Information Technology Reference
In-Depth Information
average of 124 images (range 60
145). Patients had a CT of the chest, abdomen, and
pelvis (44 patients), abdomen and pelvis (2 patients), chest (1 patient), chest and
abdomen (1 patient), abdomen (1 patient), or pelvis (1 patient). The standard recon-
struction kernel was used for 49 patients and the
-
“
B
”
kernel for 1 patient. The patients
received 110
130 cc Iopamidol (Isovue-300, Bracco Diagnostics, Princeton, NJ)
intravenous contrast material given via power injector. Using a two-reader consensus,
all thoracolumbar spine lesions were identi
-
ed and qualitatively characterized as lytic,
sclerotic, or mixed
at least 20% lytic voxels in a predominately sclerotic lesion, vice
versa, or lesion without a predominant voxel type. The lytic lesions were further
characterized as probable or unlikely metastases; only the probable metastases were
included in this study. For example, unlikely metastases including Schmorl
—
'
s nodes,
degenerative disc disease, osteopenia, and hemangiomas, were excluded.
The largest lytic area of each lesion was measured by the largest area enclosed
by the contour in the x-y plane and if lesions merged at some point, they were
considered one lesion. Due to the 5 mm thickness of slices, the z plane was not
used. The lesions were strati
ed by the maximum area on x-y plane. They were put
in three size categories, >0, >0.2 and >0.8 cm
2
. Among these, we set 0.8 cm
2
minimum lytic area threshold (equivalent to 1 cm in diameter) for lesions of sub-
stantial size and clinically critical. There were 28 probable lytic metastases with
lytic areas >0.8 cm
2
, which corresponds to a circle with a diameter of >1 cm (12 in
the training set, 16 in the test set). Patients had between zero and four probable lytic
metastases (average 0.6) with areas 0.9
10.6 cm
2
(average 2.7 cm
2
). 33 patients
with lytic lesions did not have any lesions characterized as probable metastases.
A total of 35 mixed lesions and 37 sclerotic lesions were also present in the study
group. Lesions were manually segmented by a trained student, who drew a contour
along the voxels on the edge of each lesion on each slice that it appeared. The
manual segmentation was used as the reference standard segmentation in our study.
The cohort was divided into training and test sets (29 patients in training and 21
in test). There were in total 90 lytic bone metastases (58 in training and 32 in test
set). The data are summarized in Table
2
.
-
10.2 Sclerotic Metastasis Cohort
The sclerotic metastasis cohort consisted of CT examinations from 60 patients
(mean age 56.2 years, range 12
77 years; 19 females, 41 males). 50 of them
demonstrated one or more sclerotic lesions of the spine. 10 of them were control
-
Table 2 Data summary for lytic metastasis cohort
Number
of patients
Lesions
>0cm
2
Lesions
> 0.2 cm
2
Lesions
> 0.8 cm
2
Training set
29
58
44
12
Test set
21
32
27
16
