Abstract:
HTML
Purpose: To assess the effect of a dedicated software environment with
alternative display methods and advanced visualization tools (LungCare,
Siemens) on the detection and classification of focal lung disease at
multidetector-row CT (MDCT) of the chest.
Methods and Materials: 25 thoracic CT studies were acquired by 16-slice MDCT
(Siemens) and reconstructed with a slice thickness of 0.75mm. Studies contained
between 3 and 12 lung nodules (166 total). As an independent gold standard,
twentynine 2-4mm lung nodules had been artificially inserted. Over the course
of 4 months, 3 radiologists with different levels of experience independently
evaluated all data sets, each displayed with 9 different techniques: 0.75mm
original axial slices, 3mm and 5mm axial multiplanar reformats (MPRs), 5mm and
10mm sliding axial maximum intensity projections (MIPs), 3mm and 10mm coronal
MPRs, and 5mm and 10mm sliding coronal MIPs. Confidence of lesion detection was
rated on a 4-point scale. For size determination of lung lesions use of
electronic calipers with axials/MPRs and 3D automated nodule volumetry with
MIPs was available. Readers were asked to evaluate each lesion and recommend
immediate invasive work-up (dł10mm), or
3 months (dł5mm) or 6 months (d<5mm)
follow-up. Reading time for each display mode was recorded.
Results: Average sensitivities were highest with 5mm and 10mm coronal MIPs
(both 98%) and lowest with 0.75mm axials and 3mm or 5mm axial MPRs (all 51%) in
the assessment of artificial lung lesions. The total number of identified
lesions was highest (99%) and lowest (69%-70%) with the same display methods.
The average confidence score for lesion detection was highest (3.9) with 10mm
coronal MIPs and lowest (2.7) with 3mm axial MPRs. Average inter-reader
agreement on recommended lesion/patient management was k=0.77 when electronic
calipers and k=0.89 when 3D nodule volumetry were available. Average reading
time for each case was shortest with 5mm coronal MIPs (4.02 min) and longest
with the original 0.75mm axial images (6.04 min). The 3 readers did not differ
significantly in nodule analysis, but reading time decreased with more clinical
experience with all methods.
Conclusion: Dedicated software environments, with alternative display methods
and advanced visualization tools, improve reader performance in the diagnosis
of focal lung disease. Sliding coronal MIP reconstructions provide highest
sensitivity and confidence in lesion detection and improve inter-reader
consistency in recommended patient management. (U.S. received a grant from
Siemens Medical Solutions; B.B. is the 2002/2003 International
BWH/Harvard-Siemens CT Research Fellow.)
Questions about this event email: schoepf@bwh.harvard.edu
Baumert MD, B,
MIPs, Multiplanar Reformats and Dedicated Display Tools for the Detection and Evaluation of Focal Lung Disease. Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL.
http://archive.rsna.org/2003/3101446.html