Abstract
Introduction
Peripheral stem-cell collection is an essential step/prerequisite for high-dose treatment of patients with multiple myeloma (MM). Radiomics provides numerous analytic parameters from imaging modalities and can characterize tissues in a quantitative manner. The present study used radiomics derived parameters based on computed tomography (CT) images to identify prognostic factors for stem-cell mobilization in patients with MM.
Methods
Between May 2020 and September 2022 all patients who had undergone quadruplet induction therapy, were scheduled for stem-cell mobilization in preparation for autologous stem cell transplantation, and had CT scans prior to chemomobilization were retrospectively analyzed. Total 34 patients (25 males [74%], median age 60 ± 8 years) were analyzed. Whole-body CT images routinely obtained before the start of mobilization were analyzed with texture analysis.
Results
Of the investigated CT radiomics features, three CT textures were associated with the concentration of CD34+ cell count in peripheral blood at begin of apheresis. The second-order texture feature “S(1,0)AngScMom” and the wavelet transform feature “WavEnHH_s-5” were positively correlated (r = 0.375, p = 0.031 and r = 0.432, p = 0.012, respectively), whereas the autoregressive feature “Teta1” was inversely correlated (r = −0.375, p = 0.031). The significant CT radiomics features were used to build a model with a good diagnostic accuracy with an area under the curve of 0.77 (95% CI: 0.59–0.96).
Conclusion
CT radiomics features can predict apheresis yield in patients undergoing hematopoietic cell mobilization in patients with MM. Further analyses are needed to validate the identified radiomics signature in clinical routine and to test the predictive abilities.
Keywords: Multiple myeloma, Computed tomography, Radiomics, Hematopoietic cell mobilization
Introduction
Quadruplet induction therapy followed by high-dose chemotherapy with autologous stem-cell transplantation (ASCT) is a cornerstone in the treatment younger patients with newly diagnosed multiple myeloma (MM) [1–5]. Unfortunately, conventional stem-cell mobilization with granulocyte colony stimulating factor with or without chemotherapy is insufficient in a proportion of patients [6]. For those “poor mobilizers,” the CXCR-4 antagonist plerixafor has proven to be highly effective as a rescue regime [7–10]. Various factors including age, sex, prior therapy, or remission status before mobilization have previously been described to have an influence on the mobilization capacity and stem cell yield in patients with MM [10–12]. As sufficient number of CD34+ cells in the graft is the most important prerequisite for ASCT, prior identification of poor mobilization patients could allow interventions such as pre-emptive plerixafor administration.
During the induction treatment and especially prior to ASCT, most patients with MM undergo computed tomography (CT) imaging, which is frequently used to exclude inflammation, pulmonal findings, and osteolysis/lytic bone lesions prior to ASCT. Radiomics analysis derived from CT imaging provides new quantitative imaging markers and can reflect the microstructure of the investigated tissues. The radiomics features consist of different groups including first-order (histogram) features and second order (texture) features. First-order features, including skewness, kurtosis, and entropy, are derived from pixel intensity distributions without considering spatial relationships [13, 14]. Second-order features analyze spatial relationships between voxels with similar gray levels, offering insights into intralesional heterogeneity. As a result, they may provide even greater diagnostic value than first-order features [15]. To summarize, these quantitative methods can provide new imaging biomarkers and aid to predict clinically relevant outcomes, which was demonstrated in different tumor entities [15].
In MM, radiomics has been used to improve characterization of osseous manifestations and was associated with the overall prognosis [16]. Moreover, it has been used for determination of visible osteolysis with the aim to provide more insight into the active disease state [17]. However, there is only few data on radiomics applications besides characterization of MM manifestations themselves. Notably, due to its ability to reflect the microstructure of tissues, radiomics could also be used to characterize the bone and bone marrow and may be able to even possibly predict the stem-cell mobilization in patients with MM before ASCT.
The purpose of this retrospective analysis was to investigate the associations between CT radiomics derived from iliac bone and the stem-cell graft yield of MM patients undergoing stem-cell mobilization after quadruplet induction therapy. Our group has already reported on the modalities of stem-cell collection in a bicentric analysis of this patient population (steady-state versus chemomobilization) [5]. Here we report on radiomics and stem-cell collection in the cohort of patients from our center.
Methods
This retrospective single center study was approved by the local ethics committee (Ethics Committee of the University of Leipzig, No. 344-2007 and No. 245/23-ek). All patients with newly diagnosed MM undergoing stem-cell mobilization between February 01, 2020, and September 01, 2022, with CT imaging prior to induction of chemomobilization were investigated in the present study. In total, 34 patients (25 males [74%]) have been included in this analysis. The median age at apheresis was 60 ± 8 years (range 36–71 years).
Stem-Cell Mobilization
All 34 patients had received CD38 antibodies and chemotherapy-based stem-cell mobilization. The median number of induction cycles prior to apheresis was 2 (varying from 2 to 4).
In 4 (11.8%) patients, apheresis could not be initiated due to medical reasons. These cases were successfully collected after a second mobilization attempt per steady-state.
CD34+ cell counts were measured prior to apheresis to estimate the mobilization success and to assess the necessity for additional administration of plerixafor. The collection of stem cells was performed at >10/μL CD34+ in peripheral blood with a target yield of ≥4 × 106 CD34 + cells/kg body weight. A total of 7 (20.6%) patients underwent additional treatment with plerixafor.
Stem-Cell Apheresis
All stem cell collections were performed on Spectra Optia (Terumo BCT) instruments using continuous MNC program, version 11 as previously described [5].
Evaluation of Graft Composition
The CD34+ cell counts in peripheral blood and in grafts were determined using a single platform flow cytometry assay (FACS Lyric, BD Sciences, Heidelberg, Germany). Immunophenotyping was performed with the CD34PE/CD45FITC reagent (8G12/2D1, BD Sciences, Heidelberg, Germany) according to previously described guidelines [18]. Release criteria included ≥2 × 106 CD34+ cells/kg body weight, negative sterile controls, ≥1 CFU-GM/105 cells, and ≥50% vitality after thawing.
Image Analysis
Image analysis was performed blinded to the clinical information by a trained reader with 5 years of general experience in radiology. In every patient, the contrast-enhanced CT of the abdomen and pelvis was used at baseline before chemomobilization. The CT imaging was performed on a 128-slice or 256-slice clinical CT scanner (Ingenuity or iCT256, Philips, Hamburg, Germany). The used imaging parameters were 120 kVp, 36 mAs, collimation of 64 × 0.6 mm and pitch of 0.8. The scan length included the following body regions: head, neck, chest, abdomen/pelvis, upper limb, and the proximal half of the lower limb. The minimal slice thickness was 1 mm.
The CT images were first analyzed with the Picture Archiving and Communication System (PACS) workstation (iDS7, Sectra AB, Linköping, Sweden). Then, one representative axial CT slice was chosen and extracted as a DICOM file. A round region of interest (ROI) was drawn in a standardized manner in both iliac bones. In every case, there were no osteolytic lesions included in the ROI. Texture analysis of the predefined ROIs was then conducted using MaZda software (version 4.7, available at http://www.eletel.p.lodz.pl/mazda/) [19, 20]. For each ROI, gray-level normalization was applied by constraining dynamics to μ ± 3 standard deviations, minimizing the influence of contrast and brightness variations, consistent with methods used in similar texture analysis studies [21]. For every patient, overall, 279 radiomics features were extracted from the CT images, of different groups derived from the first-order and second-order features (Gray-level histogram, Co-occurrence matrix, Run-length matrix, Absolute gradient, Autoregressive model, Wavelet transform). Figure 1 provides a representative patient of the patient cohort with the drawn ROI.
Fig. 1.
Axial CT image of a patient (58-year-old male) of our cohort. The CD34 cell count was 62/µL. The region of interest located within the iliac wings is highlighted in red.
Statistical Analyses
Statistical analysis was performed using the Anaconda Python distribution package (version 2024.06) and the statistics software SPSS (IBM SPSS Statistics for Windows, version 28: IBM corporation, NY, USA). To mitigate the potential impact of multicollinearity and to streamline the dataset for subsequent analyses, we performed a systematic feature reduction procedure. The process involved the following steps: (I) Correlation Matrix Computation: A correlation matrix was generated to assess the pairwise linear relationships between all features. The absolute values of Pearson’s correlation coefficients were considered to ensure that both positive and negative correlations were accounted for in the reduction process. (II) Identification of Redundant Features: To identify features that provide redundant information, the upper triangle of the correlation matrix was examined. Features that exhibited a high degree of correlation (Pearson’s |r| ≥ 0.7) with any other feature were flagged for removal. The choice of a 0.7 threshold was informed by standard practices in statistical analysis, where this level of correlation often suggests significant redundancy. (III) Feature Elimination: Features identified as redundant were systematically removed from the dataset. These remaining features were deemed sufficiently independent, providing unique information for subsequent modeling efforts. This approach ensured that the final dataset was composed of features that provided unique information and improving the interpretability of our analysis.
The collected data were evaluated by means of descriptive statistics (absolute and relative frequencies). A comparison of texture features and histopathological parameters in groups was performed by the Mann-Whitney-U testing. The correlation between texture features and the investigated stem-cell parameters was calculated by Spearman’s rank correlation coefficient. The diagnostic accuracy of the CT texture features was investigated with receiver operating-characteristics analysis with the outcome parameter of the area under the curve (AUC). A multivariate model was built to predict stem-cell parameters. In all instances, p values below 0.05 were considered statistically significant.
Results
At primary diagnosis, 15 (44%) patients showed an extensive plasma cell bone marrow infiltration (≥60%) by histology. R-ISS stages varied from I to III (stage I in 6 cases, stage II in 11 cases, stage III in 13 cases, nonavailable in 4 cases). Prior to apheresis five (14.7%) patients had already achieved a very good partial remission (VGPR) or better, according to International Myeloma Working Group criteria [22]. In one of the patients (3%), an additional irradiation therapy was performed prior to apheresis. Prior to apheresis, median CD34+ cell count in peripheral blood was 49/µL (IQR 42), for WBC it was 28.25/µL (IQR 34.3). Collection efficiency had a median of 0.46 (IQR 0.14). The demographic overview of the patient sample is provided in Table 1.
Table 1.
Overview of the patient sample
| | Total (n = 34) | No administration of Plerixafor (n = 27) | Prior administration of Plerixafor (n = 7) | p value |
|---|---|---|---|---|
| Age, years | 60.21±8.01 | 54.60±10.95 | 64.71±5.59 | 0.05 |
| Sex, male, n (%) | 25 | 20 (72) | 5 (71.4) | 0.43 |
| Weight, kg | 80.00±14.60 | 73.20±19.64 | 77.57±10.21 | 0.88 |
| MM subtype, n (%) | | | | 0.44 |
| IgG | 12 (35) | 11 | 1 | |
| IgA | 12 (35) | 8 | 4 | |
| Light chain MM | 10 (29) | 8 | 2 | |
| R-ISS-stage | | | | 0.87 |
| I | 6 (18) | 5 | 1 | |
| II | 11 (32) | 8 | 3 | |
| III | 13 (38) | 10 | 3 | |
| Unknown | 4 (12) | 4 | 0 | |
| Initial bone marrow plasma-cell infiltration | | | | 0.67 |
| <60% | 15 (44) | 12 | 3 | |
| ≥60% | 15 (44) | 12 | 3 | |
| Unknown | 4 (12) | 4 | 0 | |
| Number of induction cycles prior to apheresis, n (%) | | | | 0.82 |
| 2 | 8 (24) | 3 | 5 | |
| >2 | 26 (77) | 25 | 1 | |
| Remission prior to apheresis, n (%) | | | | 0.97 |
| ≥VGPR | 29 (85) | 24 | 5 | |
| <VGPR | 5 (15) | 4 | 1 | |
| CD34+ in peripheral blood, /μL | 62.47±34.64 | 51.59±15.96 | 38.11±18.89 | 0.20 |
| WBC peripheral blood, ×103/µL | 34.90±21.28 | 32.72±12.57 | 31.94±18.89 | 0.76 |
| Total CD34+ collected, ×106 | 538.41±354.72 | 363.13±154.26 | 260.00±112.35 | 0.27 |
| CD34+/kg body weight, ×106/kg | 6.59±3.79 | 4.85±1.39 | 3.45±1.73 | 0.11 |
| Collection efficacy | 149.87±61.39 | 151.10±101.06 | 108.78±18.91 | 0.64 |
VGPR, very good partial remission.
Correlation Analysis
Of the investigated CT radiomics features, three CT texture features (second-order features) were associated with the CD34+ cell count in peripheral blood. “S(1,0)AngScMom” and “WavEnHH_s-5” were positively correlated (r = 0.375, p = 0.031 and 0.432, p = 0.012, respectively), whereas “Teta1” was inversely correlated (r = −0.375, p = 0.031) (Table 2; Fig. 2–4). The other parameters were not associated with the CD34 cell count.
Table 2.
CT texture features (second-order features) associated with CD34+ cell count in peripheral blood
| Feature | Correlation coefficient (r) | p value |
|---|---|---|
| S(1,0)AngScMom | 0.375 | 0.03 |
| WavEnHH_s-5 | 0.432 | 0.01 |
| Teta1 | −0.375 | 0.03 |
Fig. 2.
Correlation analysis of texture feature “S(1,0)AngScMom” with CD34+ cell count in peripheral blood (r = 0.375, p = 0.03).
Fig. 4.
Correlation analysis of texture feature “WavEnHH_s-5” with CD34+ cell count in peripheral blood (r = 0.432, p = 0.01).
Fig. 3.
Correlation analysis of texture feature “Teta1” with CD34+ cell count in peripheral blood (r = −0.375, p = 0.03).
The parameter “HU mean” (Hounsfield units mean values) was statistically significant associated with white blood counts in peripheral blood on the day of apheresis (r = 0.361, p = 0.039). Collection parameters like absolute count of CD34+ cells in the collection, weight adapted CD34+ concentration and count of total nucleated cells were not associated with the CT radiomics features.
The need for plerixafor administration was associated with the texture feature “WavEnLL_s-1”. It was statistically significantly higher in patients with plerixaflor administration compared to those without (17,874.11 ± 132.28 versus 17,399.22 ± 174.37, p = 0.003).
The CD3+ cell yields were statistically significant associated with parameters_”MaxNorm” (r = 0.368, p = 0.038) and “HU mean” (r = 0.355, p = 0.046) and yields of natural-killer (NK) cells was associated with “WavEnLH_s-6” (r = 0.383, p = 0.028).
Prediction Analysis
The CT radiomics features (n = 3) were used to build a multivariate model to predict patients with a high mobilization defined by CD34+ cell count ≥40/µL. This model achieved an AUC of 0.77 (95% CI: 0.59–0.96) with a resulting optimized sensitivity of 0.87, specificity of 0.70 and an accuracy of 0.82. The corresponding graph is displayed in Figure 5.
Fig. 5.
Receiver operating-characteristics graph for the prediction of a high mobilization with CD34+ cell count ≥40/µL, using multivariate regression analysis incorporating texture features “S(1,0)AngScMom,” “Teta1,” and “WavEnHH_s-5.” This model achieved an AUC of 0.77 (95% CI: 0.59–0.96) with a resulting sensitivity of 0.87 and specificity of 0.70.
Considered independently, “S(1,0)AngScMom” reached an AUC of 0.72 (95% CI: 0.53–0.90) with an optimized sensitivity of 0.74, specificity of 0.70 and accuracy of 0.73. For “Teta1,” it reached an AUC of 0.77 (95% CI: 0.58–0.96) with a resulting sensitivity 0.80, specificity of 0.65 and accuracy of 0.75, while the parameter “WavEnHH_s-5” achieved an AUC of 0.68 (95% CI: 0.48–0.88) with a resulting sensitivity and specificity of both 0.70 and accuracy of 0.70.
Discussion
In this retrospective analysis, we elucidated the associations between CT radiomics features of the pelvic bone marrow and the hematopoietic cell mobilization in patients with MM. In short, promising CT radiomics features were identified, which could help to improve characterization and to provide new imaging markers in patients with MM. In a previous bicentric analysis, our group reported that stem-cell mobilization is feasible in the MM patient population treated with quadruplet induction therapy, both with and without chemotherapy [5]. In this present study, we analyzed the patients from our center under the hypothesis that radiomics as a novel quantitative imaging analysis could identify parameters that influence collection yields.
Radiomics offers a noninvasive tool that has been proven to be able to capture subtle tissue characteristics and therefore might provide data associated with stem-cell reserve and mobilization capacity [13, 14]. Radiomics provides new quantitative imaging markers and was used in several different clinical settings in MM using CT and MRI images [13].
All but one patient in our cohort were treated with a quadruplet induction therapy with daratumumab, bortezomib, thalidomide, and dexamethasone (DaraVTd) in analogy to the CASSIOPEIA phase III trial; the remaining patient was treated with isatuximab-based regimen [4]. In the daratumumab-containing arm of the CASSIOPEIA trial, the mean CD34+ yields were significantly lower with 6.7 × 106/kg body weight compared to 10.0 × 106/kg in the VTd arm [23]. Furthermore, trials like MASTER and GRIFFIN demonstrated also lower median CD34+ yields in anti-CD38 antibody containing arm with median values of 6 × 106/kg CD34+ and 8.3 × 106/kg CD34+ cells, respectively [24]. Finally, recently published PERSEUS trial, comparing quadruplet treatment containing daratumumab (Dara) bortezomib lenalidomide dexamethasone (VRd) vs. VRd, also showed lower yields of CD34+ cells in the experimental arm with 5.5 × 106/kg CD34+ and 7.4 × 106 CD34+/kg, respectively [4].
Therefore, it is of clinical importance to better identify patients at risk for low target collection, with noninvasive tools like quantitative imaging modalities. Radiomics has still not been used to reflect bone marrow characteristics of hematological diseases. However, there are several promising reports about the use of radiomics to improve diagnostic imaging and prognostication in patients with MM [16].
Another recent study used radiomics features derived from T1- and T2-weighted images from MM patients to predict high-risk cytogenetic abnormalities with a high accuracy [25]. In that study, the largest myeloma manifestation was segmented as a target lesion for the analysis and used the tissue composition of the myeloma infiltration defined by MRI images. It was further used to enhance the diagnostic accuracy and the discrimination from spinal metastasis in an MRI and CT study, respectively, to explore the potential diagnostic benefit of radiomics analyses in clinical routine [26, 27].
Furthermore, an interesting study used a segmentation of the whole skeleton to extract CT radiomics features of 98 patients with MM, which resulted in a model being able to predict the patient’s prognosis with, however, only a moderate accuracy [28]. The whole skeleton segmentation method to extract radiomics features from FDG-PET/CT images was used in another study from Spain for assessment of measurable residual disease in bone marrow biopsies of patients with MM [29].
Our present analysis, however, sought to investigate the potential benefit of the radiomics features extracted from the iliac bone, the common localization for bone marrow analyses in clinical routine. In the present study, radiomics features were associated with CD34+ cells in peripheral blood prior to apheresis, CD3+ cell collection and with NK cell collection. It seems promising that CT radiomics may also be associated with distinctive cell fractions of the serum.
To date, there are only studies investigating associations between CT radiomics and different cellular fractions of the primary tumor, which cannot be directly compared to the approach of the present study [30–32]. It can therefore be presumed that CT radiomics features can reflect the underlying histopathology of the bone and bone marrow.
The most promising texture features in the present study are from the wavelet transform group, which decomposes the images and allows the extraction of more features from the images [33]. It seems reasonable to assume that more complex texture parameters – reflecting a higher degree of heterogeneity in CT images – may better reflect the composition of bone marrow compared to the simpler Hounsfield unit measurements used in conventional CT analysis. A preliminary study previously explored the relationship between Hounsfield density on CT images and bone marrow characteristics in patients with systemic mastocytosis [34]. The study reported a moderate positive correlation between the Hounsfield units of the iliac bone and both the mast cell proportion and the number of CD177-positive cells. However, further research is clearly needed to better understand the complex associations between quantitative imaging parameters and bone marrow composition in different hematological disorders.
Beyond the scope of the present analysis is that only the combination of multiple methods, such as bone marrow aspirates, biopsy, and flow cytometry, enhances accuracy of diagnosis, and classification of plasma cell neoplasms. Presumably, the incorporation of CT radiomics in addition to these methods may refine existing methods and ultimately enhance clinical care.
The present study is not free from limitations. First, it is a retrospective analysis with possible inherent bias. However, the imaging analysis was performed blinded to the clinical results to reduce possible bias. Second, the imaging analysis was only performed by one trained reader. There may be some inter reader heterogeneity for the ROI placement. Third, a validation cohort is missing to test our identified CT radiomics features in another patient cohort.
Furthermore, the utilization of CT examination in the routine praxis for stem-cell mobilization is limited by reimbursement issues, challenges regarding the management of appointments and the radiation exposure. This should especially be acknowledged for routine clinical work-up outside clinical trials. As another aspect, the evaluation of CT findings needs to be standardized and the external validation of the radiomics features needs to be tested. Another additional issue is the aspect of the modality of stem-cell mobilization, i.e., chemo- vs. steady-state. The examination of the constitutive parameters of chemomobilization, i.e., chemotherapy and granulocyte colony stimulating factor component on radiomics parameters is challenging, among other reasons due to application of X-rays in short interval. Therefore, it would be relevant to validate our results in a steady-state cohort.
Conclusion
CT radiomics analysis can be used as to assess the distinct cell populations in MM patients undergoing stem-cell mobilization. This needs further exploration in other diseases and patient cohorts, and a further validation is needed whether CT radiomics could help better characterize the bone marrow status in a noninvasive manner.
Statement of Ethics
This retrospective single center study was approved by the Local Ethics Committee (Ethics Committee of the University of Leipzig, No. 344-2007 and No. 245/23-ek). Written informed consent from participants was not required in accordance with local/national guidelines.
Conflict of Interest Statement
The authors have no conflicts of interest to declare.
Funding Sources
The authors received no financial support for the research, authorship, and/or publication of this manuscript.
Author Contributions
J.L., V.V., and H.J.M. wrote the manuscript. T.D. provided the calculations and the radiomics data. E.B. and M.B. provided the data regarding flow cytometry. S.Y.W., G.N.F., M.J., S.H., K.H.M., U.P., T.D., and M.M. provided administrational support. All authors contributed to the interpretation of the results and editing of the manuscript and agreed on the final version.
Funding Statement
The authors received no financial support for the research, authorship, and/or publication of this manuscript.
Data Availability Statement
All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding authors.
References
- 1. Attal M, Lauwers-Cances V, Hulin C, Leleu X, Caillot D, Escoffre M, et al. Lenalidomide, bortezomib, and dexamethasone with transplantation for myeloma. N Engl J Med. 2017;376(14):1311–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Cavo M, Gay F, Beksac M, Pantani L, Petrucci MT, Dimopoulos MA, et al. Autologous haematopoietic stem-cell transplantation versus bortezomib-melphalan-prednisone, with or without bortezomib-lenalidomide-dexamethasone consolidation therapy, and lenalidomide maintenance for newly diagnosed multiple myeloma (EMN02/HO95): a multicentre, randomised, open-label, phase 3 study. Lancet Haematol. 2020;7(6):e456–68. [DOI] [PubMed] [Google Scholar]
- 3. Richardson PG, Jacobus SJ, Weller EA, Hassoun H, Lonial S, Raje NS, et al. Triplet therapy, transplantation, and maintenance until progression in myeloma. N Engl J Med. 2022;387(2):132–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Sonneveld P, Dimopoulos MA, Boccadoro M, Quach H, Ho PJ, Beksac M, et al. Daratumumab, Bortezomib, lenalidomide, and dexamethasone for multiple myeloma. N Engl J Med. 2024;390(4):301–13. [DOI] [PubMed] [Google Scholar]
- 5. Teipel R, Rieprecht S, Trautmann-Grill K, Röllig C, Klötzer C, Zimmer K, et al. Steady-state versus chemotherapy-based hematopoietic cell mobilization after anti-CD38-based induction therapy in newly diagnosed multiple myeloma. Transfusion. 2023;63(11):2131–9. [DOI] [PubMed] [Google Scholar]
- 6. Bittrich M, Kriegsmann K, Tietze-Stolley C, Movassaghi K, Grube M, Vucinic V, et al. A German-wide systematic study on mobilization and collection of hematopoietic stem cells in poor mobilizer patients with multiple myeloma prior to autologous stem cell transplantation. Transfus Med Hemother. 2023;50(6):475–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Attolico I, Pavone V, Ostuni A, Rossini B, Musso M, Crescimanno A, et al. Plerixafor added to chemotherapy plus G-CSF is safe and allows adequate PBSC collection in predicted poor mobilizer patients with multiple myeloma or lymphoma. Biol Blood Marrow Transplant. 2012;18(2):241–9. [DOI] [PubMed] [Google Scholar]
- 8. Gopal AK, Karami M, Mayor J, Macebeo M, Linenberger M, Bensinger WI, et al. The effective use of plerixafor as a real-time rescue strategy for patients poorly mobilizing autologous CD34(+) cells. J Clin Apher. 2012;27(2):81–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hundemer M, Engelhardt M, Bruckner T, Kraeker S, Schmitt A, Sauer S, et al. Rescue stem cell mobilization with plerixafor economizes leukapheresis in patients with multiple myeloma. J Clin Apher. 2014;29(6):299–304. [DOI] [PubMed] [Google Scholar]
- 10. Wuchter P, Ran D, Bruckner T, Schmitt T, Witzens-Harig M, Neben K, et al. Poor mobilization of hematopoietic stem cells-definitions, incidence, risk factors, and impact on outcome of autologous transplantation. Biol Blood Marrow Transplant. 2010;16(4):490–9. [DOI] [PubMed] [Google Scholar]
- 11. Sauer S, Erdmann K, Jensen AD, Wennmann M, Pavel P, Jordan K, et al. Local radiation therapy before and during induction delays stem cell mobilization and collection in multiple myeloma patients. Transplant Cell Ther. 2021;27(10):876.e1–1. [DOI] [PubMed] [Google Scholar]
- 12. Ray GK, Jena RK, Panda T, Sethy S. Prospective identification of potential factors influencing stem cell mobilization and the necessity for plerixafor use in newly diagnosed multiple myeloma patients undergoing autologous stem cell transplantation. Hematol Transfus Cell Ther. 2021;43(4):402–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Michalska-Foryszewska A, Rogowska A, Kwiatkowska-Miernik A, Sklinda K, Mruk B, Hus I, et al. Role of imaging in multiple myeloma: a potential opportunity for quantitative imaging and radiomics? Cancers. 2024;16(23):4099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Zhong H, Huang D, Wu J, Chen X, Chen Y, Huang C. 18F-FDG PET/CT based radiomics features improve prediction of prognosis: multiple machine learning algorithms and multimodality applications for multiple myeloma. BMC Med Imaging. 2023;23(1):87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Incoronato M, Aiello M, Infante T, Cavaliere C, Grimaldi AM, Mirabelli P, et al. Radiogenomic analysis of oncological data: a technical survey. Int J Mol Sci. 2017;18(4):805. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Sachpekidis C, Goldschmidt H, Edenbrandt L, Dimitrakopoulou-Strauss A. Radiomics and artificial intelligence landscape for 18FFDG PET/CT in multiple myeloma. Semin Nucl Med. 2025;55(3):387–95. [DOI] [PubMed] [Google Scholar]
- 17. An S, Huang G, Yu X, Liu J, Chen Y. The added diagnostic value of 18 F-FDG PET/CT radiomic analysis in multiple myeloma patients with negative visual analysis. Nucl Med Commun. 2024;45(3):244–52. [DOI] [PubMed] [Google Scholar]
- 18. Sutherland DR, Anderson L, Keeney M, Nayar R, Chin-Yee I. The ISHAGE guidelines for CD34+ cell determination by flow cytometry. International society of hematotherapy and graft engineering. J Hematother. 1996;5(3):213–26. [DOI] [PubMed] [Google Scholar]
- 19. Strzelecki M, Szczypinski P, Materka A, Klepaczko A. A software tool for automatic classification and segmentation of 2D/3D medical images. Nucl Instr Methods Phys Res Section A: Acc Spectrometers, Detectors Associated Equipment. 2013;702:137–40. [Google Scholar]
- 20. Szczypiński PM, Strzelecki M, Materka A, Klepaczko A. MaZda: a software package for image texture analysis. Comput Methods Programs Biomed. 2009;94(1):66–76. [DOI] [PubMed] [Google Scholar]
- 21. Meyer HJ, Leonhardi J, Höhn AK, Pappisch J, Wirtz H, Denecke T, et al. CT texture analysis of pulmonary neuroendocrine tumors-associations with tumor grading and proliferation. J Clin Med. 2021;10(23):5571. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Kumar S, Paiva B, Anderson KC, Durie B, Landgren O, Moreau P, et al. International myeloma working Group consensus criteria for response and minimal residual disease assessment in multiple myeloma. Lancet Oncol. 2016;17(8):e328–46. [DOI] [PubMed] [Google Scholar]
- 23. Moreau P, Attal M, Hulin C, Arnulf B, Belhadj K, Benboubker L, et al. Bortezomib, thalidomide, and dexamethasone with or without daratumumab before and after autologous stem-cell transplantation for newly diagnosed multiple myeloma (CASSIOPEIA): a randomised, open-label, phase 3 study. Lancet. 2019;394(10192):29–38. [DOI] [PubMed] [Google Scholar]
- 24. Chhabra S, Callander N, Watts NL, Costa LJ, Thapa B, Kaufman JL, et al. Stem cell mobilization yields with Daratumumab- and Lenalidomide-Containing quadruplet induction therapy in newly diagnosed multiple myeloma: findings from the MASTER and GRIFFIN trials. Transplant Cell Ther. 2023;29(3):174.e1–0. [DOI] [PubMed] [Google Scholar]
- 25. Liu S, Liu C, Pan H, Li S, Teng P, Li Z, et al. Magnetic resonance imaging-based nomograms predict high-risk cytogenetic abnormalities in multiple myeloma: a two-centre study. Clin Radiol. 2025;82:106768. [DOI] [PubMed] [Google Scholar]
- 26. Cao J, Li Q, Zhang H, Wu Y, Wang X, Ding S, et al. Radiomics model based on MRI to differentiate spinal multiple myeloma from metastases: a two-center study. J Bone Oncol. 2024;45:100599. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Özgül HA, Akin IB, Mutlu U, Balci A. Diagnostic value of machine learning-based computed tomography texture analysis for differentiating multiple myeloma from osteolytic metastatic bone lesions in the peripheral skeleton. Skeletal Radiol. 2023;52(9):1703–11. [DOI] [PubMed] [Google Scholar]
- 28. Ni B, Huang G, Huang H, Wang T, Han X, Shen L, et al. Machine learning model based on optimized radiomics feature from 18F-FDG-PET/CT and clinical characteristics predicts prognosis of multiple myeloma: a preliminary Study. J Clin Med. 2023;12(6):2280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Milara E, Gómez-Grande A, Tomás-Soler S, Seiffert AP, Alonso R, Gómez EJ, et al. Bone marrow segmentation and radiomics analysis of 18FFDG PET/CT images for measurable residual disease assessment in multiple myeloma. Comput Methods Programs Biomed. 2022;225:107083. [DOI] [PubMed] [Google Scholar]
- 30. Zhou J, Zou S, Kuang D, Yan J, Zhao J, Zhu X. A novel approach using FDG-PET/CT-Based radiomics to assess tumor immune phenotypes in patients with non-small cell lung cancer. Front Oncol. 2021;11:769272. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Tong H, Sun J, Fang J, Zhang M, Liu H, Xia R, et al. A machine learning model based on PET/CT radiomics and clinical characteristics predicts tumor immune profiles in non-small cell lung cancer: a retrospective multicohort study. Front Immunol. 2022;13:859323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Park C, Jeong DY, Choi Y, Oh YJ, Kim J, Ryu J, et al. Tumor-infiltrating lymphocyte enrichment predicted by CT radiomics analysis is associated with clinical outcomes of non-small cell lung cancer patients receiving immune checkpoint inhibitors. Front Immunol. 2022;13:1038089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Chang T, Kuo CJ. Texture analysis and classification with tree-structured wavelet transform. IEEE Trans Image Process. 1993;2(4):429–41. [DOI] [PubMed] [Google Scholar]
- 34. Meyer HJ, Pönisch W, Monecke A, Gundermann P, Surov A. Can diagnostic low-dose whole-body CT reflect bone marrow findings in systemic mastocytosis? Anticancer Res. 2020;40(2):1015–22. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding authors.