Value of third-generation of VNCa dual-energy CT for differentiating diffuse marrow infiltration of multiple myeloma from red bone marrow

Abstract

This study aims to investigate the ability of bone marrow imaging using third-generation dual-energy computed tomography (CT) virtual noncalcium (VNCa) to differentiate between multiple myeloma (MM) with diffuse bone marrow infiltration and red bone marrow (RBM). Bone marrow aspiration or follow-up results were used as reference. We retrospectively reviewed 188 regions of interests (ROIs) from 21 patients with confirmed MM and diffuse bone marrow infiltrations who underwent VNCa bone marrow imaging between May 2019 and September 2022. At the same time, we obtained 98 ROIs from 11 subjects with RBM for comparative study, and 189 ROIs from 20 subjects with normal yellow bone marrow for the control group. The ROIs were delineated by 2 radiologists independently, the interobservers reproducibility was evaluated by interclass correlation coefficients. The correlation with MRI grade results was analyzed by Spearman correlation coefficient. Receiver operating characteristic (ROC) curve analysis was used to determine the optimal threshold for differentiating between these groups and to assess diagnostic performance. There were statistically significant differences in VNCa CT values of bone marrow among the MM, RBM, and control groups (all P < .001), with values decreasing sequentially. A strong positive rank correlation was observed between normal bone marrow, subgroup MM with moderately and severe bone marrow infiltration divided by MRI and their corresponding CT values (ρ = 0.897, 95%CI: 0.822 to 0.942, P < .001). When the CT value of VNCa bone marrow was 7.15 HU, the area under the curve (AUC) value for differentiating RBM and MM was 0.723, with a sensitivity of 50.5% and a specificity of 89.8%. When distinguishing severe bone marrow infiltration of MM from RBM, the AUC value was 0.80 with a sensitivity 70.9% and a specificity 78.9%. The AUC values for MM, RBM, and the combined group compared to the control group were all >0.99, with all diagnostic sensitivity and specificity exceeding 95%. VNCa bone marrow imaging using third-generation dual-energy CT accurately differentiates MM lesions from normal bone marrow or RBM. It demonstrates superior diagnostic performance in distinguishing RBM from MM with diffuse bone marrow infiltration.

1. Introduction

Multiple myeloma (MM) is a prevalent hematological malignancy characterized by the accumulation of monoclonal plasma cells in the bone marrow, producing excessive monoclonal immunoglobulins and resulting in systemic multi-organ damage.^[1,2] Based on imaging examinations, the Durie/Salmon plus staging system is commonly used for MM staging.^[3,4] Imaging allows for quick and intuitive assessment of the number and distribution of systemic lesions, particularly in hyposecretory or non-secretory types of MM.^[5] As observed on imaging, diffuse bone marrow infiltration is closely associated with a high tumor burden and serves as an independent prognostic factor within the Durie/Salmon plus staging system.^[4]

The 2014 International Myeloma Working Group (IMWG) guidelines recommend whole-body CT, MRI, or positron emission tomography/computed tomography for evaluating MM bone lesions.^[6,7] Among these modalities, MRI, which utilizes hydrogen proton imaging, exhibits superior sensitivity in detecting bone marrow lesions rich in water and fat.^[8,9] Normal adult bone marrow primarily consists of yellow bone marrow, comprising approximately 80% fat, 15% water, and 5% protein. It demonstrates higher signal intensity than muscle on T1-weighted imaging (T1WI).^[8,10] In MM with diffuse bone marrow infiltration, there is a diffuse decrease in signal intensity on T1WI due to increased myeloma cells and decreased fat content. However, under certain benign or physiological conditions, yellow bone marrow can convert to red bone marrow (RBM), which contains less fat, resulting in a similar diffuse decrease in signal intensity on T1WI. These conditions may include proliferative anemia, smoking, and high-intensity exercise.^[8,10] Therefore, except for invasive bone marrow aspiration or long follow-up, conventional MRI imaging is still not fully accurate in distinguishing diffuse bone marrow infiltration of MM from diffuse RBM.

Dual-energy CT virtual noncalcium (VNCa) bone marrow imaging is broadly applied in bone marrow assessment currently. By removing the high attenuation of trabeculae with material separation technology, the generated VNCa image allows for visual assessment of bone marrow attenuation through color coding while providing a superimposed calcium-containing gray map to evaluate both bone marrow and bone; at the same time, it allows for quantitative analysis by measuring CT values.^[11,12] Thomas et al^[11] reported a detection rate of localized MM lesions without bone destruction of 75% using VNCa bone marrow imaging, with MRI serving as the reference standard, compared 0% using traditional CT. VNCa bone marrow imaging has also demonstrated the ability to depict various forms of bone marrow infiltration observed on MRI.^[13] However, there are few studies on the application of this technique to the identification of diffuse MM and RBM.

In this study, we aim to investigate the ability of bone marrow imaging using third-generation dual-energy CT VNCa to differentiate between MM with diffuse bone marrow infiltration and RBM. Follow-up results or bone marrow aspiration were used as reference. On the one hand, for patients with diagnosed MM, D-S staging can be performed more accurately. On the other hand, for patients with sporadic diffuse bone marrow signal reduction on T1WI, it may provide clinicians with evidence on whether the patient needs active management.

2. Materials and methods

2.1. Subjects

All subjects provided informed consent, and the study was approved by the Ethics Committee of our hospital. A total of 21 MM patients with diffuse bone marrow infiltration were enrolled between May 2019 and September 2022. The group consisted of 16 males and 5 females, with a median age of 61.54 ± 9.28 years. All of them were newly diagnosed with active or symptomatic MM by the diagnostic criteria established by the IMWG in 2014.^[6] The primary treatment modalities included bisphosphonate therapy, adjuvant therapy, and autologous stem cell transplantation for 2 patients. Diffuse bone marrow infiltration of MM was defined as follows: Firstly, all of them were newly diagnosed with active or symptomatic MM by the diagnostic criteria established by the IMWG in 2014; Secondly, the MRI T1WI signal of the vertebral body was uniformly lower than or similar to the signal of the vertebral space; Lastly and importantly, follow-up with progression or bone marrow aspiration demonstrated bone marrow infiltration.^[11,14]

We selected 306 patients over 50 years old who underwent lumbar MRI examination due to low back pain from March 2021 to October 2022 and underwent lumbar dual-source CT examination during the same period (interval <1 week), excluding trauma, tumor, and inflammatory lesions. Among these patients, we screened 20 normal yellow bone marrow patients as the control group, 14 males and 6 females, with a median age of 69.3 ± 6.5 years. The screening conditions were that the MRI T1WI signal of the vertebral body was uniformly higher than that of the intervertebral disc or similar to that of subcutaneous fat. 11 patients were selected as the RBM group, 7 males and 4 females, with a median age of 58.6 ± 4.7 years (Table 1). The screening criteria of RBM were as follows: the MRI T1WI signal of the vertebral body was uniformly lower than or similar to the signal of the vertebral space; Follow-up for more than 1 year to exclude tumor and hematological diseases.^[15–17]

Table 1.

Demographic and clinical characteristics of patients.

	MM (n = 21)	RBM (n = 11)	Control (n = 20)	P value
Males	16	7	14	P = .750*
Females	5	4	6
Age (yr)	61.54 ± 9.28	58.6 ± 4.7	69.3 ± 6.5	P < .05†
Clinical characteristics
	IgG type (n = 7); IgA type (n = 4); IgG type (n = 3); Lighe chain type (n = 2)	discogenic pain (n = 5); myofascial injury (n = 2); both (n = 4)	discogenic pain (n = 13) myofascial injury (n = 3); both (n = 4)

^*The sex distribution between groups was tested by Chi-square test, Pearson chi-square = 0.575.

^†Group differences in age were tested by one-way analysis of variance (ANOVA), MM vs RBM: P = .238; MM vs Control: P = .01; RBM vs Control: P = .000.

Because both MM and RBM had reduced T1WI signals, they were sometimes combined as signal reduction group later in this study.

2.2. Dual-energy CT and the post-processing technology

All subjects underwent dual-energy CT scans using a third-generation dual-energy CT scanner (Somatom Force; VA50, Siemens Healthineers, Erlangen, Germany). Non-enhanced single- or multi-site scans were performed from the cervical to the pelvis, with the hands raised above the top of the skull. The scanning parameters were as follows: collimator, 128 × 0.6 mm; pitch, 0.6; rotation time, 0.5 seconds. The tube voltage was adjusted based on the patient body weight. For body weights below 90 kg, the tube voltage ranged from 90 to 150 kV, using a 0.6 mm tin filter. For body weights of 90 kg or more, the tube voltage ranged from 100 to 150 kV, using a 0.6 mm tin filter. The tube current ratio at the peak of 90 kV and 150 kV with a tin filter was 1.6:1, with 220 milliamperes (mAs) for tube A and 138 mAs for tube B. The tube current ratio at the peak of 100 kV and 150 kV with a 0.6 mm tin filter was 2.0, with 276 mAs for tube A and 138 mAs for tube B. Tube current modulation based on automatic attenuation (Care dose 4D; Siemens Healthineers, Erlangen, Germany) was used. The mean volume CT dose index was 7.7 mGy ± 3.8 (range, 4.1–18.3 mGy), and the mean dose-length product was 1181.9 mGy · cm ± 517.6 (range, 489–2113mGy · cm).

Post-processing of bone marrow imaging was performed using a dedicated workstation (Syngo.via, VB10, Siemens Healthineers, Erlangen, Germany). The bone marrow post-processing application was employed, which utilizes a noise reduction algorithm and bone segmentation algorithm for thresholding and erosion. A color overlay image was then generated, with VNCa color-coded and superimposed on a grayscale calcium content map, allowing the fusion of bone marrow attenuation content with the anatomical image. Blue color represented negative attenuation, indicating normal yellow bone marrow, while red color indicated positive attenuation, representing enhanced decay of RBM or other bone marrow space material. All images were analyzed and evaluated using thin-slab maximum intensity projection mode with a layer thickness of 5 mm. The color lookup table for the overlays was consistent for all patients, with a window width of 200 and a window level of −50 (Fig. 1B, D, F, H).

Figure 1.

Manifestations of T1WI of traditional MRI and VNCa in different groups. (A) Normal signal of yellow bone marrow on T1WI is higher than intervertebral disc and similar to that of fat. MM with moderate diffuse bone marrow infiltration (C*) presents with decreased signal on T1WI but remains higher than that of the intervertebral disc. The signals of MM with severe diffuse bone marrow infiltration (E) and RBM reverse conversion T1WI signal (G) are similar, both lower than that of the intervertebral disc, similar to that of muscle, which is difficult to distinguish between them. (B), (D), (F), and (H) are Dual-Energy CT images with VNCa color overlay, in which blue and red pixels indicate negative and positive attenuation, corresponding to (A), (C), (E), and (G), respectively. L1 vertebral compression fracture. CT = computed tomography, MM = multiple myeloma, MRI = magnetic resonance imaging, ROI = region of interest, T1WI = T1-weighted imaging, VNCa = virtual noncalcium.

2.3. MRI protocol

All subjects underwent MRI scans with a body coil using a Magnetom Avanto 1.5 T MR imaging system (Siemens Healthineers, Erlangen, Germany). The conventional scanning sequences included sagittal T1WI, T2-weighted imaging (T2WI), and fat suppression Turbo Inversion Recovery Magnitude (TIRM). The scanning parameters for the spine were as follows: T1WI, repetition time/echo time (TR/TE) = 537 ms/8.8 ms; T2WI, TR/TE = 3600 ms/101 ms; TIRM, TR/TE = 4200 ms/50 ms. The field of view was set at 320 mm, layer thickness was 4 mm, and the number of layers was 11. Additionally, axial T2WI sequences were selectively included.

The MRI signals of diffuse bone marrow infiltration in MM were categorized into 3 grades based on Baur and Staebler methods.^[18,19] Normal or mild infiltration refers to bone marrow signal intensity on T1WI that is normal or slightly lower than that of subcutaneous fat. Moderate infiltration indicates a reduction in signal intensity on T1WI, but it remains higher than that of the intervertebral disc. Severe infiltration is defined when the signal intensity of bone marrow on T1WI is similar to or lower than that of the intervertebral disc (Fig. 1A, C, E). The “pepper sign” without localized lesions on TIRM represents a lower tumor burden, typically classified as mild infiltration.^[20] Normal or only a slight decrease in bone marrow signal intensity on T1WI does not affect the Durie/Salmon plus staging of MM, and it is easily distinguishable from RBM with significantly reduced signal. Therefore, this study selected subjects with moderate or severe bone marrow infiltration, which is difficult to distinguish from RBM during visual evaluation (Fig. 1E, G).

2.4. Drawing region of interest (ROI)

The ROI on the VNCa bone marrow pseudo-color map was delineated by 2 radiologists with 10 and 8 years of MR diagnostic experience independently (W TT and Y C). ROI was sketched for each vertebra from T10 to S1, some patients’ image showed better ilium bones, and ROI was also sketched after negotiation. Both radiologists were blinded to the diagnosis of patients at the time of measurement.

The following principles guided ROI delineation: avoid areas affected by the bone cortex, blood sinus, and hemangioma following MRI reference, as they can exhibit higher attenuation even in the presence of a normal vertebral body; select the delineation range in areas with relatively high attenuation, overlapped by yellow or red pseudo-color. The conversion of red and yellow bone marrow or the infiltration of malignant tumor cells is a gradual trade-off process, and the density or signal of bone marrow is not uniform. The ROI should be selected in the RBM or the high attenuation area of the lesion while artificially controlling the effects of low attenuation in yellow bone marrow or RBM containing more fat through visual observation. This approach aligns with the primary objective of the study. Lastly, while meeting the previous 2 conditions, efforts were made to outline a more considerable ROI to reduce errors and improve data repeatability. The area is >1.0 cm² at least.

2.5. Statistical analysis

Interclass correlation coefficients were calculated to evaluate the interobserver reproducibility of measuring ROI. The averages of the measurements from the 2 radiologists were used in further analyses. An interclass correlation coefficient value <0.60 indicates poor or moderate agreement; 0.61 to 0.80, good agreement; and 0.81 to 1.00, excellent agreement.

One-way analysis of variance was employed to compare the measurement data among different groups. The Spearman correlation coefficient was used to evaluate the correlation between the measurement and ordinal data. Receiver operating characteristic (ROC) curve analysis was performed to determine the optimal thresholds for different data sets and their corresponding sensitivity, specificity, and area under the curve (AUC). An AUC value <0.7 indicates a low diagnostic value, an AUC of 0.7 to 0.9 indicates moderate accuracy, and an AUC >0.9 indicates a high diagnostic value. All statistical analyses were conducted using SPSS 17.0 statistical software (Armonk, New York), with a significance level of P < .05.

3. Results

3.1. Statistical description

188 ROIs were obtained from 21 MM patients with diffuse bone marrow infiltrations, the mean VNCa CT value of bone marrow was 6.45, (95%CI: 4.02, 8.88). 98 ROIs were obtained from11 subjects, the mean VNCa CT value of bone marrow was −5.76 (95%CI: −8.04, −3.48). 189 ROIs were obtained from 20 subjects with normal yellow bone marrow, the mean VNCa CT value of bone marrow was −62.68, (95%CI: −65.34, −60.01) (Table 2). The CT values of MM, RBM, and control groups were decreased sequentially, with statistically significant differences. The interobserver agreements were excellent. The interclass correlation coefficient for the MM group was 0.908; for the RBM group, 0.890; for the control group, 0.948.

Table 2.

The VNCa CT values of bone marrow from different groups.

Groups	ROIs	Mean (HU)	95% CI
			Lower bound	Upper bound
Control	189	−62.68	−65.34	−60.01
RBM	98	−5.76	−8.04	−3.48
MM with moderate DI	71	−0.10	−3.48	3.28
MM with severe DI	117	10.43	7.28	13.57
MM	188	6.45	4.02	8.88

CI = confidence interval, CT = computed tomography, DI = diffuse infiltration, HU = Hounsfield unit, MM = multiple myeloma, RBM = red bone marrow, ROI = region of interest, VNCa = virtual noncalcium.

3.2. Rank correlation between VNCa CT values and MM grades of MRI

According to MRI, MM was divided into 3 grades: normal, moderate and severe bone marrow infiltration (the control group was treated as normal bone marrow in this study). The corresponding VNCa CT values of bone marrow was increased sequentially. The Spearman correlation coefficient between the 2 was 0.897 (95%CI: 0.822–0.942, P < .001).

3.3. ROC analysis

The VNCa CT values of bone marrow in the control group were significantly lower than those in the MM, RBM, and the combination of them (P < .001). The AUC values between the control group and the preceding groups were 0.993, 0.997, and 0.996, respectively, indicating excellent diagnostic values. The differences between RBM, MM, and subgroups of MM were also significant (all P < .001) (Table 3). At a cutoff CT value of 7.15HU for VNCa bone marrow, the AUC value for differentiating RBM and MM was 0.723, suggesting a specific diagnostic value. While in the differentiation between RBM and subgroup MM with severe bone marrow infiltration, the AUC value increased to 0.80 at a cutoff CT value of 8.0 HU, indicating an improved diagnostic value.

Table 3.

The pairwise comparison of bone marrow CT Values between different groups using ROC analyses.

Intergroup comparison	AUC	P	Youden index	cutoff value (HU)	Sensitivity (%)	Specificity (%)
MM vs Con Group	0.997	.001	0.958	−25.85	0.995	0.963
RBM vs Con Group	0.993	.001	0.952	−29.85	1	0.952
Signal Reduction* vs Con Group	0.996	.001	0.956	−25.95	0.993	0.963
MM vs RBM	0.723	.001	0.403	7.15	0.505	0.898
MM with severe DI vs RBM	0.80	.001	0.498	1.8	0.709	0.789
MM with moderate DI vs RBM	0.609	.001	0.249	12.1	0.31	0.939
MM with moderate DI vs severe DI	0.672	.001	0.31	−0.35	0.761	0.549

AUC = area under the curve, CI = confidence interval, CT = computed tomography, DI = diffuse infiltration, MM = multiple myeloma, RBM = red bone marrow, ROI = region of interest, VNCa = virtual noncalcium.

^*The signal reduction group was the combination of MM group and RBM group.

However, the AUC values for differentiating RBM and subgroup MM with moderate bone marrow infiltration, differentiating subgroup MM with differentiating moderate and subgroup MM with severe bone marrow infiltration were <0.7, indicating lower diagnostic values (AUC = 0.609, AUC = 0.672) (Fig. 2).

Figure 2.

ROC analyses of different groups. (A) RBM vs Control Group; (B) MM vs Control Group; (C) Signal Reduction* vs Control Group; (D) MM with moderate DI vs RBM Group; (E) MM with severe DI vs RBM Group; (F) MM vs RBM Group. * The signal reduction group was a combination of MM and RBM groups. DI = diffuse infiltration, MM = multiple myeloma, RBM = red bone marrow, ROC = receiver operating characteristic.

4. Discussion

In this study, we demonstrated VNCa bone marrow imaging using third-generation dual-energy can identify bone marrow states with different T1WI signal levels by quantitative analysis, and can well distinguish RBM from MM, especially MM with severe bone marrow infiltration. At a CT value of 7.15HU for VNCa bone marrow, the AUC value for differentiating RBM and MM was 0.723, with a sensitivity of 50.5% and a specificity of 89.8%. When distinguishing severe bone marrow infiltration of MM from RBM, the AUC value increased to 0.80, with a sensitivity of 70.9% and a specificity of 78.9%. The AUC values for MM, RBM, and their combined signal reduction group with the control group were all >0.99 (ranging from 0.993 to 0.997), with diagnostic sensitivities and specificities exceeding 95% (ranging from 95.2% to 100%).

Currently, VNCa bone marrow imaging is primarily used to assess bone marrow edema in the axial and appendicular skeleton, such as diagnosing bone marrow edema without displaced fractures in trauma,^[21–24] with diagnostic thresholds ranging from −13HU to −80HU.^[21,22] Some studies have compared MM with a normal control group to detect MM lesions. Thomas et al^[11] reported detection thresholds of −3HU and 4HU for bone marrow imaging to detect MM lesions without or with bone destruction, respectively, using MRI as the reference standard. The sensitivities were 78.9% and 92.6%, and the specificities were 94.8% and 89.2%, respectively. Kosmala et al^[25] found that MM lesions without distinguishing bone destruction could be detected well when the CT value of VNCa bone marrow imaging of localized lesions was -49HU, with a diagnostic sensitivity of 93.3% and specificity of 92.4%. In a subsequent study by Kosmala et al,^[13] they reported that the CT value of localized lesions on VNCa bone marrow imaging was higher than that of diffuse lesions in MM, and the difference between them was significant. The threshold for detecting diffuse bone marrow infiltration of MM was −35.7HU, with a sensitivity of 100% and a specificity of 97%. In this study, the diagnostic threshold for diffuse bone marrow infiltration of MM and the normal control group was −25.85HU, which is more similar to the results reported by Kosmala et al.^[13] The similarity in machine used and generally consistent imaging parameters between the studies, as well as potential differences in ROI delineation regions between study subjects and the control group, may contribute to the slight discrepancy.

The second-generation dual-energy CT (Somatom Definition Flash, Siemens Healthineers) used by Thomas et al^[11] employed parameters of 100 kV for low energy and 140 kV for high energy, with a tin filter used to harden the high energy spectrum and increase the spectral separation between the high- and low energy spectra. Kosmala et al^[25] utilized the same third-generation dual-energy CT (Somatom Force, Siemens Healthineers) as this study, with parameters of 90 kV for low energy and 150 kV for high energy, along with a stronger tin filter (0.6 mm). This approach reduced spectral overlap compared to the second-generation dual-energy CT and improved the accuracy of material separation. Although Kosmala et al^[25] and this study both used third-generation dual-energy CT with similar parameters, there were differences in the study subjects. In Kosmala et al previous study,^[25] patients had localized bone marrow infiltration patterns, and the control group included pelvises with higher fat content in the ROI delineation, resulting in a lower threshold.

However, in their later study,^[13] they found no significant difference in the CT values of VNCa bone marrow between the normal healthy control group and lesion-free areas of MM patients. They distinguished between localized and diffuse lesions and discovered that the detection threshold for diffuse lesions was higher than that for previously localized lesions. In this study, the demarcation threshold was further increased compared to the previous studies.^[13,25] We speculate that this may be due to the exclusion of the pelvis in ROI delineation in the control group and the majority (62.2%) of patients having severe diffuse bone marrow infiltration in MM (with higher CT values in patients with severe diffuse bone marrow infiltration compared to those with moderate infiltration in this study). Although the demarcation thresholds in these studies are not identical, they all conclude that bone marrow lesions of MM can be effectively detected using dual-energy CT VNCa bone marrow imaging.

In imaging diagnosis, it is often challenging to distinguish malignant bone marrow lesions from benign red myelohyperplasia or reverse conversion, even with conventional MRI, which is the most sensitive modality for observing signal changes in the bone marrow. The results of this study demonstrate that dual-energy CT VNCa bone marrow imaging can effectively differentiate RBM from MM with diffuse bone marrow infiltration when the threshold is set at 7.15 HU, with an AUC value of 0.72, a sensitivity of 50.5%, and a specificity of 89.8%. When distinguishing severe bone marrow infiltration of MM with a higher CT value from RBM, the AUC increases to 0.80, with a sensitivity of 70.9% and a specificity of 78.9%. Currently, there are no research reports on using dual-energy CT VNCa bone marrow imaging technology to differentiate MM with diffuse bone marrow infiltration from RBM, whereas MRI is the most commonly used modality for obtaining Signal Intensity Ratios or Fat Fraction through water and lipid imaging to measure changes in bone marrow signal quantitatively.^[26–28] Akman B used these techniques to differentiate the reverse conversion of RBM from bone marrow malignant lesions and reported an AUC of 0.825 when using a signal intensity ratios of 0.82 as the optimal threshold.^[17] Although the diagnostic performance of quantitative MRI measurement is higher than that of dual-energy CT in current research, the longer examination time of MRI and contraindications such as intrareceptor implants and life support devices limit its application. Dual-energy CT VNCa bone marrow imaging technology can be considered an option depending on the specific clinical situation.

There are some limitations to this study. Firstly, this study did not consider the treatment condition of MM patients with bone marrow infiltration, which may affect the CT values of bone marrow. Secondly, although the difference in CT values between MM with moderate and severe bone marrow infiltration groups was significant in bone marrow imaging, the AUC value was low (<0.7), and the diagnostic performance was poor. Therefore, it could not effectively distinguish the degree of diffuse bone marrow infiltration with different tumor burdens, further affecting the accuracy of Durie/Salmon plus staging.^[3,29] In future studies, adopting machine learning with texture analysis may improve identification accuracy,^[30] and increasing the analytical dimension or sample size and adding single-energy post-processing data to form different energy spectrum curves could be considered.

5. Conclusions

VNCa bone marrow imaging on the third-generation dual-energy CT accurately distinguishes MM lesions from normal bone marrow and has a specific value in differentiating RBM from MM with diffuse bone marrow infiltration.

Author contributions

Conceptualization: Tiantian Wang, Kui Zhang, Chen Yan, Xiangzhen Guan.

Data curation: Biao Zhou.

Formal analysis: Tiantian Wang.

Investigation: Biao Zhou, Chen Yan.

Project administration: Kui Zhang, Xiangzhen Guan.

Resources: Chen Yan.

Software: Chen Yan.

Validation: Biao Zhou, Kui Zhang, Xiangzhen Guan.

Writing – original draft: Tiantian Wang.

Writing – review & editing: Tiantian Wang.