Abstract:
HTML
Purpose: To evaluate the performance of a computer aided diagnosis tool for
automated detection of peripheral pulmonary emboli in large volume
multidetector-row CT data sets.
Methods and Materials: 33 consecutive multidetector-row CT pulmonary angiography
studies of varying diagnostic quality with a clinical diagnosis of pulmonary
embolism were retrospectively identified. All studies had been acquired with
contrast enhanced 4-slice or 16-slice multidetector-row CT (Siemens) with a
slice thickness of 0.75-mm - 1.25-mm. Studies contained between 320 and 540
individual axial sections. Consensus reading for the identification of
segmental and subsegmental pulmonary emboli by two experienced radiologists
with a third radiologist as an adjudicator was used to establish the standard
of reference. The performance of a computer aided diagnosis tool (ImageChecker
CT, R2 Technology, Sunnyvale, CA) for the automated detection of pulmonary
emboli was tested on all data sets. The computer aided diagnosis tool identifies
a list of candidates and analyzes them according to shape, density, size,
texture, and other properties. A likelihood of representing pulmonary embolism
is associated with each analyzed structure and findings are marked as a
candidate lesion by the computer aided diagnosis system if the finding meets
the detection threshold.
Results: Consensus reading revealed 186 pulmonary emboli on the segmental and
120 pulmonary emboli on the subsegmental pulmonary arterial level in the 33
patient studies. In 5 of 33 patients (15%) pumonary emboli were isolated to the
subsegmental pulmonary arterial level. Sensitivity of the computer aided
diagnosis tool tested was 88% (164/186) for the detection of segmental
pulmonary emboli and 78% (94/120) for the detection of subsegmental pulmonary
emboli. The average false positive rate of the computer aided diagnosis tool
was 4/case. All patients, including those with isolated subsegmental pulmonary
emboli, were correctly identified as having pulmonary embolism by the computer
aided diagnosis tool (no false negative cases).
Conclusion: Highly sensitive and specific computer aided detection of pulmonary
embolism is feasible based on high-resolution multidetector-row CT pulmonary
angiography. Application of computer aided diagnosis tools may increase
diagnostic accuracy by detecting isolated peripheral pulmonary emboli and may
aid radiologists in the interpretation of large volume multidetector-row CT
pulmonary angiography studies. (A.C.S., S.A.W. are employees of R2 Technology,
Sunnyvale, CA. P.C. received a research grant from R2 Technology, Sunnyvale, CA.)
Questions about this event email: schoepf@bwh.harvard.edu
Das MD, M,
Computer-aided Diagnosis of Peripheral Pulmonary Emboli. Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL.
http://archive.rsna.org/2003/3102033.html
