Artifacts and abnormalities: computer-aided densitometry requires skilled interpretation.
In fact, the technology's accuracy is largely operator dependent, and its findings should be interpreted with caution, according to the conclusions of two posters presented at the annual meeting of the International Society for Clinical Densitometry.
The technology, which uses a pattern-detecting software to analyze dual-energy x-ray absorptiometry (DEXA), requires that physicians ensure that patients are positioned properly, the correct scan mode is selected, and the region of interest is identified precisely during analysis. Operators must be able to identify structural pathology that can be associated with unusual elevations in bone mineral density, such as osteophytes, vertebral fractures, and osteoarthritis.
GE Medical Systems' recently released computer-assisted densitometry (CAD) software for its Lunar Prodigy device helps zero in on areas that require closer review by the interpreting physician. The system also flags atypical conditions having to do with the measurement, analysis, and vertebral anatomy. The software is also available on some other GE products, such as the Lunar DPX Bravo densitometer.
The other major manufacturer of densitometers, Hologic, offers a CADfx program to evaluate instant vertebral assessment exams for compression fractures. Unlike the GE program, Hologic software is intended only for use by physicians or DEXA readers, not as an aid for technician training.
Dr. Elliott N. Schwartz and Dee M. Steinberg compared the GE CAD software's interpretation with their own assessment of 71 spine scans and 35 bilateral femur scans. About two-thirds of the scans contained an abnormality, and one-third were normal. The two experienced densitometry reviewers agreed on their interpretations before learning the CAD findings. Neither Dr. Schwartz nor Ms. Steinberg have a conflict of interest regarding GE or the CAD software.
There was strong agreement between the experienced physicians and CAD, noted Dr. Schwartz, co-medical director, and Ms. Steinberg, bone density testing program manager at the Foundation for Osteoporosis Research and Education in Oakland, Calif.
In the spine study, the "blinded" users and the software agreed on 76%-86% of scans, depending on what was being evaluated, for example, whether the spine was centered, whether there was unusually high bone density, and whether the spine was curved.
In the femur study, there was a wider range of agreement, with 63%-97%, again depending on what was being evaluated--for example, the degree of shaft-pelvis separation, whether the area of interest was correctly identified, and whether borders between tissue types were accurately identified.
After reviewing the CAD results, the two experienced densitometry reviewers agreed with the software in 83%-92% of spine scans and 66%-97% of femur scans.
BY DAMIAN McNAMARA
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Clinical Rounds|
|Publication:||Internal Medicine News|
|Date:||May 1, 2004|
|Previous Article:||Forearm densitometry: room in primary care practice for BMD tests.|
|Next Article:||Reverse sequence doesn't work: PTH followed by alendronate shows synergistic effect.|