Value of Spectral CT Techniques for the Assessment of Bone Marrow Infiltration in Multiple Myeloma: A Systematic Literature Review

Abstract

Background

Multiple myeloma (MM) is the second most common hemato-oncological malignancy, characterized by the clonal proliferation of malignant plasma cells and bone marrow infiltration. The degree of bone marrow infiltration, which is crucial for diagnosis and treatment initiation, is determined through biopsy. While MRI and CT are considered standard imaging methods for detecting focal lesions and osteolytic changes, CT has limitations, particularly in detecting diffuse infiltration patterns without osteolysis. Spectral CT techniques offer a promising alternative for assessing bone marrow infiltration through material decomposition.

Method

A systematic literature search was conducted in the PubMed database for relevant keywords in articles published between 01/2010 and 12/2024. Original studies evaluating spectral CT techniques for the assessment of MM bone marrow infiltration were included. Articles with a different focus, such as fracture detection, were excluded. A qualitative synthesis of the study results was performed.

Results and Conclusion

Spectral CT techniques improve the differentiation between healthy and infiltrated bone marrow. Particularly, the commonly applied virtual calcium suppression showed good sensitivity and specificity compared to histology, serology, or MRI. Spectral CT also shows potential for distinguishing different bone marrow infiltration patterns, assessing disease activity, and evaluating treatment response. Limitations included reduced sensitivity for detecting moderate infiltration within red bone marrow and small cohort sizes. Multicenter analyses are required to compare different device manufacturers, evaluate the utility of spectral CT biomarkers, the potential of currently less intensively studied material density maps and radiomics features, as well as, of photon-counting CT.

Key Points

• Spectral CT techniques can detect bone marrow infiltration in MM and allow for differentiation of infiltration patterns in CT imaging.

• Spectral CT parameters appear to have potential as biomarkers for tumor activity and treatment response.

Citation Format

• Melzer YF, Erley J, Ristow I et al. Value of Spectral CT Techniques for the Assessment of Bone Marrow Infiltration in Multiple Myeloma: A Systematic Review. Rofo 2026; 198: 36–54

Keywords

myeloma - bones - marrow - spine

Introduction

Multiple myeloma (MM) is the second most common hemato-oncological systemic disease, characterized by a clonal proliferation of malignant plasma cells within the bone marrow [1]. This proliferation disrupts the bone marrow microenvironment and leads to a variety of clinical symptoms and complications [1]. The most frequent manifestations include anemia, increased susceptibility to infections, hypercalcemia, and characteristic osteolytic lesions that may cause bone pain or pathological fractures [1].

The diagnosis of multiple myeloma is based on the SLiM-CRAB criteria, which encompass hypercalcemia (C), renal impairment (R), anemia (A), and bone lesions (B) [2]. In addition, specific biomarkers are considered, such as elevated serum free light chains and the extent of plasma cell infiltration in the bone marrow [2]. Imaging plays a central role in disease staging and therapy monitoring [3].

Magnetic resonance imaging (MRI) and computed tomography (CT) are key modalities for detecting both bone marrow infiltration and osteolytic lesions [3]. MRI is regarded as the gold standard due to its superior soft-tissue contrast and high sensitivity for detecting diffuse infiltration patterns [4]. Conventional CT, while offering excellent visualization of bone destruction, has limited capability for assessing the bone marrow itself [5]. According to the German S3 guideline, the detection of more than one typical myeloma lesion > 5 mm on MRI is diagnostically relevant – even in the absence of mineralized bone destruction [6]. Such lesions, however, often remain undetectable with conventional CT techniques.

This limitation has spurred interest in advanced spectral CT methods, such as dual-source and dual-layer detector CT, for bone marrow assessment [5] [7]. Spectral CT enables the differentiated visualization of various tissues and materials. By exploiting the atomic number dependence of the photoelectric effect and the energy-specific attenuation properties of calcium, fat, and soft tissue, virtual calcium subtraction (VNCa) can selectively eliminate bone mineral from CT images [8]. This enhances the ability to differentiate between healthy and infiltrated bone marrow with greater sensitivity. An additional advantage is the extraction of quantitative parameters from bone marrow, which may serve as imaging biomarkers for tumor burden and disease monitoring. For example, texture-based features derived from calcium-subtracted images have shown potential diagnostic and prognostic value [9] [10]. Spectral CT may thus play a particularly important role in patients for whom MRI is not feasible or as part of routine CT-based therapy monitoring where osteolysis is already being assessed.

The aim of this systematic literature review is to examine current evidence on the use of spectral CT techniques for detecting and characterizing bone marrow infiltration in multiple myeloma.

Materials & Methods

The systematic literature review was conducted using PubMed, one of the world's leading biomedical databases. PubMed is maintained by the U.S. National Center for Biotechnology Information (NCBI), part of the National Institutes of Health (NIH), and provides access primarily to the MEDLINE database, which includes references from the fields of biomedicine and life sciences. As this review is based exclusively on previously published data, no ethics approval was required. A formal review protocol was not prepared. This manuscript follows the guidelines outlined in the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Appendix, Table 1) [11].

Search strategy

The search strategy was based on a combination of keywords related to myeloma and spectral CT techniques, including “myeloma” or “plasmocytoma” and “spectral CT”, “dual-energy CT” or “DECT” and “dual-layer CT”. To account for the increase in publications since 2010, the search was restricted to studies published between 2010 and December 2024.

The PubMed database was searched for relevant results in titles and abstracts. The exact search combinations using Boolean operators were as follows:

(myeloma AND spectral CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (myeloma AND DECT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (myeloma AND dual-energy CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (myeloma AND dual-layer CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (plasmocytoma AND DECT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (plasmocytoma AND dual-energy CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (plasmocytoma AND spectral CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication])); (plasmocytoma AND dual-layer CT) AND (("2014/01/01"[Date – Publication] : "3000"[Date – Publication]))

The value "3000" was used as an upper limit to ensure that all recent publications were included. Initially, 55 search results were retrieved.

Selection process

Inclusion and exclusion criteria were pre-specified before the review process began. Studies were included if they were original articles, focused on bone marrow infiltration in multiple myeloma, and applied spectral CT techniques. Studies were excluded if they were preprints, reviews, not available in English full text, or focused on unrelated topics (e.g., radiation dose comparisons). All search results were screened by one reviewer (FM) based on titles and abstracts to determine whether they met the predefined inclusion and exclusion criteria. Potentially eligible studies were then assessed in a second step using the full texts. Review articles identified during the primary search were excluded (n = 6), but their full texts and references were reviewed to identify additional potentially relevant studies. A second independent reviewer (IM) confirmed the final selection of studies. [Fig. 1] illustrates the study selection process in the form of a flowchart.

Fig. 1 PRISMA flowchart for study selection. No automation tools were used during study selection. *e.g. reviews on imaging used to illustrate body composition; **excluded because only dose comparison was performed, evaluation of calcified bone, focus on image quality in photon-counting CT, focus on radiation dose, focus on signal-to-noise ratio and osteolyses detection but no bone marrow assessment in photon-counting CT; ***e.g. excluded because extramedullary myeloma manifestations were examined, lack of focus on spectral CT, focus on AI-supported segmentation, focus on developing an AI model based on spectral CT lesion values, focus on AI methods to improve image quality, focus on fracture detection; ****excluded because bone tumors in general were studied, focus on AI models for fracture detection.

Assessment of study quality and risk of bias

The quality of all included studies was assessed using the QUADAS-2 tool (Quality Assessment of Diagnostic Accuracy Studies) [12] (Appendix, Table 2). This instrument was used to evaluate both the risk of bias and the applicability of each study in relation to the review question, based on four domains: patient selection, index test, reference standard, and flow and timing [12]. Bias may arise from systematic shortcomings in study methodology, while limited applicability may occur if, for example, a study investigates a different patient population or has objectives that do not align with the review question. The assessment was performed by a single reviewer (FM) without the use of automated tools.

Data synthesis

Two independent reviewers (FM, IM) extracted data from all included studies. Extracted variables included publication year, study objectives, cohort definitions, CT parameters relevant for material decomposition, the type of decomposition used, reference standards, measurement parameters, statistical methods, and outcomes relevant to the research questions. No discrepancies between the reviewers occurred. Study results were reported in the original formats, including absolute values, ROC AUC results with 95% confidence intervals, sensitivity, specificity, correlation coefficients, and p-values. These results are presented in detail in [Table 1] and [Table 2]. The findings were subsequently grouped and summarized based on shared research themes: the most commonly used calcium suppression technique (VNCa), the application of spectral CT in detecting bone marrow infiltration, and the evaluation of its diagnostic value in differentiating various infiltration patterns. Due to the heterogeneity of spectral techniques evaluated, a meta-analysis was not conducted. Similarly, no data conversions were performed, and inconsistent or missing outcome variables were excluded from the synthesis.

Table 1 Methodology of the studies included.

Author, year	Objectives	Cohorts	Spectral CT characteristics	Type of material decomposition	(MRI) comparative technique
Abbreviations: ADC = apparent diffusion coefficient, DECT = dual-energy computed tomography, DWI = diffusion imaging, FS = fat saturation, GE = General Electric, GRE = gradient echo, GSI = Gemstone Spectral Imaging, HD = high definition, ID = initial diagnosis, IMWG = International Myeloma Working Group, LVC = lumbar vertebral column, MGUS = monoclonal gammaopathy of undetermined significance (precursor to multiple myeloma), MM = multiple myeloma, RBM = red bone marrow, SE = spin echo, STIR = short tau inversion recovery sequence, SVB = sacral vertebral bodies, TIRM = turbo inversion recovery magnitude, TSE = turbo spin echo, TVB = thoracic vertebral bodies, VB10, VB20, VB30 = Syngo.via platform software versions, VNCa = virtual calcium subtraction
Chen et al., 2024	Detection of bone marrow infiltration and differentiation of infiltration patterns using monoenergetic and material density maps	n=41 patients and n=41 control subjects MM according to the IMWG criteria No osteolysis in the CT scan	Dual-source CT (Revolution CT scanner, GE Healthcare) Scan area: Thorax 80 and 140 kV, adaptive tube current modulation, native 70 keV reconstructions	Two-material decomposition with calcium or hydroxyapatite as dense material, and fat, water, and muscle as less dense material	1.5 T Thoracic spine T1w and STIR Purpose: to identify myeloid infiltration and its pattern (diffuse vs. focal)
Xiong et al., 2024	VNCa for the detection of bone marrow infiltration, prognosis of tumor risk profile and disease severity	n=47 untreated myeloma patients Histologically confirmed myeloma	Dual-layer detector spectral CT (IQon Spectral CT, Philips Healthcare) Scan area: Skull to knee 120 kV, 70 mAs, native	Calcium subtraction (VNCa maps) for a suppression index in 10% increments from 25% retained calcium to 95% Manufacturer software without further details on decomposition (IntelliSpace Portal, Spectral Diagnostics Suite, Philips Healthcare)	3 T Head to thigh TIRM and DWI Serum free light chain ratio Histology: Plasma cell infiltration rate, cytogenetic status (high-risk (del(17p), t(4;14), t(14;16), t (14;20), gain(1p), p53 mutation))
Wang et al.,2024	VNCa for differentiation of diffuse bone marrow infiltration from red bone marrow	From n=306 patients with lumbar spine MRI were selected: n=21 patients with diffuse bone marrow infiltration (newly diagnosed active or symptomatic MM according to IMWG criteria, 2014) n=11 patients with red bone marrow n=20 patients with yellow bone marrow	Dual-source CT (Somatom Force, Siemens Healthineers) Scan area: Neck/thorax/abdomen <90 kg: 90 and Sn150 kV, 220 mAs (tube A), 138 mAs (tube B) > 90 kg: 100 and Sn150 kV, 276 mAs (tube A), 138 mAs (tube B)	Calcium subtraction via manufacturer software without further information on decomposition (Syngo.via, VB10, Siemens Healthineers)	1.5 T Spine T1, T2, TIRM Purpose: Identification of myeloma patients with diffuse bone marrow infiltration or healthy individuals with red bone marrow, defined by homogeneously more iso or hypointense T1w signal intensity within the vertebral bodies compared to the intervertebral disc spaces, as well as patients with yellow bone marrow (T1 hyperintense bone marrow compared to intervertebral discs or isointense to subcutaneous fat). Furthermore, confirmation of bone marrow infiltration via bone marrow aspiration and follow-up examinations
Jiang et al., 2024	Fat density maps for the detection of bone involvement in the absence of CT correlates	MM patients n=32 (newly diagnosed or suspected MM according to IMWG criteria) without detectable bone lesions on conventional CT Control group: n=64 patients without MM, with back pain and/or lumbar symptoms and CT chest, abdomen	Rapid-switching CT (Revolution CT scanner, GE Healthcare) Scan area: Thorax, abdomen 80–140 kV, automatic tube current modulation, native	Hydroxyapatite subtraction maps = fat density maps and 70 keV maps via manufacturer software without further details on decomposition (GSI Viewer, GE Healthcare)	No MRI reference technique MM was diagnosed using the IMWG criteria.
Brandelik et al., 2021	Assessment of VNCa for detecting bone marrow infiltration and distinguishing infiltration patterns	n=32 patients (n=27 with MM, n=4 with smoldering myeloma, n=1 with MGUS) Diffuse infiltration: n=14 patients; non-diffuse infiltration: n=18 patients	Dual-layer detector spectral CT (IQon Spectral CT, Philips Healthcare) Scan area: from the crown of the head to the knees 120 kV; 93 mAs;	Calcium subtraction via manufacturer software without further details on decomposition (IntelliSpace Portal Version 11, Philips) Calcium suppression indices: 25–95 in increments of 10	1.5 T T1 TSE, T2 STIR, DWI with ADC maps to determine the infiltration patterns Full body scan Conventional CT
Liang et al., 2024	Evaluation of VNCa for detecting bone marrow infiltration across different skeletal sites	MM patients (n=72) based on the criteria of the 2020 Chinese Guidelines for the Diagnosis and Treatment of MM Control group: n = 10 healthy individuals	Dual-source CT (SOMATOM Force, Siemens Healthineers) Scan area: Entire spine including pelvis 100 kV/Sn150 kV, 190 mAs (tube A), 380 mAs (tube B), native	Three-material decomposition (bone mineral, yellow and red bone marrow) Post-processing using Syngo.via VB20 (Siemens Healthineers)	3 T T1 and T2 Spine and pelvis MRI used as a reference standard for assessing bone marrow infiltration
Werner et al., 2022	Evaluation of VNCa for assessing osteolytic lesion activity	n=32 patients with MM with active/inactive disease status according to IMWG criteria 37 examinations of the 32 patients with analysis of 103 focal osteolytic lesions	Dual-source CT (SOMATOM Definition Flash, Siemens Healthineers) Scan area: Base of skull to knee 100 kV and 140 kV, 160 mAs (tube A), 139 mAs (tube B), native	Three-material decomposition (fat, soft tissue, and calcium) Post-processing using Syngo.via VB30A (Siemens Healthineers)	1.5 T T1 and T2 spin echo, DWI Head to thigh Hematological markers: M-gradient in serum and urine, immunofixation tests to determine disease status Bone marrow biopsies from the iliac crest in 24 of 37 cases to determine plasma cell infiltration
Gu et al., 2022	Development of automated segmentation of the entire skeleton Evaluation of VNCa for detecting activity of bone lesions	n=21 patients with suspected or confirmed newly diagnosed or recurrent MM according to IMWG criteria; n=16 with osteolysis	Dual-source CT (SOMATOM Force, Siemens Healthineers) Scan area: Vertex to proximal tibia 90 kV and 150 kV, automatic dose modulation, native	Three-material decomposition (bone mineral, yellow and red bone marrow) Post-processing using Syngo.via VB30 (Siemens Healthineers)	Biopsy in 15 patients to determine bone marrow plasma cell infiltration as a reference standard Correlation with hemoglobin level and age
Hu et al., 2021	Application of spectral parameters to detect bone marrow infiltration and distinguish infiltration patterns	n = 35 patients with newly diagnosed MM (IMWG criteria); control group: n = 15 healthy individuals	Rapid kVp-switching CT (Discovery CT750 HD, GE Healthcare) Scan area: Head to thigh (for MM patients); T11 to S1 (for the control group) 80 kV/140 kV, 260 mA, native	Two-material decomposition for calcium and water or hydroxyapatite and fat Post-processing using manufacturer software (GSI Viewer, GE Healthcare)	3 T T1 Fast SE, T2 FS Fast SE, DWI Spine and pelvis Consensus of two independent readers
Fervers et al., 2021	Use of VNCa for monitoring during radiation therapy and early identification of treatment failure	n=33 patients diagnosed with MM according to IMWG criteria Status post Radiation of at least one osteolysis, no more than 5 irradiated osteolysis Classification into two groups (stable/partial remission vs. progression/relapse) Total 170 lesions	Dual-layer detector spectral CT (IQon, Philips Healthcare) Scan area: Whole body 120 kV, 70 mAs, native	Calcium subtraction (VNCa) with high suppression index (25%) Post-processing via manufacturer software, no further details on decomposition (IntelliSpace Portal, Spectral Diagnostics Suite, Philips Healthcare)	Conventional CT
Reinert et al., 2021	Identification of VNCa texture features for assessing disease activity and treatment response	n=110 patients with MM according to IMWG criteria Subgroup with histology: 56 patients	Dual-source CT (SOMATOM Definition Flash, Siemens Healthineers) Scan area: Whole body with arms elevated from elbow to knee 100 kV and Sn140 kV, 230 mAs (tube A), 178 mAs (tube B), native	Three-material decomposition (fat, soft tissue, calcium) Post-processing using Syngo.via VB30A (Siemens Healthineers)	Segmentation by three independent readers in consensus Histology to assess bone marrow infiltration (iliac crest biopsy) Serological markers: Serum-free light chains, free light chain ratio, kappa/lambda ratio Radiomics via manufacturer software (Pyradiomics); 92 features
Reinart et al., 2020	Application of VNCa texture features to predict treatment response	n = 44 patients with MM (diagnosis confirmed by laboratory parameters) and CT before and after therapy Treatment breakdown: n = 29: preparation for stem cell transplantation with bortezomib, lenalidomide, and dexamethasone n = 5: consolidation therapy with cyclophosphamide, adriamycin, and dexamethasone n = 5: maintenance therapy with lenalidomide and dexamethasone n = 5: active surveillance following prior therapy	Dual-source CT (SOMATOM Definition Flash, Siemens Healthineers) Scan area: Whole body with arms elevated from elbow to knee 100 kV and Sn140 kV, 230 mAs (tube A), 178 mAs (tube B), native	Three-material decomposition (fat, soft tissue, calcium) Post-processing using Syngo.via VB30A (Siemens Healthineers)	Vertebral body segmentation by three independent readers in consensus Laboratory values: M-protein in serum and urine Classification of treatment response per IMWG criteria: complete/partial remission, stable disease, progression Radiomics via manufacturer software (Pyradiomics); 41 features
Kosmala et al., 2018 (Radiology)	Evaluation of VNCa for detecting bone marrow infiltration	n=34 patients, including: n=29 with known MM, n=5 with monoclonal gammopathy of unclear significance	Dual-source CT (SOMATOM Force, Siemens Healthineers) Scan area: Crown of head to proximal tibia < 90 kg: 90 kV and Sn150 kV with 220 mAs (tube A) and 138 mAs (tube B), > 90 kg: 100 kV and Sn150 kV with 276 mAs (tube A) and 138 mAs (tube B), native	Three-material decomposition (bone mineral, yellow and red bone marrow) Post-processing using Syngo.via version VA30A (Siemens Healthineers)	3 T T1 TSE, T2 TIRM Entire spine and pelvis Bone involvement assessed on MRI by one reader CT images also evaluated for bone involvement by two additional readers
Kosmala et al., 2018 (Eur Radiol)	Assessment of VNCa for differentiating infiltration patterns	n=53 patients, including n=34 from the study by Kosmala et al. 2018 in Radiology; including: n=45 with MM, n=8 with monoclonal gammopathy of undetermined significance according to IMWG criteria Control group: n=21 healthy individuals	Dual-source CT (SOMATOM Force, Siemens Healthineers) Scan area: Crown of head to proximal tibia < 90 kg: 90 kV and Sn150 kV with 220 mAs (tube A) and 138 mAs (tube B), > 90 kg: 100 kV and Sn150 kV with 276 mAs (tube A) and 138 mAs (tube B), native	Three-material decomposition (bone mineral, yellow and red bone marrow) Post-processing using Syngo.via version VA30A (Siemens Healthineers)	3 T T1 TSE, T2 TIRM Entire spine and pelvis Infiltration pattern defined according to IMWG by two independent readers
Thomas et al., 2015	Suitability of VNCa for detecting bone marrow infiltration and differentiating infiltration patterns	n=32 patients, including n=22 with MM, n=10 with monoclonal gammopathy of unclear significance according to IMWG criteria	Dual-source CT (SOMATOM Definition Flash, Siemens Healthineers) Scan area: from elbow to knee 100 kV and 140 kV, 230 mAs (tube A), 178 mAs (tube B), native	Three-material decomposition (fat, soft tissue, calcium) Post-processing using a self-developed Matlab tool	1.5 T T1 TSE and T2* GRE of the spine Infiltration assessed on MRI by two expert readers by consensus Infiltration categories: High: Bone marrow iso-/hypointense to intervertebral discs Moderate: More hypointense than normal, but hyperintense vs. discs Pattern: diffuse, multifocal, salt-and-pepper (not observed in this cohort)

Table 2 Key findings of the studies included.

Author Year	Parameters	Statistics	Results	Conclusion
Abbreviations: 95% CI = 95% confidence interval, ADC = apparent diffusion coefficient, AUC = area under the curve, CVB = cervical vertebral bodies, DECT = dual-energy computed tomography, DWI = diffusion imaging, FS = fat saturation, GE = General Electric, GRE = gradient echo, GSI = Gemstone Spectral Imaging, HD = high definition, HU = Hounsfield unit, ICC = intraclass correlation coefficient, ID = initial diagnosis, IMWG = International Myeloma Working Group, LVB = lumbar vertebral bodies, MGUS = monoclonal gammaopathy of undetermined significance (precursor to multiple myeloma), MM = multiple myeloma, RBM = red bone marrow, ROI = region of interest, ROC = receiver operating characteristic, SE = spin echo, STIR = short tau inversion recovery sequence, SVB = sacral vertebral bodies, TIRM = turbo inversion recovery magnitude, TSE = turbo spin echo, TVB = thoracic vertebral bodies, VB10, VB20, VB30 = Syngo.via software versions, VNCa = virtual calcium subtraction
Chen et al., 2024	ROIs in TVB 11, TVB 12 and LVB 1 Material density values (mg/cm³) on maps after subtraction of calcium or hydroxyapatite HU values on 70 keV images	t-test/Mann-Whitney U test/Chi-square test/Fisher's exact test ROC analysis for MM diagnosis Positive predictive value, negative predictive value Youden Index to define optimal cut-off Delong test to compare AUCs	Before calcium or hydroxyapatite subtraction, there were no differences in material density or HU values among patients with diffuse or focal MM, and healthy controls. After subtraction of calcium or hydroxyapatite, significant differences were observed in material density and HU values among patients with diffuse or focal MM, and healthy controls. Material density maps following hydroxyapatite subtraction were superior to material density maps after calcium subtraction: AUC MM vs. healthy hydroxyapatite: (95%-CI), 0.874 (0.800–0.949) vs. calcium: 0,737(0.630, 0.844); p=0.02; AUC diffuse infiltration vs. focal infiltration hydroxyapatite 0.809 (0.654, 0.964) vs. calcium 0.736 (0.566, 0.907); p=0.049. Diagnostic value remained stable regardless of fat, muscle, or water as reference material.	Material density values and HU values after hydroxyapatite or calcium subtraction are useful for detecting bone marrow infiltration and distinguishing infiltration patterns.
Xiong et al., 2024	ROIs in LVB 1 to LVB 5 HU values on calcium subtraction maps	Intraclass correlation coefficient (ICC) between two reviewers for VNCa HU and ADC values Pearson correlation between VNCa HU and ADC values ROC analysis for predicting serum free light chain ratio and high-risk cytogenetic status	ICC for VNCa HU values: 0.824−0.970 HU values on VNCa images correlate with ADC values for calcium subtraction from 75 to 95% (Pearson’s r 0.342−0.612, p < 0.05). There was no correlation with greater calcium subtraction (calcium suppression index 25%-45%). For a calcium subtraction index of 85, there was a correlation with plasma cell infiltration (r = 0.835, p < 0.001). AUC for predicting the light chain ratio for calcium suppression index of 85: 0.876 (0.736−0.958) AUC for predicting high-risk cytogenetic status for a calcium suppression index of 85: 0.760 (0.603−0.878)	HU values on VNCa images are diagnostically and prognostically informative.
Wang et al., 2024	ROIs in TVB 10 to SVB 1 (plus ilium if visible) HU values on VNCa maps in patients with yellow bone marrow (control), red bone marrow (RBM), and diffuse bone marrow infiltration (MM)	Intraclass correlation coefficient between two reviewers for VNCa HU values One-way ANOVA for group comparisons ROC analysis for diagnostic thresholds and sensitivity/specificity Spearman correlation between MRI grades and VNCa HU values	ICC for VNCa HU values in MM: 0.908; in controls: 0.948 HU values increased from yellow to red marrow to diffuse infiltration (p < 0.001) Correlation with MRI grading (Baur/Stäbler): r = 0.897 (95% CI: 0.822–0.942; p < 0.001) ROC cut-off: −25.85 HU to distinguish yellow marrow from MM (AUC = 0.997; sensitivity 99.5%; specificity 96.3%) HU threshold 7.15 HU to differentiate RBM from MM (AUC = 0.723) Diagnostic performance of 70 keV HU values alone was low (AUC = 0.427; p < 0.001) Fat density values showed better diagnostic accuracy, especially at L2–5 (AUC = 0.837; sensitivity 80%; specificity 82.4%) Bone marrow aspiration confirmed CT-based diagnosis	VNCa HU values are sensitive and specific for detecting diffuse bone marrow infiltration. Moderate bone marrow infiltration can be differentiated from red bone marrow based on HU values.
Jiang et al., 2024	HU values on 70 keV maps and fat density values (mg/cm³) on hydroxyapatite subtraction maps Mean values from ROIs in T1–4, T5–10, T11-L1, and L2–5 Combination of CT-based fat density (DFat_HAP) and HU values for diagnosing multiple myeloma with negative bone density (MNBD; i.e., MM without visible osteolytic lesions on conventional CT)	Intraclass correlation coefficient between two reviewers ROC analysis to assess diagnostic accuracy of fat density (segment-based sensitivity/specificity) Definition of optimal cut-off values per spinal segment (Youden Index)	Lower HU values and higher fat density values in MM patients compared to healthy controls for all spinal segments (p < 0.001). However, HU values in 70 keV maps are not diagnostically suitable for detecting bone marrow infiltration (AUC = 0.427; p < 0.001; sensitivity 60.3%; specificity 27.5%; cut-off −0.121 HU). Based on fat density maps, there is a moderate diagnostic accuracy (AUC = 0.733; p < 0.001; sensitivity 58.8%; specificity 77.8%, cut-off 958 (mg/cm3). Diagnostic accuracy was highest for fat density values at the level of L2–5 (AUC = 0.837; sensitivity 80%; specificity 82.4%).	Fat density maps are superior to 70 keV HU values for detecting bone marrow infiltration.
Brandelik et al., 2021	Use of the VNCa technique with calcium suppression indices ranging from 25% to 95% for bone marrow infiltration analysis ROIs in the vertebral bodies C7, T12, L1 to L5 and the ilium	ROC analysis for diagnostic performance of VNCa (e.g., CaSupp Index 65) Pearson correlation between VNCa HU values and ADC values	The highest correlation between VNCa HU values and ADC values was found for a CaSupp Index 65 (r = 0.68, p < 0.001). The ROC analysis showed an AUC of 0.819 for LVB1–5 to distinguish diffuse vs. non-diffuse infiltration (cut-off: –1.6 HU; sensitivity = 78.6%, specificity = 75.0%). Conventional CT showed lower diagnostic performance (AUC = 0.641). Interrater agreement for quantitative VNCa measurements was excellent (ICC = 0.98).	VNCa HU values with moderate calcium suppression enable detection and classification of bone marrow infiltration.
Liang et al., 2024	ROIs in the spine (cervical, thoracic, lumbar) and pelvis HU values from regular and VNCa maps	Fleiss’ kappa for agreement between VNCa and MRI findings Bland-Altman analysis for interobserver consistency ROC analysis with AUC, sensitivity, specificity, and optimal cut-off values Tamhane's T2 test for group comparisons	HU values were higher in MM with bone marrow infiltration (169.7 HU; 95% CI 164.6–174.7) than in MM without bone marrow infiltration (163.6 HU; 95% CI, 158.1–169.0) or in the control group (139.6 HU; 95% CI, 131.4–147.8; p < 0.001). HU values on VNCa maps were higher in MM with bone marrow infiltration (−28.3 HU; 95% CI, −32.1 to −24.6) than in MM without bone marrow infiltration (−97.5 HU; 95% CI, −104.7 to −90.3) or in the control group (−89.1 HU; 95% CI, −95.1 to −83.1; p < 0.001), although the difference in VNCa HU between MM without bone marrow infiltration and the control group was not significant (p= 0.240). The diagnostic performance of VNCa HU at a cut-off of –42.2 HU (AUC = 0.839, sensitivity 70.6%, specificity 84.4%) was higher than that of HU values before calcium subtraction (AUC = 0.532; sensitivity 80.6%; specificity 29%). VNCa HU values were superior to HU values before calcium subtraction in all anatomical regions (cervical spine, thoracic spine, lumbar spine, pelvis). The optimal VNCa HU cut-off values varied by region (cervical spine –21.9 HU; thoracic spine –42.8 HU; lumbar spine –56.9 HU; pelvis –66.3 HU)	Compared to regular HU values, VNCa HU values are more effective in detecting bone marrow infiltration. Body-region specific cut-off values are required for accurate diagnosis.
Werner et al., 2022	VNCa HU values of focal osteolytic lesions in the spine and pelvis Comparison with MRI (T1w and ADC values) for validation	ROC analysis to evaluate diagnostic accuracy Spearman correlation between VNCa HU values, ADC values, and T1w signal intensity T-tests and Mann-Whitney U-tests to compare groups Intraclass correlation coefficient to assess measurement reproducibility.	Mean VNCa HU in active disease (higher plasma cell infiltration in the biopsy) 12.4 HU (SD, 24.6), in inactive disease –25.3 HU (SD, 32.0) (p < 0.001) AUC value of the ROC analysis: 0.823 (95% CI 0.739–0.907; p<0.001) for differentiating between active and inactive disease. The ideal VNCa cut-off for detecting active cases was –21.4 HU (sensitivity 92%, specificity 58%). In comparison, for a visual assessment of disease activity, the sensitivity was only 57% and the specificity 70%. There was a negative correlation of VNCa values with T1w signal intensity (r = –0.617) and a positive correlation with ADC values (r = 0.521), each p < 0.001.	VNCa HU values allow accurate differentiation between active and inactive lesions.
Gu et al., 2022	Segmentation of the entire skeletal system VNCa HU values of the whole skeleton ROIs in L3, the iliac crest, and up to five focal osteolytic lesions (if present)	Spearman correlation between HU values and plasma cell infiltration Wilcoxon test for differences between ROIs and the entire skeletal system	Median VNCa HU of the total skeleton was –59.9 HU (–66.3 to –51.8 HU); median of focal ROIs in LVB3 and iliac crest was –51.3 HU, median within osteolyses was –11.8 HU (each p < 0.001) Significant differences between total skeleton VNCa HU values and focal ROI HU values (e.g. LVB3 vs. iliac crest, p < 0.001). Positive correlation between VNCa HU values of the whole skeleton and plasma cell infiltration in the bone marrow (r = 0.79; p < 0.001) and negative correlation with Hb value (r = –0.66; p = 0.001)	VNCa HU correlate with lesion activity and vary by skeletal region
Hu et al., 2021	ROIs in the lumbar spine: For focal patterns: five of the largest lesions (≥5 mm diameter) For all other infiltration patterns and healthy subjects: one 10 mm² ROI in the vertebral body HU values on 70 keV maps Effective atomic number, material density of water, calcium, hydroxyapatite, and fat	Spearman correlations Bonferroni-adjusted ANOVA for group comparisons ROC analyses for bone marrow infiltration prediction Logistic regression with material density pairs as predictors	Strong positive correlation between VNCa HU values and plasma cell infiltration (r = 0.79; p < 0.001), and negative correlation with Hb levels (r = –0.66; p = 0.001) Significant differences between infiltrated and healthy bone marrow for effective atomic number, calcium density, hydroxyapatite density, fat density, and water content (p < 0.001); not significant for 70 keV HU (p = 0.427) Best ROC performance when using fat density alone (AUC = 0.846; 95% CI: 0.84–0.88; sensitivity 62%; specificity 93%) Two-parameter ROC: Calcium + water: AUC = 0.856 (95% CI: 0.81–0.89); sensitivity 84%; specificity 77% Hydroxyapatite + fat: AUC = 0.850 (95% CI: 0.80–0.88); sensitivity 79%; specificity 81% Infiltration pattern differentiation (focal, diffuse, salt-and-pepper) successful with calcium, hydroxyapatite, fat, and atomic number (p < 0.01) Regression using calcium + water: AUC = 0.951 (95% CI: 0.91–0.95); sensitivity 90.5%; specificity 93.3% ADC-based differentiation (reference standard): AUC = 0.954 (95% CI: 0.94–0.96); sensitivity 90.5%; specificity 95.2%	Combined material density values are diagnostically superior to individual spectral parameters for detecting bone marrow infiltration and distinguishing infiltration patterns. Diagnostic accuracy is comparable to MRI ADC maps.
Fervers et al., 2021	ROIs in all irradiated and non-irradiated osteolytic lesions of the spine and pelvis (copied from conventional CT onto the VNCa map) HU values on the conventional images and VNCa maps Tumor response assessment based on MD Anderson criteria Percent change in HU values pre- vs. post-radiation Calculation of bone mineral density by subtracting VNCa HU values from conventional HU values	Subset of 30% of cases used for interrater reliability ROC analysis comparing irradiated vs. non-irradiated lesions using percentage change in VNCa and regular HU values. HU comparison from pre-irradiation to post-irradiation CT; if unavailable, post-irradiation HU only Pearson correlation between radiotherapy dose and density values Intraclass correlation for measurement reproducibility	Significant drop in regular HU values after radiation therapy: from 22.0 HU (30.5–47.0) to 8.5 HU (–3.1 to 33.0) Significant drop in VNCa HU values: from 4.5 HU (–3.8 to 7.0) to –53.5 HU (–94.8 to –20.5) In percentage terms, the change in values from before to after radiation was –48% (–7 to –92) of the regular HU values and –228% (–23 to –583) of the VNCa HU values. Slightly better performance of the VNCa HU values regarding the identification of a pre-irradiated lesion (AUC 0.57–0.85 depending on the time interval from irradiation) than the regular HU values (AUC = 0.56–0.75 depending on the time interval from irradiation). The diagnostic suitability of the VNCa HU values for the detection of irradiated lesions increased significantly for bone lesions with high calcium content (for these, for example, with >90% calcium subtraction, AUC 0.96 (0.91–1.00)). In individual cases with progression of the lesions despite radiation (3 of 75 lesions), sometimes only after 1800 days after irradiation, VNCa HU values were consistently higher in all follow-up examinations than in stable or regressed lesions. The difference was more pronounced in VNCa HU values (–3.0 HU (IQR –21.8–8.3) vs. –62.5 HU (IQR –99.8 to –30.3)) than in regular HU values (32.5 HU (IQR 29.0–39.8) vs. 7.0 HU (IQR –35.0–29.3). Negative correlation between radiation dose and density values (regular HU r=–0.40 (95 CI –0.55 to –0.23), p<0.001; VNCa HU r=–0.21 (95 CI –0.38 to –0.02), p=0.03). ICC between ROI measurements and clinical tumor response according to MD Anderson criteria 0.85 each	VNCa HU values are superior to regular HU values in identifying irradiated lesions and predicting treatment response.
Reinert et al., 2021	Volumetric ROIs from TVB 10 to LVB 5 Texture analyses of VNCa maps	ICC and Bland-Altman analyses for interrater reliability Mann-Whitney U test for differences in texture features by myeloma stage, serum markers, and osteolysis presence Pearson correlation for the relationship between texture metrics and bone marrow infiltration Multivariable linear and logistic regression to predict bone marrow infiltration and serum abnormalities based on texture features Z-transformation of texture parameters	Significant positive correlation between bone marrow infiltration (%) and 1st order feature “10th percentile” and “uniformity” (each p < 0.001) as well as negative correlation with the 1st order feature “entropy” (p < 0.001) and 2nd order feature “grayscale coincidence matrix contrast” and “grayscale coincidence matrix average difference” (each p < 0.0001). In the regression models, these same features as well as “grayscale coincidence matrix difference in entropy” showed a significant influence (r from –0.52; –0.23 to 0.49). Osteolysis was associated with the 1st order characteristics “mean” (p < 0.004), “minimum” (p < 0.004), “10th percentile” (p < 0.003) and “grayscale coincidence matrix run variance” (p < 0.007), resulting in a regression model with an r2 = 0.33. Myeloma stages were associated with the 1st order feature “10th percentile” (p<0.01), “90th percentile (p < 0.01), ”mean” (p < 0.02), and the 2nd order feature “grayscale coincidence matrix cluster prominence” (p < 0.004) Furthermore, there were smaller associations with the kappa/lambda ratio (r2 = 0.18) and elevated serum light chains (r2 = 0.18) ICC 0.85 (0.72–0.95) to 0.87 (0.65–0.94)	VNCa radiomic features may serve as non-invasive biomarkers to predict disease activity and treatment response.
Reinart et al., 2020	Volumetric ROIs from TVB 10 to LVB 5 Texture analyses of the VNCa maps	Wilcoxon test for differences in texture characteristics before and after therapy ROC analyses for prediction based on texture features Z-transformation of texture parameters	n = 29 complete remission, n = 10 progression, n = 5 stable findings In case of progression, decrease in 1st order features “10th percentile” (0.8±1.0 vs. 0.1±1.2, p>0.05), “median” (0.7±1.0 vs. 0.1±1.2, p>0.05), and “minimum” (0.5±0.1 vs.–0.5±1.5, p>0.05) as well as increase in the 1st order features “range” (–0.6±0.1 vs. 0.5±1.5, p>0.05) and 2nd order “grayscale coincidence matrix difference of variance” (–0.3±0.1 vs. 0.1±0.6, p>0.05). 2nd order feature “grayscale coincidence matrix difference of variance” differentiated with a cut-off of –0.28 between complete remission and progression (AUC = 0.76; sensitivity 93%; specificity 70%). In stable disease, texture characteristics were also unchanged.	VNCa adiomics features are associated with disease progression.
Kosmala et al., 2018 (Radiology)	Up to five ROIs per patient in bone lesions from MRI and additional five ROIs of 100 mm2 in unaffected bone marrow regions in vertebral bodies and pelvis Evaluation of bone marrow involvement in VNCa maps by two independent readers	ROC analysis for diagnostic performance of regular vs. VNCa HU values Interobserver agreement (kappa) Sensitivity, specificity, positive and negative predictive values	Interobserver reliability of the visual analysis of bone marrow involvement on regular CT data k 1.00; for VNCa maps k 0.93 Visual analyses: regular CT: Sensitivity 69.6%; specificity 90.9%; VNCa maps: sensitivity 91.3%, specificity 90.9%, thus a particularly higher negative predictive value of VNCa of 83.3% vs. 58.8% in regular CT. Quantitative analyses: regular CT (AUC = 0.734; SD 0.035; at a cut-off of 78.5 HU; sensitivity 52.0%; specificity 84.7%), VNCa (AUC 0.978; SD 0.007; at a cut-off of 244.9 HU; sensitivity 93.3%; specificity 92.4%)	Both qualitatively and quantitatively, VNCa outperforms conventional CT in the detection of bone marrow infiltration.
Kosmala et al., 2018 (Eur Radiol)	For focal patterns, up to five ROIs per patient in bone lesions identified on MRI. For normal and diffuse infiltration patterns: five ROIs per patient (100 mm²) in TVB12, LVB4–SVB1, or the ilium. VNCa HU values	ROC analysis of diagnostic performance of VNCa HU values in distinguishing between infiltration patterns.	VNCa HU values were significantly higher in patients with diffuse bone marrow infiltration than in those with normal bone marrow or focal pattern or in healthy controls (p < 0.001) or in focal lesions (p = 0.002). VNCa HU values in focal lesions were different from healthy controls (p < 0.001). There were no differences between controls and patients with normal pattern (p = 0.672). VNCa-based differentiation of diffuse vs. normal infiltration pattern (AUC = 0.997; cut-off value –35.7 HU; sensitivity 100% (95% CI 100–100%); specificity 97% (95% CI 94–100%). VNCa-based differentiation of diffuse vs. focal pattern (AUC = 0.726; cut-off value –14.4 HU; sensitivity 33% (95% CI 15–52%); specificity 25% (95% CI 16–33%). VNCa-based differentiation of focal vs. normal infiltration pattern (AUC = 0.996; cut-off value –31.9 HU; sensitivity 97% (95% CI 94–100%); specificity 99% (95% CI 98–100%)	VNCa HU are well suited for distinguishing infiltration patterns.
Thomas et al., 2015	10–20 ROIs in the spine (affected and unaffected regions), defined by MRI Comparison between regular HU and VNCa HU values	ROC analyses for predicting bone marrow infiltration in osteolytic and non-osteolytic lesions Matthew correlation coefficient for evaluating prediction quality	In osteolytic lesions, regular HU values (AUC = 0.877) and VNCa HU values (AUC = 0.916) were suitable to detect bone marrow infiltration. In non-osteolytic lesions, only VNCa HU values were suitable to detect bone marrow infiltration. (AUC = 0.932), but not regular HU values (AUC = 0.577). VNCa HU cut-off values for detection of bone marrow infiltration: 4 HU in osteolytic lesions (89% sensitivity, 85% specificity, Matthew r = 0.7288) and –3 HU in non-osteolytic lesions. VNCa HU values showed a diagnostic performance for differentiating the infiltration patterns than regular HU values (diffuse infiltration: VNCa HU sensitivity 40%, specificity 85.7% vs. regular HU 0% and 100%; multifocal infiltration: VNCa HU sensitivity 87.5%, specificity 78.3% vs. regular HU 87.5% and 73.9%).	VNCa HU values outperform conventional HU values in detecting bone marrow infiltration, particularly in non-osteolytic lesions. They also allow accurate differentiation between focal, diffuse, and normal infiltration patterns.

Results

Study overview

A total of n = 15 studies published between 2015 and 2025 were included in the final analysis ([Fig. 1]). The majority of studies used dual-source spectral CT (10/15), followed by dual-layer detector CT (3/15) and fast kVp-switching CT (2/15). Although the examined anatomical regions varied, all studies included measurements within the spine. The most frequently evaluated method was the use of attenuation values in Hounsfield units (HU) based on VNCa images (10/15). Less frequently used were material density maps for calcium, hydroxyapatite, fat, or water (4/15 in total) and HU values on monoenergetic maps (70 keV each, 3/15). One study also included atomic number (Z-effective) maps. Two studies focused on radiomic analysis based on VNCa data. A summary of the study characteristics and main findings is shown in [Fig. 2]. Additional details regarding study objectives, methodology, and results are presented in [Table 1] and [Table 2].

Fig. 2 Overview of study characteristics and findings. Created in BioRender. Molwitz, I. (2025) https://BioRender.com/x71k794. [rerif].

Assessment of study quality and risk of bias

Among the 10 studies evaluating VNCa, the risk of bias was predominantly low (28/40 domains rated as “low risk”, nine “high risk”, and three “unclear”). Applicability to the review question was consistently rated as “low concern” across all domains (30/30). In the four studies using material density maps, 10 of 16 domains were rated as “low risk”, five as “high risk”, and one as “unclear”. Concerns regarding applicability to the review question were low in all cases. The three studies evaluating monoenergetic HU values showed no domains rated as “low risk”, three as “high risk”, but all had low applicability concerns. Full details are provided in Appendix, Table 2 and illustrated in Appendix, Figure 1.

Detection of bone marrow infiltration using VNCa

VNCa-based HU values were shown to effectively detect bone marrow infiltration in multiple myeloma [13] [14] [15], significantly outperforming conventional CT HU values [14] [16] [17]. For example, Kosmala et al. reported a sensitivity of 52.0% and specificity of 84.7% for conventional CT, compared to 93.3% and 92.4%, respectively, for VNCa-based HU values [17].

The advantage of VNCa was especially pronounced in non-osteolytic lesions (AUC for VNCa: 0.932; vs. regular CT: 0.577) [18]. Even in osteolytic lesions, VNCa improved detection of disease activity compared to conventional CT (active vs. inactive lesions: AUC 0.823 (95% CI: 0.739–0.907), p<0.001) [14]. Cut-off values for HU in VNCa images varied depending on the anatomical location (e.g., cervical vs. thoracic vs. lumbar spine) [16], but quantitative assessments were generally superior to visual interpretations [14] [17]. In contrast, results using monoenergetic HU values at 70 keV were poor (AUC = 0.427; sensitivity 60.3%; specificity 27.5%) [19]. VNCa HU values were particularly effective in detecting fatty marrow infiltration (cut-off: –25.85 HU: AUC = 0.997; sensitivity 99.5%; specificity 96.3%) [20]. However, in cases of moderate red marrow infiltration, sensitivity dropped from 70.9% (specificity 78.9%) to 31% (specificity 93.9%) [20].

A moderate correlation between VNCa HU values and MRI-derived apparent diffusion coefficient (ADC) values was observed in regions with high calcium subtraction (75–95% subtraction; r = 0.342–0.612; p < 0.05) [13].

Detection of bone marrow infiltration using material density maps

Material density maps were analyzed for calcium, hydroxyapatite, fat, and water, depending on the base materials used for decomposition. Material density maps appear to be well suited, in principle, to detect bone marrow infiltration. For example, Hu et al., who examined regular HU values, VNCa HU values, 70 keV HU values, and atomic number maps (Z-effective), showed a strong diagnostic performance, particularly for fat material density maps (AUC = 0.846 (95% CI 80.4–88.3%); sensitivity 62%; specificity 93%) [21]. Chen et al. demonstrated that diagnostic performance was influenced by the type of dense material in the decomposition model (hydroxyapatite AUC: 0.874 (95% CI: 0.800–0.949) vs. calcium AUC: 0.737 (95% CI: 0.630–0.844)) [15]. Regression models combining multiple material maps yielded good diagnostic accuracy – for instance, combining calcium and water achieved an AUC of 0.856 (95% CI: 0.814–0.891); sensitivity 84%; specificity 77%. Similarly, combining hydroxyapatite and fat resulted in an AUC of 0.850 (95% CI: 0.807–0.886); sensitivity: 79%; specificity: 81% [21]. Jiang et al. reported that fat material density maps enabled detection of bone marrow infiltration even in the absence of osteolysis on conventional CT (lumbar vertebrae L2–L5: AUC = 0.837; sensitivity 80%; specificity 82.4%) [19].

Value of spectral CT techniques for differentiating bone marrow infiltration patterns

Five different bone marrow infiltration patterns were identified: Normal marrow appearance, homogeneous diffuse infiltration, focal infiltration, mixed diffuse and focal infiltration, and “salt and pepper” pattern [22]. Several studies confirmed the utility of VNCa-based HU values for differentiating between these patterns [18] [23] [24]. Chen et al. found that hydroxyapatite maps outperformed calcium maps in distinguishing diffuse from focal infiltration (AUC diffuse infiltration vs. focal infiltration: hydroxyapatite 0.809 (95% CI 0.654–0.964) vs. calcium 0.736 (95% CI 0.566–0.907)) [15]. Hu et al. also demonstrated significant differences across all infiltration patterns using fat maps and Z-effective maps. Notably, even marrow infiltrations with normal appearance could be identified this way [21]. A regression model combining calcium and water densities achieved diagnostic performance comparable to MRI ADC values according to Hu et al. (CT: AUC = 0.951 (95% CI 0.905–0.932); sensitivity 90.5%; specificity 93.3%); MRI ADC: AUC = 0.954 (95% CI 0.842–0.884); sensitivity 90.5%; specificity 95.2%). However, as the given 95% CI does not include the CT AUC value, these results should be critically reflected.

Spectral CT parameters as imaging biomarkers in multiple myeloma

VNCa HU values were also linked to tumor risk profiles as imaging biomarkers [13] and showed strong correlation with serological markers such as plasma cell infiltration (r = 0.79; p < 0.001) [25] and light chain ratio as an indicator of disease activity (prediction AUC: 0.876 (95% CI: 0.736–0.958)) [13]. Higher VNCa HU values in osseous lesions predicted lack of treatment response after radiation therapy, with a greater distinction than seen with regular HU values (mean HU difference between responding and non-responding lesions in VNCa: 59.5 HU vs. 25.5 HU with conventional CT) [26]. This was particularly evident in lesions with high calcium content (AUC 0.96 (95% CI: 0.91–1.00)) [26]. Radiomics analyses of VNCa maps identified texture features predictive of disease progression and treatment response [10] [27].

Discussion

This systematic review investigated how spectral CT techniques are used to detect and characterize bone marrow infiltration in patients with multiple myeloma.

Fifteen studies met the inclusion criteria. Their findings suggest that:

• Spectral CT techniques significantly improve the differentiation between healthy and infiltrated bone marrow compared to conventional CT. Diagnostic value was demonstrated for virtual calcium suppression (VNCa) and material density maps – both for dense materials (e.g., calcium, hydroxyapatite) and less dense materials (e.g., fat).

• Spectral CT techniques allow for the differentiation of the various infiltration patterns of multiple myeloma. Both VNCa and material density maps are helpful for differentiating the infiltration patterns of multiple myeloma. The only study that also examined atomic number maps also revealed meaningful differences between infiltration patterns.

• Spectral CT parameters may serve as imaging biomarkers for tumor activity, risk stratification, and therapy monitoring. VNCa values correlated strongly with clinical parameters such as plasma cell infiltration and enabled the differentiation of high-risk patients.

Predictive relevance was also demonstrated for texture parameters (radiomics).

Classification of the results

It is important to note that some studies reported even weak correlations as positive findings. In addition, the number of studies per parameter examined was very heterogeneous. Given that two-thirds of the studies analyzed VNCa data (10/15), these results are considered the most robust. Calcium and hydroxyapatite maps were also used several times (4/15). In contrast, atomic number maps (1/15) and radiomic features (2/15) were investigated only in a small number of studies. Their positive assessment for detection of bone marrow infiltration, differentiation of infiltration patterns, and predictive relevance should therefore be considered initially with reservations.

With regard to clinical applications of the study results, it is important to keep in mind that despite the first included study being published in 2015, major barriers remain for integrating these findings into routine clinical practice. Spectral bone marrow analysis requires commercially available software solutions integrated with the PACS system (e.g. Spectral Magic Glass, Philips) or a separate login to manufacturer-specific software for spectral analysis (e.g. Syngo.via, Siemens Healthineers). These systems have to allow users to retrieve the examination, actively upload it, if necessary, apply the proper tool, and send the results to the PACS system for documentation purposes. Moreover, many studies still depend on labor-intensive approaches, such as manual measurement of HU or density values in individual lesions. The time needed for these methods is generally not available in clinical routines. Furthermore, several studies relied on material density maps that are not offered as standard by manufacturers. Additionally, interpreting spectral maps requires substantial experience. Spectral CT also has limited sensitivity in cases of mild bone marrow infiltration, particularly when predominantly red bone marrow is involved [13]. Thus, despite the sufficient sensitivity and specificity demonstrated in the included studies, spectral CT is unlikely to replace MRI.

Nevertheless, spectral CT can provide valuable additional information in clinically indicated CT examinations. The combination of high-quality imaging and quantitative parameters makes spectral CT a promising complement to existing methods. Particularly when MRI is contraindicated or unavailable, spectral CT offers a time-efficient alternative.

Limitations of the studies included

To further classify the study results, limitations of the studies included should be considered. For example, in studies that used histological analyses as a reference for plasma cell infiltration, histology was only available for a subgroup of the cohort [10] [25]. In general, the cohort size of the studies was small, with a maximum of 110 patients and a median of 35. Larger studies are needed to validate these results, particularly regarding diagnostic accuracy compared to MRI. Future analyses should also incorporate subgroups corresponding to different spectral CT scanner types (dual-source, dual-layer detectors, etc.) to assess the robustness of findings. The lack of standardized material decomposition methods further complicates study comparisons (e.g., two- vs. three-material decomposition; calcium vs. hydroxyapatite as dense materials). International consensus recommendations to harmonize these methodologies could significantly enhance their clinical utility.

Longitudinal analyses are also necessary, particularly with regard to the predictive value of spectral CT parameters demonstrated in individual cases [10] [27].

Regarding the studies by Kosmala et al., it should be noted that they involve overlapping patient cohorts. For instance, the authors initially published data on 34 patients regarding the suitability of VNCa for detecting bone marrow infiltration in Radiology in 2018 [17]. Subsequently, they published on VNCa’s ability to differentiate infiltration patterns in European Radiology, adding an expanded cohort of 19 additional patients [23].

Limitations of the literature review

This review did not examine the potential benefit of spectral CT techniques for assessing extramedullary myeloma manifestations. For example, treatment responses could theoretically be assessed through iodine quantification in extramedullary lesions. The literature search was conducted exclusively via PubMed, albeit a leading biomedical database. Consequently, this review does not include a quantitative meta-analysis of study results.

Outlook on new techniques and other diseases

Photon-counting CT is another emerging technology that inherently provides spectral data, differentiating high- and low-energy photons using predefined signal thresholds [28]. Although several studies have applied photon-counting CT in multiple myeloma, these have predominantly focused on osteolytic lesion detection, image quality, or signal-to-noise ratio [29] [30]. To date, no studies have employed photon-counting CT specifically for assessing bone marrow infiltration. Furthermore, none of the studies reviewed used split-filter CT or sequential acquisitions at different tube voltages for spectral data generation. Due to the temporal delays in data acquisition and consequent motion artifacts, sequential acquisitions hold limited clinical relevance compared to other spectral CT techniques (e.g., rapid kV-switching, split-filter, dual-source, dual-layer detectors, photon-counting CT).

Besides exploring spectral CT as an alternative to MRI for bone marrow evaluation in multiple myeloma, research might also consider the inverse possibility – employing synthetic CT derived from MRI for evaluating mineralized bone [31] [32].

Future research should therefore adopt a comprehensive approach, evaluating not only the utility of individual imaging modalities but also their comparative strengths and weaknesses for diagnosing and managing multiple myeloma. While this review focused specifically on spectral CT for multiple myeloma, spectral CT techniques have also demonstrated utility in detecting other malignant bone marrow lesions, such as metastases [33], or distinguishing osteoblastic metastases from osteomas [34].

Conclusion

This systematic review demonstrates that spectral CT represents a promising imaging modality for multiple myeloma, providing detailed, quantitative information potentially beneficial for diagnosis and treatment monitoring. Future research should focus on larger multicenter studies involving various manufacturers to further validate spectral CT’s utility, particularly relative to MRI. Additionally, consensus on optimal spectral parameters and streamlined clinical workflows will be essential to successfully integrate these techniques into routine clinical practice.

Conflict of Interest

G.M. Campbell is employed as a Clinical Scientist at Philips GmbH Market DACH. The remaining authors declare that they have no conflicts of interest. The analysis of the literature findings was conducted by the independent authors.

• Literatur

• 1 Malard F, Neri P, Bahlis NJ. et al. Multiple myeloma. Nat Rev Dis Primers 2024; 10: 45
• 2 Rajkumar SV, Dimopoulos MA, Palumbo A. et al. International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. The Lancet Oncology 2014; 15: e538-e548
• 3 Weber M-A, Baur-Melnyk A. Radiologische Diagnostik des multiplen Myeloms: Rolle der Bildgebungsverfahren und aktuelle S3-Leitlinie. Radiologe 2022; 62: 35-43
• 4 Zamagni E, Tacchetti P, Cavo M. Imaging in multiple myeloma: How? When?. Blood 2019; 133: 644-651
• 5 Cheraya G, Sharma S, Chhabra A. Dual energy CT in musculoskeletal applications beyond crystal imaging: Bone marrow maps and metal artifact reduction. Skeletal Radiol 2022; 51: 1521-1534
• 6 Leitlinienprogramm Onkologie. S3-Leitlinie Diagnostik, Therapie und Nachsorge für Patienten mit monoklonaler Gammopathie unklarer Signifikanz (MGUS) oder Multiplem Myelom. 2022 Accessed April 14, 2025 at: https://www.leitlinienprogramm-onkologie.de/fileadmin/user_upload/Downloads/Leitlinien/Multiples_Myelom/LL_Multiples_Myelom_Langversion_1.0.pdf
• 7 Schierenbeck M, Grözinger M, Reichardt B. et al. Detecting Bone Marrow Edema of the Extremities on Spectral Computed Tomography Using a Three-Material Decomposition. Diagnostics 2023; 13: 2745
• 8 D’Angelo T, Albrecht MH, Caudo D. et al. Virtual non-calcium dual-energy CT: Clinical applications. Eur Radiol Exp 2021; 5: 38
• 9 McCollough CH, Leng S, Yu L. et al. Dual- and Multi-Energy CT: Principles, Technical Approaches, and Clinical Applications. Radiology 2015; 276: 637-653
• 10 Reinert CP, Krieg E, Esser M. et al. Role of computed tomography texture analysis using dual-energy-based bone marrow imaging for multiple myeloma characterization: Comparison with histology and established serologic parameters. Eur Radiol 2021; 31: 2357-2367
• 11 Page MJ, McKenzie JE, Bossuyt PM. et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021; n71
• 12 Whiting PF, Rutjes AWS, Westwood ME. et al. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Ann Intern Med 2011; 155: 529-536
• 13 Xiong X, Hong R, Fan X. et al. Quantitative assessment of bone marrow infiltration and characterization of tumor burden using dual-layer spectral CT in patients with multiple myeloma. Radiology and Oncology 2024; 58: 43-50
• 14 Werner S, Krauss B, Horger M. Dual-Energy CT-Based Bone Marrow Imaging in Multiple Myeloma: Assessment of Focal Lesions in Relation to Disease Status and MRI Findings. Academic Radiology 2022; 29: 245-254
• 15 Chen J, Qiu Z, Jiang N. et al. Detecting bone marrow infiltration in nonosteolytic multiple myeloma through separation of hydroxyapatite via the two-material decomposition technique in spectral computed tomography. Quant Imaging Med Surg 2024; 14: 2345-2356
• 16 Liang L, Xiao F, Liang L. et al. Visual assessment and quantitative analysis of dual-energy CT virtual non-calcium in imaging diagnosis of multiple myeloma. Skeletal Radiol 2025; 54: 1059-1070
• 17 Kosmala A, Weng AM, Heidemeier A. et al. Multiple Myeloma and Dual-Energy CT: Diagnostic Accuracy of Virtual Noncalcium Technique for Detection of Bone Marrow Infiltration of the Spine and Pelvis. Radiology 2018; 286: 205-213
• 18 Thomas C, Schabel C, Krauss B. et al. Dual-Energy CT: Virtual Calcium Subtraction for Assessment of Bone Marrow Involvement of the Spine in Multiple Myeloma. American Journal of Roentgenology 2015; 204: W324-W331
• 19 Jiang N, Xia Y, Luo M. et al. Diagnosis of newly developed multiple myeloma without bone disease detectable on conventional computed tomography (CT) scan by using dual-energy CT. Journal of Bone Oncology 2024; 48: 100636
• 20 Wang T, Zhou B, Zhang K. et al. Value of third-generation of VNCa dual-energy CT for differentiating diffuse marrow infiltration of multiple myeloma from red bone marrow. Medicine 2024; 103: e38003
• 21 Hu C, Zhang Y, Xiong X. et al. Quantitative evaluation of bone marrow infiltration using dual-energy spectral computed tomography in patients with multiple myeloma. Journal of X-Ray Science and Technology: Clinical Applications of Diagnosis and Therapeutics 2021; 29: 463-475
• 22 Kosmala A, Bley T, Petritsch B. Imaging of Multiple Myeloma. Fortschr Röntgenstr 2019; 191: 805-816
• 23 Kosmala A, Weng AM, Krauss B. et al. Dual-energy CT of the bone marrow in multiple myeloma: Diagnostic accuracy for quantitative differentiation of infiltration patterns. Eur Radiol 2018; 28: 5083-5090
• 24 Brandelik SC, Skornitzke S, Mokry T. et al. Quantitative and qualitative assessment of plasma cell dyscrasias in dual-layer spectral CT. Eur Radiol 2021; 31: 7664-7673
• 25 Gu R, Amlani A, Haberland U. et al. Correlation between whole skeleton dual energy CT calcium-subtracted attenuation and bone marrow infiltration in multiple myeloma. European Journal of Radiology 2022; 149: 110223
• 26 Fervers P, Celik E, Bratke G. et al. Radiotherapy Response Assessment of Multiple Myeloma: A Dual-Energy CT Approach With Virtual Non-Calcium Images. Front Oncol 2021; 11: 734819
• 27 Reinert CP, Krieg E-M, Bösmüller H. et al. Mid-term response assessment in multiple myeloma using a texture analysis approach on dual energy-CT-derived bone marrow images — A proof of principle study. European Journal of Radiology 2020; 131: 109214
• 28 Srinivas-Rao S, Cao J, Marin D. et al. Dual-Energy Computed Tomography to Photon Counting Computed Tomography: Emerging Technological Innovations. Radiologic Clinics of North America 2023; 61: 933-944
• 29 Rau A, Neubauer J, Taleb L. et al. Impact of Photon-Counting Detector Computed Tomography on Image Quality and Radiation Dose in Patients With Multiple Myeloma. Korean J Radiol 2023; 24: 1006
• 30 Schwartz FR, Vinson EN, Spritzer CE. et al. Prospective Multireader Evaluation of Photon-counting CT for Multiple Myeloma Screening. Radiology: Imaging Cancer 2022; 4: e220073
• 31 Chong LR, Lee K, Sim FY. 3D MRI with CT-like bone contrast – An overview of current approaches and practical clinical implementation. European Journal of Radiology 2021; 143: 109915
• 32 Lombardi AF, Ma Y-J, Jang H. et al. Synthetic CT in Musculoskeletal Disorders: A Systematic Review. Invest Radiol 2023; 58: 43-59
• 33 Liu M-C, Ho C-C, Lin Y-T. et al. Opportunistic screening with multiphase contrast-enhanced dual-layer spectral CT for osteoblastic lesions in prostate cancer compared with bone scintigraphy. Sci Rep 2024; 14: 5310
• 34 Luo H, Zou L, Yang Q. et al. Spectral CT assists differentiation of osteoblastic bone metastasis from bone island in newly diagnosed cancer patients. Eur Radiol 2023; 34: 60-68

Publication History

Received: 12 February 2025

Accepted after revision: 22 May 2025

Article published online:
23 July 2025