Information Technology Reference
In-Depth Information
Fig. 13 False positive detections in sclerotic metastasis CAD. FPs (red dot) due to: a cortex of
neuroforamen, b cortex of endplate, c bone cortex/neurocentral synchondrosis, d sclerotic margin
basivertebral bundle, e degenerative sclerosis of osteophyte, f bone island, g sclerotic Schmorl
'
s
node margin
12 Discussion
While CT is not the study of choice for initial whole body screening the whole
patient for bone metastases
skeletal scintigraphy or conventional radiography are
the most common choices [
56
-
ed on CT when
possible. However, since the cost of CT is decreasing and the radiation exposure is
also reducing, CT is becoming more affordable and available for screening pur-
poses. Additionally, the ability to detection metastases on CT continues to grow in
importance with the increasingly widespread availability and use of integrated PET/
CT imaging in the detection and follow up of metastatic disease. Using bone
window settings, CT shows a high level of detail in bone, distinguishing amongst
materials of different radio densities [
61
,
62
]. For depicting metastases to the spine,
CT is superior to skeletal scintigraphy and conventional radiography [
11
,
18
] and
has performed at sensitivities ranging from 93 to 100 % [
17
,
63
,
64
]. However,
these lesions can be subtle and easily overlooked by a radiologist, especially when
bone windows are underutilized [
58
] and the radiologist has not been speci
60
]
—
metastases still must be identi
-
cally
directed by the referring physician to look for metastases.
Detecting spinal lesions by computer is challenging, owing to the variation in
bone attenuation within and amongst patients as well as the diversity of non-
metastatic abnormalities such as degenerative disk disease. The problem must be
broken down into manageable components that can be addressed sequentially. Our
system detects lytic and sclerotic metastases separately and the results can be
combined. The
first task in detecting spinal metastases is to locate and segment the
spine, excluding other structures. This is most difficult in the thoracic spine, where
the ribs are often detected along with the vertebrae. We included location criteria in
our
false
positives were detections on costovertebral joints, as they are low intensity regions
surrounded by high attenuation cortex. The synovial joints between adjacent ver-
tebrae and the nearby contrast-
filter and classi
er to account for this, but a number of the
“
outside
”
lled IVC are sometimes segmented along with the
lumbar vertebrae, resulting in
, non-bone related false positives.
While adjacent high intensity structures pose a challenge for segmentation of the
spine from non-spine structures, the intrinsically low intensity intervertebral disks
“
outside
”
