinates with their associated filter response. In addition, a visionx file with 3D vector element and object ID corresponding to XYZ coordinates of the potential nodule location and object ID respectively can be generated using of= option. A volume occupancy distribution curve for a single voxel [c=] can be plotted using [-p] option.
A useful visualization tool with potential nodule locations marked with colors overlaid on top of windowed original input image is also available (-c option). For response for all pixel locations can be dumped to a text file using [os=] option. Debugging options -v and -d are available to available for different levels
OPTIONS|
if= |
suspicious lung surface voxel file (byte type) | ||
|
ig= |
Lung mask before morphological closing operation (byte type data). | ||
|
[of=] |
visionx output file with 3D vector corresponding to XYZ coordinates of nodule location and object ID | ||
|
[k=] |
Potential attached nodules location in text file format | ||
|
[os=] |
volume occupancy response for different template sizes for all surface voxels | ||
|
[xy=] |
in-plane resolution | ||
|
[z=] |
axial direction resolution | ||
|
[-u] |
anisotropic data | ||
|
[i=] |
minimum size for the spherical filter | ||
|
[x=] |
maximum size for the spherical filter | ||
|
[s=] |
step size for the filter |
[th=th1,th2,th3,th4,sth]
Parameters used to analaysze trhe volume occupancy curve th1 minimum volume occupancy threshold (speed up heuristics) th2 volume occupnacy max threshold for large size candidates (> 4mm) th3 volume occupancy max threshold for small size nodules (<= 4mm) th4 vocc calculation stopping criteria to analyze trailing edge of the distribution curve sth size difference between the peak point and the current iteration size value mainly for speed up heuristics.
|
[sx=] |
xy scaling factor for the clustger analysis window | ||
|
[sz] |
z scaling factor for the cluster analysis window | ||
|
[-a] |
cluster analysis to achieve a single response for each nodule candidate. | ||
|
[-p] |
generate volume occupancy plot file for surface voxel [c=] | ||
|
[-v] |
selects the verbose mode to print information about intermediate results | ||
|
[-c] |
generate color marked output image for quick visualization (default name IVALdcol.vx) | ||
|
[-s] |
circle the candidate regions for visualization (IVALdcol.vx) | ||
|
[-d] |
output filter response at every pixel location (useful for debugging) |
vpsgen
AUTHORAndinet A. Enquobahrie