Information Technology Reference
In-Depth Information
cases. The medical records of the patients whose studies were selected for the study
were reviewed for demographics and pathology (27 with prostate cancer and the
rest with other types of pathology). CT examination technique varied by the
scanners, and across the time interval of the case studies. CT examination of the
chest, abdomen and pelvis was performed on 56 patients, CT of the abdomen and
pelvis on 3 patients, and CT of the chest and abdomen on 1 patient. 58 of the 60
patients received intravenous contrast as part of their examination, with speci
c
information not available on two of the 58 patients. Contrast was given as Isoview
300 (iopamidol injection 61 %; Bracco Diagnostic Inc., Princeton, NJ) to 44
patients in volumes ranging from 40 ml (n = 1) to 130 ml (n = 41), with one patient
receiving 125 ml of Isovue-370 (lopamidol Injection 76 %; Bracco Diagnostic Inc.,
Princeton, NJ), and one receiving 110 ml of Omnipaque-300 (iohexol, iodine
content 46.36 %, GE Healthcare Inc., Princeton, NJ).
CT images were reviewed by a musculoskeletal radiologist with 5-years of
experience. Manual segmentation was performed over the three-dimensional extent
of each lesion, using in-house software with a bone window setting [
55
]. These
marked lesions formed the reference standard for determination of CAD software
performance. Large heterogeneous sclerotic regions that appeared to be either an
amalgam of inseparable smaller component sclerotic lesions or a single heteroge-
neous lesion were location-marked and segmented as a single lesion. A total of 552
sclerotic lesions of the spine were greater than 0.84 cm average diameter.
The cohort was divided into training and test sets (17 patients in training and 43
in test). The data are summarized in Table
3
. It also summarizes the size and
intensity distribution of the metastases in the data set.
Table 3 Summary of sclerotic metastasis cohort
Reference standard lesion sets
Training
Testing
Number of cases
17
43
Number of lesions
180
372
Volume
1
range
Min
Max
Min
Max
0.3
52.7
0.3
56.9
Volume
1
mean
±
standard deviation (SD)
5.4
±
9.6
3.3
±
6.8
Length
2
range
Min
Max
Min
Max
0.7
8.1
0.8
7.6
Length
2
mean
±
SD
2.4
±
1.6
2.0
±
1.2
Range of the mean HU (CT attenuation) per lesion
Min
Max
Min
Max
91.8
878
161
821
Mean of the mean HU per lesion
±
SD
525
±
150
391
±
106
Range for the SD of the mean HU per lesion
Min
Max
Min
Max
24
322
28
317
Mean of SD of the mean HU per lesion
±
SD of SD
179
±
65
124
±
54
1. Units of cm
3
, 2. Units of cm
