Information Technology Reference
In-Depth Information
majority of the volume of the vertebra are constituted by these regions [ 65
67 ].
Bounding box inclusion of the vertebral body and pedicles thus assesses the most
likely regions of early metastatic tumor involvement. Future work could include the
laminae, transverse and posterior spinous processes for a more complete evaluation
and characterization of sclerotic tumor burden. Second, the false positive (FP) rate
is relatively high. Two of the three most common causes (volume averaging of
cortex at vertebral endplates and pedicles) may be decreased by a more sophisti-
cated segmentation algorithm design, and the third (degenerative sclerosis) by an
addition of an algorithm designed to detect degenerative change of the vertebrae.
Elimination of these three FP etiologies would eliminate 75 % of false positive
detections. Third, images reconstructed with a soft tissue kernel were used to
decrease the effect of image noise on software performance. Future design may
include increased robustness in the presence of image noise. Finally, sclerotic
lesions and lytic lesions were detected separately. An integrated approach to lytic
and sclerotic lesion detection and characterization is under investigation.
Since CAD technologies are still under intensive development, most studies of
CAD systems to-date have reported its performance in the laboratory setting rather
than in the radiology reading room. This CAD system is designed for clinical
application as a secondary reader to increase the sensitivity for detection of sclerotic
metastatic lesions in the spine. Potential future practical applications include
quantification of bone tumor burden and of change in individual lesions and total
tumor volumes with generation of metrics, for follow-up examinations in patients
undergoing treatment, as well as to assess for localized new or changing density
lesions.
In conclusion, we have presented a CAD system that can detect both lytic and
sclerotic metastases in the thoracolumbar spine. We have identi
-
ed some of the
common causes of false negative and false positive detections to guide further
development of bone CAD systems. The CAD framework is based on supervised
machine learning techniques and can be employed to detect other abnormalities in
the spine, such as fractures, osteophytes and epidural masses. Additional research
will be required to show whether bone CAD systems improve radiologists
'
diag-
nostic accuracy and interpretive ef
ciency.
References
1. Hitron A, Adams V (2009) The pharmacological management of skeletal-related events from
metastatic tumors. Orthopedics 32:188
2. Roodman GD (2004) Mechanisms of bone metastasis. N Engl J Med 350:1655 - 1664
3. Guillevin R, Vallee JN, Lafitte F, Menuel C, Duverneuil NM, Chiras J (2007) Spine metastasis
imaging: review of the literature. J Neuroradiol 34:311
321
4. Lee RJ, Saylor PJ, Smith MR (2011) Treatment and prevention of bone complications from
prostate cancer. Bone 48:88
-
95
5. Chirgwin JM, Guise TA (2007) Skeletal metastases: decreasing tumor burden by targeting the
bone microenvironment. J Cell Biochem 102:1333
-
1342
-
Search WWH ::




Custom Search