Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Staron, R. B.
Right arrow Articles by Haramati, N.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Staron, R. B.
Right arrow Articles by Haramati, N.
(Radiology. 1999;211:467-470.)
© RSNA, 1999

Musculoskeletal Imaging

Computerized Bone Densitometric Analysis: Operator-dependent Errors1

Ronald B. Staron, MD, Robin Greenspan, MD 2, Theodore T. Miller, MD 3, John P. Bilezikian, MD, Elizabeth Shane, MD and Nogah Haramati, MD 4

1 From the Department of Radiology, College of Physicians and Surgeons, Columbia University, New York Presbyterian Hospital: Columbia Presbyterian Center, Milstein Bldg 2-121, 177 Fort Washington Ave, New York, NY 10032 (R.B.S., R.G., T.T.M., N.H.); and the Departments of Medicine (J.P.B., E.S.) and Pharmacology (J.P.B.), Columbia University, New York, NY. Received February 26, 1998; revision requested April 16; final revision received August 7; accepted November 6. Address reprint requests to R.B.S.


   Abstract
TOP
 Abstract
Introduction
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
PURPOSE: To determine the nature and relative frequency of operator-dependent data analysis errors in dual x-ray absorptiometry.

MATERIALS AND METHODS: Over 40 months, 2,528 dual x-ray absorptiometric examinations of the forearm, femoral neck, and lumbar spine were performed by 11 technologists by using standard techniques and software. Each analysis was reviewed by a radiologist; errors were recorded and corrected.

RESULTS: There were no forearm analysis errors. There were 24 (0.9%) femoral neck analysis errors, of which 23 resulted from misplacement of the analysis region. There were 33 (1.3%) spinal analysis errors, of which 24 resulted from misplacement of intervertebral disk space markers. Analysis errors of the femur and spine resulted in six misdiagnoses (0.2%).

CONCLUSION: Misdiagnosis due to analysis errors is rare. Femoral neck analysis errors were easily detectable, but accurate spinal analyses depended on accurate identification of vertebral end plates and posterior elements. Nonetheless, these potentially serious errors can be detected and corrected if the analyses are reviewed and interpreted by a supervising physician who is familiar with the relevant anatomy, proper analysis techniques, and factors—such as artifacts—that adversely affect the accuracy of the analysis.

Index terms: Bones, absorptiometry, 30.1295, 40.1295 • Computers, diagnostic aid, 30.1295, 40.1295 • Femur, abnormalities, 443.56 • Hip, radiography, 44.1295 • Osteoporosis, 30.56, 40.56 • Spine, mineralization, 30.1295, 30.56 • Spine, radiography, 30.1295


   Introduction
TOP
 Abstract
 Introduction
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
Osteoporosis is to our knowledge the most common metabolic bone disorder (1). Dual x-ray absorptiometry has become the most widely used technique for diagnosing osteoporosis and is currently considered the analytic method of choice (2–5). When properly performed, dual x-ray absorptiometry allows precise and accurate quantification of bone mineral density (6–19). Analysis of the data is a computerized and almost completely automated process, but accurate operator input is needed during the performance of several key steps, such as the sizing and positioning of analysis regions (QDR-1000 Operator's Manual, Hologic, Waltham, Mass, 1989). Operator input errors may lead to errors in the diagnosis of osteoporosis, which in turn may lead to faulty comparisons with images from serial examinations. Incorrect treatment decisions could even result. The purpose of this study was to determine the nature and relative frequency of operator-dependent analysis errors.


   MATERIALS AND METHODS
TOP
 Abstract
 Introduction
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 References
 
Over 40 months in the Department of Radiology, Columbia-Presbyterian Medical Center, New York, NY, 11 licensed radiologic technologists performed dual x-ray absorptiometry in 2,528 patients by using a QDR-1000 Bone Densitometer (Hologic) equipped with routine software supplied by the manufacturer. Each examination was performed to evaluate the bone mineral density of three anatomic regions: the distal third of the nondominant radius, the right femoral neck, and the lumbar spine (posteroanterior projection of L2 through L4). Thus, examinations of the 2,528 patients yielded 7,584 analyses. The printed analysis for each examination was prospectively reviewed by one of four radiologists (R.B.S., R.G., T.T.M., N.H.). Errors were identified and recorded; a corrected analysis was then performed.


   RESULTS
TOP
 Abstract
 Introduction
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 References
 
Results are presented in the Table. Results were obtained in units of grams per square centimeter but are shown as milligrams per square centimeter to two significant figures for ease of presentation. The mean accuracy error magnitudes are the mean coefficients of variation (as percentages) of the paired values (erroneous bone mineral density and correct bone mineral density) that result from each error and its correction.


View this table:
[in this window]
[in a new window]
  		Operator-dependent Analysis Errors
 
There were 57 analysis errors; none were forearm analysis errors. There were 24 (0.9%) femoral neck analysis errors and 33 (1.3%) spinal analysis errors. Two patients' analyses had errors in both anatomic regions, so studies in 55 (2.2%) patients had analysis errors.

Two types of femoral neck analysis errors occurred. The analysis box was misplaced in 23 errors (Fig 1). The analysis region, which should have been rectangular, was misshapen in one (Fig 2).



View larger version (137K):
[in this window]
[in a new window]
  		Figure 1. Posteroanterior digital radiograph obtained during densitometry of the right hip of a 38-year-old woman. The femoral neck analysis box (white arrows) is in the wrong location; the proper location of the inferolateral corner is adjacent to the cortex of the greater trochanter (black arrow) rather than several millimeters away (white dot).

 


View larger version (138K):
[in this window]
[in a new window]
  		Figure 2. Posteroanterior digital radiograph obtained during densitometry of the right hip of a 77-year-old woman. The femoral neck analysis box (arrows) is not properly rectangular with a uniform width of 16 pixels; its lateral side is longer than its medial side.

 
There were four types of spinal analysis errors.

1. Intervertebral disk space markers were misplaced in 24 analyses when disk spaces were not properly localized (Fig 3). Fifteen (62%) of these errors resulted from improper localization of the L4-5 disk space, and four (17%) involved the T12-L1 disk space. Five of these errors had multiple levels with end-plate mislocalizations: The technologist could not adequately identify the positions of the end plates due to multiple spinal compression deformities in four cases and due to scoliosis in one case. Even when compressed vertebrae are to be excluded from the final analysis, their borders with other vertebrae must be clearly localized; yet, the deformities make such localization difficult because of the abnormal obliquity of the vertebral end plates.



View larger version (154K):
[in this window]
[in a new window]
  		Figure 3. Posteroanterior digital radiograph obtained during densitometry of the lumbar spine of a 68-year-old woman. Because of the obliquity of the L4 vertebra in relation to the x-ray beam (in spite of the proper positioning of the leg elevation cushion), the technologist did not locate the proper level for a straight line (wide white line) to indicate the L4-L5 intervertebral disk space.

 
2. Vertebrae were mislabeled in five analyses, and misidentification of the rib-bearing T12 vertebra resulted in assignment of incorrect numbers to all the vertebrae.

3. Opaque artifacts were analyzed as bone in three cases. The lumbar spinal scans in some patients revealed extraneous material within the overall analysis region, such as pills within the bowel or medical appliances overlying the abdomen. Our standard procedure was to perform repeat scanning in the patient after passage, or removal, of the extraneous material. In some cases, repeat scanning could not be performed or the materials could not be removed. If the technologist did not manually exclude these artifacts from the bone analysis region, the computer analyzed them as bone, which resulted in this error.

4. The analysis region was oversized in one case. The overall spinal analysis region incorrectly filled the entire transverse diameter of the scanned region (QDR-1000 Operator's Manual, Hologic, 1989).

By using World Health Organization diagnostic criteria (20), the 57 erroneous analyses resulted in six misdiagnoses (11%). The 23 cases of misplaced femoral neck analysis regions resulted in three misdiagnoses: Two patients with osteopenia had a misdiagnosis—one of osteoporosis and one of normal findings, and one patient with normal bone mineral density had a misdiagnosis of osteoporosis. Among the 33 spinal analysis errors, three resulted in misdiagnosis in patients with osteopenia: two as normal findings and one as osteoporosis.


   DISCUSSION
TOP
 Abstract
 Introduction
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
References
 
The accuracy error magnitudes in the Table give a scale by which to judge each error's relative severity and allow comparison with published precision values. The overall mean value of 2.7% for femoral neck misanalyses is greater than our specific scanner's measured short-term in vivo precision error of 1.36% (21) and greater than other reported values of 1.2%–2.6% for the manufacturer's model (6,10,13,17,18). Similarly, the mean value of 2.3% for our spinal misanalyses is more than our machine's measured precision error of 0.68% (21) and is more than the reported values of 0.8%–1.3% (10,13,17) for this model. Thus, operator-dependent analysis errors are of greater magnitude than the machine's intrinsic precision errors for both the hip and the spine.

Most femoral neck analysis box placement errors resulted from insufficient attention to the exact location of the box's inferolateral corner, which should have been placed immediately adjacent to the greater trochanter (Fig 1). Recognition of the relevant anatomy was not the problem; the error resulted from mistaken judgement. This type of error was clearly visible on the printed images and was easily corrected.

In contrast, the proper localization of disk space margins proved difficult. During the era of dual photon absorptiometry, which produced lower resolution images than dual x-ray absorptiometry, this had been a serious source of error (7,22). In 1990, Ho et al (7) postulated that, even with dual x-ray absorptiometry, incorrect localization of disk spaces could contribute to the accuracy error. Our results confirm that prediction. Such errors represented 24 (73%) of the 33 spinal analysis errors. Most (15 [62%] of 24) such errors arose from misplacement of the L4-5 disk space marker, most often (10 [67%] of 15) because of obscured visualization resulting from the obliquity caused by the normal lordosis of the lumbar spine. A cushion was placed beneath the legs of all patients to minimize the lordosis, but this maneuver was not always effective. To correct the errors, the reviewer placed the disk space markers in their most logical positions, taking into account the spinal contour and using the visible positions of the posterior elements as a guide (Fig 3). In such cases, posteroanterior digital images are similar enough to anteroposterior radiographs that expertise with the latter is helpful in interpreting the former.

Major errors result from misidentification of the rib-bearing T12 vertebra, since all the vertebrae are thereby mislabeled. This error, however, is most easily correctable by a person familiar with the normal anatomy and its variations, including the occasional riblets that articulate with L1, as well as the occasional absence of ribs on T12 (23). Yet, it may be difficult for anyone to correctly number the vertebrae, so comparison with any available old scans or plain radiographs is desirable.

The most severe error occurred when radiopaque artifacts within the region of interest were erroneously analyzed as bone. Previous literature indicates that aortic calcifications superimposed over the spine cause only an unsubstantial increase in calculated bone mineral density (24–26). Unlike spotty aortic calcifications, our larger artifactual opacities in three cases resulted in a mean error in bone mineral density of 170 mg/cm2 and a mean accuracy error of 10%. If such artifacts overlie part of a vertebra, one might consider excluding the entire vertebra from the analysis, rather than manually altering the computer-generated vertebral contour in the attempt to exclude the artifact, since such manual alteration of bone contours is itself a source of error (QDR-1000 Operator's Manual, Hologic, 1989). Also, it is important to recognize that in some cases the erroneous lines of analysis are difficult to see on the computer printout because the opaque artifacts obscure visualization of the lines, and it deceptively appears as if the computer has excluded the artifact. If the operator decides not to exclude that vertebra from analysis, careful inspection of the printouts should be supplemented by review of the computer monitor, where the offending line is displayed in color and is distinguished clearly from the underlying artifact. Some scanner models have C-arm gantries that allow lateral scanning and thus exclusion of such artifacts.

Clinical decisions to treat or withhold treatment for osteoporosis are heavily dependent on the densitometric results in this often asymptomatic condition, for which there are increasingly numerous treatment options, each with its own risks and complications. In this study, the erroneous results caused cases in six patients to be placed in the wrong diagnostic categories, according to World Health Organization criteria (20). Likewise, analysis errors may confound comparison with previous or future scans by giving the wrong impression of a "trend" in the patient's densitometric results. One limitation of this study was that such potential long-term effects on patient care could not be studied because all errors were immediately corrected.

Also, this study was designed to identify types of operator-dependent analysis errors and to look at their relative frequencies, not to evaluate prospectively the training or performance of individual technologists. Two technologists had been trained by representatives of the manufacturer; the others were trained by those two. All were licensed radiologic technologists. Detailed records of service could not be maintained. Nevertheless, informal review of the data indicate that the total of 57 errors occurred sporadically throughout the 40-month study; temporally focal clusters of errors were generally not found, so there is no evidence that the errors were usually clustered at the beginning of each technologist's service experience, and there is no evidence that the errors related to the amount of training received by each technologist or to the technologist's subsequent experience. In the future, it would be of interest to study the number of errors produced by a site's single, dedicated densitometric technologist.

Correction of operator-dependent errors does not slow patient throughput, since correction is performed during the postprocessing of data while other patients are undergoing scanning. No matter how trustworthy the technologist, a radiologist or other physician with expertise in the interpretation of spinal radiographs should carefully review all analyses to identify and correct the few, but potentially serious, operator-dependent analysis errors.


   Footnotes
 
2 Current address: Department of Diagnostic Imaging, Valley Health System, Medical Center Campus, Beaver, Pa. Back

3 Current address: North Shore Radiology, Great Neck, NY. Back

4 Current address: Unified Department of Radiology, Albert Einstein College of Medicine and Montefiore Medical Center, Bronx, NY. Back

Author contributions: Guarantor of integrity of entire study, R.B.S.; study concepts, R.B.S., R.G., J.P.B., E.S.; study design, R.B.S., R.G.; definition of intellectual content, R.B.S., R.G., J.P.B., E.S., T.T.M., N.H.; literature research, R.B.S.; clinical studies, R.B.S., R.G., T.T.M., N.H.; data acquisition and analysis, R.B.S., R.G., T.T.M., N.H.; manuscript preparation, R.B.S., T.T.M.; manuscript editing and review, R.B.S., R.G., T.T.M., J.P.B., E.S., N.H.


   References
TOP
 Abstract
 Introduction
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 

  1. Melton LJ, Eddy DM, Johnston CC. Screening for osteoporosis. Ann Intern Med 1990; 112:516-528.
  2. Jergas M, Genant HK. Quantitative bone mineral analysis. In: Resnick D, eds. Diagnosis of bone and joint disorders. Philadelphia, Pa: Saunders, 1995; 1854-1880.
  3. . Consensus development conference: diagnosis, prophylaxis, and treatment of osteoporosis. Am J Med 1993; 94:646-650.[Medline]
  4. Grampp S, Jergas M, Gluer CC, Lang P, Brastow P, Genant HK. Radiologic diagnosis of osteoporosis: current methods and perspectives. Radiol Clin North Am 1993; 31:1133-1145.[Medline]
  5. Sartoris DJ. Coding and reimbursement issues for dual-energy x-ray absorptiometry. AJR 1994; 163:137-139.[Abstract/Free Full Text]
  6. Gluer CC, Steiger P, Selvidge R, Elliesen-Kliefoth K, Hayashi C, Genant HK. Comparative assessment of dual-photon absorptiometry and dual-energy radiography. Radiology 1990; 174:223-228.[Abstract/Free Full Text]
  7. Ho CP, Kim RW, Schaffler MB, Sartoris DJ. Accuracy of dual-energy radiographic absorptiometry of the lumbar spine: cadaver study. Radiology 1990; 176:171-173.[Abstract/Free Full Text]
  8. Lang P, Steiger P, Faulkner K, Gluer C, Genant HK. Osteoporosis: current techniques and recent developments in quantitative bone densitometry. Radiol Clin North Am 1991; 29:49-76.[Medline]
  9. Larcos G, Wahner HW. An evaluation of forearm bone mineral measurement with dual-energy x-ray absorptiometry. J Nucl Med 1991; 32:2101-2106.[Abstract/Free Full Text]
  10. Laskey MA, Flaxman ME, Barber RW, et al. Comparative performance in vitro and in vivo of lunar DPX and Hologic QDR-1000 dual energy x-ray absorptiometers. Br J Radiol 1991; 64:1023-1029.[Abstract]
  11. Lai KC, Goodsitt MM, Murano R, Chesnut CH. A comparison of two dual-energy x-ray absorptiometry systems for spinal bone mineral measurement. Calcif Tissue Int 1992; 50:203-208.[Medline]
  12. Mitlak BH, Schoenfeld D, Neer RM. Accuracy, precision, and utility of spine and whole-skeleton mineral measurements by DXA in rats. J Bone Miner Res 1994; 9:119-126.[Medline]
  13. Nelson DA, Brown EB, Flynn MJ, Cody DD, Shaffer S. Comparison of dual photon and dual energy x-ray bone densitometers in a clinic setting. Skeletal Radiol 1991; 20:591-595.[Medline]
  14. Orwoll ES, Oviatt SK. Longitudinal precision of dual-energy x-ray absorptiometry in a multicenter study: the Nafarelin/Bone Study Group. J Bone Miner Res 1991; 6:191-197.[Medline]
  15. Orwoll ES, Oviatt SK, Biddle JA. Precision of dual-energy x-ray absorptiometry: development of quality control rules and their application in longitudinal studies. J Bone Miner Res 1993; 8:693-699.[Medline]
  16. Pacifici R, Rupich R, Vered I, et al. Dual energy radiography (DER): a preliminary comparative study. Calcif Tissue Int 1988; 43:189-191.[Medline]
  17. Slosman DO, Rizzoli R, Buchs B, Piana F, Donath A, Bonjour JP. Comparative study of the performances of x-ray and gadolinium 153 bone densitometers at the level of the spine, femoral neck and femoral shaft. Eur J Nucl Med 1990; 17:3-9.[Medline]
  18. Svendsen OL, Marslew U, Hassager C, Christiansen C. Measurements of bone mineral density of the proximal femur by two commercially available dual energy x-ray absorptiometric systems. Eur J Nucl Med 1992; 19:41-46.[Medline]
  19. Wahner HW, Dunn WL, Brown ML, Morin RL, Riggs BL. Comparison of dual-energy x-ray absorptiometry and dual photon absorptiometry for bone mineral measurements of the lumbar spine. Mayo Clin Proc 1988; 63:1075-1084.[Medline]
  20. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N. The diagnosis of osteoporosis. J Bone Miner Res 1994; 9:1137-1141.[Medline]
  21. Shane E, Rivas M, McMahon DJ, et al. Bone loss and turnover after cardiac transplantation. J Clin Endocrinol 1997; 82:1497-1506.[Abstract/Free Full Text]
  22. Borders J, Kerr E, Sartoris DJ, et al. Quantitative dual-energy radiographic absorptiometry of the lumbar spine: in vivo comparison with dual-photon absorptiometry. Radiology 1989; 170:129-131.[Abstract/Free Full Text]
  23. Southworth JD, Bersack SR. Anomalies of the lumbosacral vertebrae in five hundred and fifty individuals without symptoms referable to the low back pain. AJR 1950; 64:624-634.
  24. Frye MA, Melton LJ, Bryant SC, et al. Osteoporosis and calcification of the aorta. Bone Miner 1992; 19:185-194.[Medline]
  25. Drinka PJ, DeSmet AA, Bauwens SF, Rogot A. The effect of overlying calcification on lumbar bone densitometry. Calcif Tissue Int 1992; 50:507-510.[Medline]
  26. Reid IR, Evans MC, Ames R, Wattie DJ. The influence of osteophytes and aortic calcification on spinal mineral density in postmenopausal women. J Clin Endocrinol Metab 1991; 72:1372-1374.[Abstract]



This article has been cited by other articles:


Home page
 RadiologyHome page
B. C. Lentle and J. C. Prior
Osteoporosis: What a Clinician Expects to Learn from a Patient's Bone Density Examination
Radiology, September 1, 2003; 228(3): 620 - 628.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
J. A. Jacobson, D. A. Jamadar, and C. W. Hayes
Dual X-Ray Absorptiometry: Recognizing Image Artifacts and Pathology
Am. J. Roentgenol., June 1, 2000; 174(6): 1699 - 1705.
[Full Text]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Staron, R. B.
Right arrow Articles by Haramati, N.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Staron, R. B.
Right arrow Articles by Haramati, N.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE