Orthopedics

Feature Article 

The Utility of the Energy Subtraction Method for Thoracic Spine Radiography in Clinical Practice: An Analysis of 25 Patients With Multiple Myeloma

Takaki Shimizu, MD; Takeshi Sasagawa, MD, PhD; Naohiro Kawamura, MD, PhD; Shigeru Masuyama, MD; Naohiro Tachibana, MD; Haruka Emori, MD, PhD; Masaaki Iizuka, MD; Hisatoshi Ishikura, MD; Kenshi Suzuki, MD, PhD; Takeshi Kuwabara, MS; Hiroyuki Tsuchiya, MD, PhD; Junichi Kunogi, MD

Abstract

Interpretation of thoracic spine radiographs is difficult because they cannot clearly depict the vertebrae due to overlap with soft tissues. This study aimed to evaluate whether thoracic spine radiographs obtained using the energy subtraction method could improve the accuracy of a diagnosis of thoracic osteolytic lesions. The authors analyzed 300 thoracic vertebrae from 25 patients with multiple myeloma who underwent thoracic spine radiography. All patients underwent thoracic spine radiography with 2 views. Two sets of images were prepared: computed radiography images (CR images) acquired using conventional processing parameters; and processed images for specifically visualizing bone, using the energy subtraction method (ES images). The CR images (CR group) and paired CR and ES images (CR+ES group) were interpreted in parallel by 5 orthopedic surgeons. The presence of osteolytic lesions was evaluated for each of the 12 thoracic vertebrae, and the sensitivity and specificity of the method were compared with computed tomography (CT), which is considered the gold standard. Subgroup analysis was also performed based on location. Osteolytic lesions were found on CT in 28 (9.3%) vertebrae of 12 patients. The overall sensitivities and specificities of the CR and CR+ES groups were 17.2% and 54.3%, respectively, and 95.6% and 98.0%, respectively, with statistically significant differences. Subgroup analysis showed particular improvement in the sensitivity for the CR+ES group in the middle thoracic spine compared with that at other locations. Thoracic spine radiographs generated using this method may improve the accuracy of diagnosis of thoracic osteolytic lesions. [Orthopedics. 2021;44(1):e31–e35.]

Abstract

Interpretation of thoracic spine radiographs is difficult because they cannot clearly depict the vertebrae due to overlap with soft tissues. This study aimed to evaluate whether thoracic spine radiographs obtained using the energy subtraction method could improve the accuracy of a diagnosis of thoracic osteolytic lesions. The authors analyzed 300 thoracic vertebrae from 25 patients with multiple myeloma who underwent thoracic spine radiography. All patients underwent thoracic spine radiography with 2 views. Two sets of images were prepared: computed radiography images (CR images) acquired using conventional processing parameters; and processed images for specifically visualizing bone, using the energy subtraction method (ES images). The CR images (CR group) and paired CR and ES images (CR+ES group) were interpreted in parallel by 5 orthopedic surgeons. The presence of osteolytic lesions was evaluated for each of the 12 thoracic vertebrae, and the sensitivity and specificity of the method were compared with computed tomography (CT), which is considered the gold standard. Subgroup analysis was also performed based on location. Osteolytic lesions were found on CT in 28 (9.3%) vertebrae of 12 patients. The overall sensitivities and specificities of the CR and CR+ES groups were 17.2% and 54.3%, respectively, and 95.6% and 98.0%, respectively, with statistically significant differences. Subgroup analysis showed particular improvement in the sensitivity for the CR+ES group in the middle thoracic spine compared with that at other locations. Thoracic spine radiographs generated using this method may improve the accuracy of diagnosis of thoracic osteolytic lesions. [Orthopedics. 2021;44(1):e31–e35.]

Despite the widespread use of computed tomography (CT), radiographic examination remains indispensable to daily practice. Although CT is beneficial for diagnosis and treatment, it is associated with high radiation exposure and financial costs. Because CT contributes to the total cancer risk,1 improvements in the quality of radiography are needed to reduce the use of CT.

Thoracic spine radiographs are typically used for primary evaluation of patients with suspected thoracic spine disorders. However, interpretation of thoracic spine radiographs is difficult because they cannot clearly depict the thoracic spine due to overlap with soft tissues, such as the heart, lung, trachea, aorta, and pulmonary blood vessels. To improve the clarity of these images, Sasagawa et al2 applied a one-shot energy subtraction method, effectively used in chest radiography,3–6 to thoracic spine radiographs. This method involves acquiring images from a single radiation exposure at 2 radiograph energies. Lowand high-energy information is obtained by sandwiching a copper filter, which absorbs low-energy radiograph photons, between 2 imaging plates. Weighted subtraction is applied to the information obtained from these images to eliminate materials with specific radiograph absorption properties. Because bones and soft tissues have different radiograph absorption magnitude ratios, signals corresponding to bones or soft tissues may be eliminated by subtraction.2–6

Sasagawa et al2 reported that thoracic spine radiographs comprised exclusively of bone signals generated by this method depicted the vertebrae more clearly than the corresponding original radiographs, without additional workload for the radiological technician or patient exposure to additional radiation. The clarity of the endplate and trabeculae in the images was evaluated in the study; however, the utility of this method in improving interpretational accuracy in clinical practice has not been evaluated.

Therefore, the authors conducted this study to assess bone-lesion detection in actual clinical settings using the images generated by this one-shot energy subtraction method. The purpose of this study was to evaluate whether the use of thoracic spine radiographs generated using this method improved the accuracy of diagnosis of thoracic osteolytic lesions in patients with multiple myeloma (MM).

Materials and Methods

This prospective study included consecutive patients diagnosed with MM who underwent thoracic spine radiography for screening of bone lesions between November 1, 2015, and August 31, 2016, at the authors' institution. The institutional ethics committee approved the study. Written informed consent was obtained from all patients.

All patients underwent thoracic spine radiographic examination, including both frontal and lateral views. The authors used FCR XU-D1 (Fujifilm Corporation) as the image reader. The radiographic conditions used were in accordance with those reported by Sasagawa et al.2 Two sets of images were prepared from each radiograph: computed radiography images (CR images) using conventional processing parameters; and processed images specialized for bone-structure visualization using the one-shot energy subtraction method (ES images; Figure 1).

A 62-year-old man with multiple myeloma. Conventional frontal (a) and lateral (b) computed radiography images. Frontal (c) and lateral (d) views processed using the one-shot energy subtraction method. Axial computed tomography scans at the T8 (e) and T12 (f) levels. The osteolytic lesions at T8 and the endplate fracture at T12 are more clearly visible in the processed image of the frontal view (c, arrowheads). The osteolytic lesion in the anterior vertebral body is more clearly visible in the processed image of the lateral view (d, arrowheads).

Figure 1:

A 62-year-old man with multiple myeloma. Conventional frontal (a) and lateral (b) computed radiography images. Frontal (c) and lateral (d) views processed using the one-shot energy subtraction method. Axial computed tomography scans at the T8 (e) and T12 (f) levels. The osteolytic lesions at T8 and the endplate fracture at T12 are more clearly visible in the processed image of the frontal view (c, arrowheads). The osteolytic lesion in the anterior vertebral body is more clearly visible in the processed image of the lateral view (d, arrowheads).

The CR images (CR group) and the paired CR and ES images (CR+ES group) were interpreted in parallel by 5 general orthopedic surgeons (S.M., N.T., H.E., M.I., H.I.) who had never seen the images of the patients in their clinical practice. Simple randomization determined which of the images from the CR and CR+ES groups would be interpreted first. The observers were double-blinded to the grouping of the images and were only informed that the images were acquired from patients with MM for bone-lesion screening. Images of one group were interpreted 4 weeks after the images of the other group. Regular image manipulation, including windowing and zooming, was possible for both images.

The frontal and lateral view radiographs were evaluated in combination to assess the presence of osteolytic lesions in each of the 12 thoracic vertebrae in each case. The accuracy of the interpretation was judged based on interpretation of the CT reports by radiologists; this was considered the gold standard. These CT scans were routinely performed by hematologists as part of whole-body screens for bone lesions in all patients diagnosed with MM. The sensitivity and specificity (which were the primary endpoints) of detecting osteolytic lesions in each interpretation group were calculated relative to those of the CT gold standard. Secondary endpoints included analysis of subgroups classified based on location: upper (T1-4), middle (T5-10), and lower (T11-12) thoracic spines.

The sensitivity and specificity of each interpretation group for each observer were statistically compared using McNemar's test with post hoc power analysis. The overall sensitivity and specificity of each interpretation group were statistically compared using the paired t test. Statistical significance was defined as P<.05. Intra- and interobserver reliabilities of the CR and CR+ES groups were assessed using Fleiss' kappa coefficient. All statistical analyses were performed using R Software for Statistical Computing, version 2.8.1 (R Foundation for Statistical Computing). The authors also used G-Power software (Franz Faul, Univesitat Kiel) to calculate the statistical power of the analysis.

Results

This study included 25 patients with MM (11 men and 14 women). Mean age was 68.2 years (range, 62 to 78 years). In 25 cases, CT revealed that 28 (9.3%) of 300 vertebrae from 12 patients had osteolytic lesions caused by MM. Table 1 shows the distribution of the osteolytic lesions at each thoracic vertebral level, as detected by CT. There was no time interval or clinical intervention between thoracic spine radiographs and CT because both tests were routinely performed at the time of diagnosis.

Distribution of the Osteolytic Lesions Detected on Computed Tomography at Each Thoracic Vertebral Levela

Table 1:

Distribution of the Osteolytic Lesions Detected on Computed Tomography at Each Thoracic Vertebral Level

With the exception of 1 observer, the sensitivity for the CR+ES group was significantly higher than that for the CR group; the specificity for the CR+ES group was significantly higher than that for the CR group for only 1 observer. Statistical power for the analysis of sensitivity and specificity was 0.92 and 0.74, respectively. Overall sensitivities and specificities for the CR and CR+ES groups were 17.2% and 54.3%, respectively, and 95.6% and 98.0%, respectively; the difference was statistically significant (Table 2).

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions Relative to a Computed Tomography Gold Standard

Table 2:

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions Relative to a Computed Tomography Gold Standard

Subgroup analysis based on location showed that the sensitivity for the CR+ES group in the middle thoracic spine was significantly higher than that for the CR group for all but 1 observer. The sensitivity and specificity for the CR+ES group tended to be higher than those for the CR group in the upper and lower thoracic spines; however, the difference was not statistically significant. Overall, the sensitivity and specificity for the CR+ES group were significantly higher than those for the CR group in the upper and middle thoracic spines (Table 3). The Fleiss' kappa coefficients for the CR and CR+ES groups were 0.84 and 0.89 for intraobserver reliability, respectively, and 0.31 and 0.60 for interobserver reliability, respectively.

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions of Each Subgroup Classified on the Basis of Location

Table 3:

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions of Each Subgroup Classified on the Basis of Location

Discussion

This study compared the diagnostic accuracies of CR and paired CR and ES images for detecting thoracic bone lesions in patients with MM. Because CR and paired ES images are obtained at the same time in clinical practice, the ES images were not interpreted alone with the one-shot energy subtraction method. Patients with MM were enrolled in this study because they exhibited bone lesions and because whole-body radiographic and CT examinations are routinely performed for bone-lesion screening in these cases.

The results of this study had certain clinical implications. First, the accuracy of diagnosis of thoracic osteolytic lesions in patients with MM may be improved by adding ES images to conventional CR images. Second, ES images are especially useful for diagnosis of lesions of the middle thoracic spine.

Sensitivity for the entire thoracic spine was significantly improved by adding ES images to conventional CR images. This indicates that ES images may be particularly helpful in reducing the frequency of oversight of osteolytic lesions. The authors believe that the increased clarity of bone visualization by ES images improved the detection of osteolytic lesions. Conversely, although the specificity was improved, significant improvement in the diagnostic accuracy was found for only 1 observer. The authors believe that small changes in the specificity and the small sample size may have contributed to the lack of statistical significance in the differences among individual specificities. Intra- and interobserver reliability improved from 0.84 to 0.89, respectively, and 0.31 to 0.60, respectively, with the addition of ES images. Use of ES images reduced the intra- and interobserver variation in detecting osteolytic lesions in the radiographs and improved the reliability of the examination.

Subgroup analysis on the basis of location showed significant improvement in sensitivity in the middle thoracic spine for the CR+ES group; however, significant differences were not observed in individual assessments of the upper and lower thoracic spines. An earlier study had reported that lateral views of the upper and lower thoracic spines were less clear than those of the middle thoracic spine, even after processing the images by the energy subtraction method.2 Because the upper thoracic spine overlaps with the head of the humerus and the scapula, the vertebrae cannot be visualized separately. The lower thoracic spine overlaps with the diaphragm, which has higher density than the other soft tissues, including the heart and pulmonary vessels. Therefore, the difference in energy between the 2 imaging plates is minimized due to low radiograph transmission throughout this area, resulting in images with lesser clarity. Insufficient information obtained from the lateral view of the upper and lower thoracic spines compelled the observer to evaluate the presence of osteolytic lesions based almost exclusively on the information from the frontal views; this resulted in limited improvement in diagnostic accuracy for this area.

The sensitivity in detecting thoracic bone lesions was 17.2% in the CR group; although the observers were informed that the images were for bone-lesion screening of patients with MM, this value was considerably low. A further decrease in sensitivity is expected in clinical practice, where patients come prior to a diagnosis of MM. Approximately half of the patients with MM initially present to orthopedic departments and are often misdiagnosed and treated as cases of vertebral fractures or spinal degenerative disease.7 However, the combined use of ES with CR images increases the sensitivity of thoracic spine radiographs and may contribute to the early detection and treatment of MM. In addition to MM, ES images may effectively diagnose bone lesions in other pathologic conditions, such as spinal tumor, trauma, ossification, deformity, degeneration, and infection. Improvement in the quality of radiographs may minimize CT use, thereby reducing radiation exposure and medical expenses. Therefore, this one-shot energy subtraction method is of particular clinical value.

This study had several limitations. The relatively small sample could have under-powered the statistical analyses, leading to a lack of statistical significance for some comparisons. The interobserver reliability in this study was not high, indicating interobserver variations in interpretation. This may be addressed by replacing general orthopedic surgeons with more specialized physicians, such as radiologists, orthopedic oncology surgeons, and spine surgeons. This study exclusively evaluated osteolytic bone lesions in patients with MM. In addition to osteolytic lesions, it is desirable to evaluate the clinical utility of this method for other conditions, including osteoblastic lesions, trauma, infection, and ossification. However, because this study enrolled consecutive patients with MM, including cases with and without lesions, cognitive bias in interpretation of the images was avoided and the utility of the one-shot energy subtraction method was evaluated in a manner relevant to daily clinical practice.

Conclusion

Thoracic spine radiographs generated using the one-shot energy subtraction method improved the accuracy of diagnosis of thoracic osteolytic lesions in patients with MM. The clinical utility of this method was remarkably high in the middle thoracic spine. Because it improves the diagnostic accuracy of CRs without increasing the burden on patients and radiological technicians, this method will be of particular value in clinical practice.

References

  1. Berrington de González A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med. 2009;169(22):2071–2077. doi:10.1001/archinternmed.2009.440 [CrossRef] PMID:20008689
  2. Sasagawa T, Kunogi J, Masuyama S, et al. The clinical utility of a one-shot energy subtraction method for thoracic spine radiography. J Orthop Sci. 2012;17(4):346–351. doi:10.1007/s00776-012-0220-1 [CrossRef] PMID:22476393
  3. Ishigaki T, Sakuma S, Horikawa Y, Ikeda M, Yamaguchi H. One-shot dual-energy subtraction imaging. Radiology. 1986;161 (1):271–273. doi:10.1148/radiology.161.1.3532182 [CrossRef] PMID:3532182
  4. Ishigaki T, Sakuma S, Ikeda M. One-shot dual-energy subtraction chest imaging with computed radiography: clinical evaluation of film images. Radiology. 1988; 168(1):67–72. doi:10.1148/radiology.168.1.3289096 [CrossRef] PMID:3289096
  5. Ishigaki T, Itoh K, Sakuma S. [Radiological assessment of pulmonary function by computed radiography with imaging plates]. Nihon Kyobu Shikkan Gakkai Zasshi. 1989;27(3):256–262. PMID:2615080
  6. Takashima T. [Clinical evaluation of single-exposure dual-energy subtraction chest radiography: with FCR 9501 ES]. Nihon Igaku Hoshasen Gakkai Zasshi. 1996;56(13):909–916. PMID:8969053
  7. Kyle RA, Gertz MA, Witzig TE, et al. Review of 1027 patients with newly diagnosed multiple myeloma. Mayo Clin Proc.2003;78(1):21–33. doi:10.4065/78.1.21 [CrossRef] PMID:12528874

Distribution of the Osteolytic Lesions Detected on Computed Tomography at Each Thoracic Vertebral Levela

Subgroup/locationNo.

Osteolytic lesion(s)Total
Upper thoracic spine
  T1210
  T23
  T32
  T43
Middle thoracic spine
  T5312
  T60
  T71
  T83
  T92
  T103
Lower thoracic spine
  T1136
  T123
Total28

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions Relative to a Computed Tomography Gold Standard

ObserverCR groupCR+ES group


SensitivitySpecificitySensitivitySpecificity
A17.9%a97.1%60.7%a98.9%
B10.7%a97.1%53.6%a99.3%
C17.9%a93.4%60.7%a96.7%
D25.0%97.8%46.4%97.8%
E14.3%a92.6%a50.0%a97.1%a
Overall17.2%a95.6%a54.3%a98.0%a

Sensitivity and Specificity for the CR and CR+ES Groups in Detecting Osteolytic Lesions of Each Subgroup Classified on the Basis of Location

Observer/locationCR groupCR+ES group


SensitivitySpecificitySensitivitySpecificity
Upper thoracic spine (T1-T4)
  A0%89.6%66.7%93.6%
  B0%89.9%100%90.9%
  C0%89.9%60.0%92.6%
  D0%90.0%100%90.9%
  E0%89.5%50.0%90.8%
  Overall0%a89.8%a75.3%a91.8%a
Middle thoracic spine (T5-T10)
  A33.3%a94.4%91.7%a99.3%
  B25.0%a93.6%91.7%a99.3%
  C41.7%94.9%75.0%97.8%
  D33.3%a94.4%91.7%a99.3%
  E33.3%a94.2%91.7%a99.3%
  Overall33.3%a94.3%a88.4%a99.0%a
Lower thoracic spine (T11-12)
  A50.0%89.6%100%91.7%
  B0%87.5%60.0%93.3%
  C11.1%87.8%44.4%95.1%
  D66.7%91.5%33.3%89.4%
  E0%86.0%28.6%90.7%
  Overall25.6%88.5%53.3%92.0%
Authors

The authors are from the Department of Orthopaedic Surgery (TShimizu, HT), Graduate School of Medical Sciences, Kanazawa University, Kanazawa; the Department of Orthopedic Surgery (TSasagawa), Toyama Prefectural Central Hospital, Toyama; the Department of Spine and Orthopedic Surgery (NK, SM, NT, HE, MI, HI, JK) and the Department of Hematology (KS), Japanese Red Cross Medical Center, Tokyo; and the Fujifilm Holdings Corporation (TK), Tokyo, Japan.

The authors have no relevant financial relationships to disclose.

The authors thank Editage ( www.editage.jp) for English language editing.

Correspondence should be addressed to: Takaki Shimizu, MD, Department of Orthopaedic Surgery, Graduate School of Medical Sciences, Kanazawa University, 13-1 Takaramachi, Kanazawa 920-8641, Japan (takaki. shimizu0928@gmail.com).

Received: August 15, 2019
Accepted: December 13, 2019
Posted Online: December 07, 2020

10.3928/01477447-20201202-05

Sign up to receive

Journal E-contents