Pulmonary
Nodule Analysis Applications
Several
Image analysis applications are available on the system for the analysis of
chest and cardiac CT images.
Image nodule
detectors
|
vmonodule |
This application runs
the MONAI lung nodule detection algorithm |
|
vtsnodule |
Runs the TotalSegmentator (Bluemind)
nodule detection algorithm |
These applications
have the following Syntax
<application> [if=]<input> [of=]output [od=<output-directory>] [-ic] [-csv]
Nodule
csv file output:
The output of
the nodule detectors is a csv file with the following columns
|
Column Heading |
Contents |
|
image |
image-code of the analyzed image |
|
nodule |
a numeric nodule identifier (starts at 1) |
|
x |
x coordinate of nodule center in pixels |
|
y |
y coordinate of nodule center in pixels |
|
z |
z coordinate of nodule center in pixels |
|
prob |
probability that the entity is a pulmonary nodule |
|
size |
estimation of the diameter of the nodule in mm |
|
xs |
x-dimension extent of the nodule in pixels |
|
ys |
y-dimension extent of the nodule in pixels |
|
xs |
z-dimension extent of the nodule in pixels |
The main program for providing input to the pulmonary nodule classifier is vnodext which has the following syntax:
vnodext
[if=]<original-image> in=<nodule-csv-file> [of=][output directory] [-ic]
When the -ic option is specified the image original-image is specified
by just the image-code. Note, this algorithm is define
for a single image analysis only. A csv file from a batch run must be partitioned
into individual image-code sets of rows for analysis.
The following experimental programs for nodule analysis are currently available:
|
vnodext |
Extract the nodule regions
from the original image and creates individual 32 x 32 x 32 images for
malignancy evaluation. For classification the nodules are magnified by
interpolation to occupy most of the space in the image. Actual unmagnified images
can also be created for visualization. |
|
vnodvis |
Create a visualization version of the original image in which the regions of all detected nodules are clearly visible. Input parameters are similar to vnodext |
|
vnview |
The images that result
from vnodext are windowed and tiled into a 6
x 6 grid for convenient visualization. The input is a 32 x 32 x 32 nodule
region image created by vnodext. |
|
vnnview |
Apply vnview to all images in a directory |