RPS 2110 - Imaging musculoskeletal tumours

Author Block: J. Hofmann, S. Afat, K. Nikolaou, J. Herrmann; Tübingen/DE
Purpose: Whole-body MRI (WB-MRI) is routine but time-intensive examination in patients with multiple myeloma. Prolonged scan duration may decrease image quality through motion artifacts, reduce throughput, and increase patient discomfort. This study aimed to evaluate the impact of deep learning–accelerated diffusion-weighted imaging (DWIDL) in 3T WB-MRI on overall and medullary image quality, sharpness, noise, lesion detectability, and diagnostic confidence.
Methods or Background: This retrospective study included 40 patients with multiple myeloma (21 men, 19 women; mean age, 62 ± 11 years; range, 35–82 years) who underwent 3T WB-MRI, including DWI with two b-values (50 and 800 s/mm²) between February 2023 and April 2024. Single-average raw data were processed using a deep learning (DL) image reconstruction algorithm, yielding a simulated acquisition time of 3 minutes 3 seconds for DWIDL compared with 7 minutes 6 seconds for standard DWI (DWIStd) by reducing the number of signal averages. Two experienced radiologists independently evaluated all DWI datasets using a 5-point Likert scale assessing noise level, artifact severity, sharpness, overall image quality, and diagnostic confidence.
Results or Findings: There were no significant differences between DWIDL and DWIStd in artifact severity, sharpness, overall image quality, lesion detectability, or diagnostic confidence. However, DWIDL demonstrated significantly lower noise levels and higher contrast compared with standard DWI (P < .05).
Conclusion: Deep learning–based image reconstruction for WB-DWI at 3T is feasible and enables a significant reduction in acquisition time without compromising image quality. Accelerated acquisition particularly benefits multiple myeloma patients by reducing motion artifacts, improving comfort, and potentially enabling MRI instead of CT for better assessment of bone marrow infiltration.
Limitations: This retrospective single-center study with a limited sample size, subjective image quality assessment, and no cost or workflow analysis limits the generalizability and strength of the findings.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: University of Tuebingen

Author Block: R. F. Valenzuela, E. Duran-Sierra, M. Antony, J. Espinoza, C. M. Costelloe, J. E. Madewell, W. Murphy, B. Amini, S. Lo; Houston, TX/US
Purpose: Desmoid fibromatosis patients are typically treated with active surveillance and drugs on progression, particularly sorafenib, and most recently, with nirogacestat. This study aimed to determine the signal and volume changes after therapy and over time, comparing both drugs using MRI T2-WI-derived features, including modified Choi (mChoi) and volumetric measurements.
Methods or Background: Retrospective study including 17 patients with single-lesion extremity desmoid fibromatosis and standard-of-care MRI, including T2-WI, from March 2021-February 2024. Volumetric tumor segmentations were created on T2-STIR images. Diameter, mChoi, and volume measurements were computed at four time points across the patient’s treatment: Pre-treatment (Pre-Rx), post-treatment 1 (Rx1), post-treatment 2 (Rx2), and post-treatment 3 (Rx3). The percentage change in diameter (RECIST), mChoi, and volume across time were computed with respect to Rx1 and compared in nirogacestat vs. sorafenib patients.
Results or Findings: Eight patients were treated with nirogacestat and nine patients with sorafenib. Nirogacestat-group: 1) mChoi detected two cases of progression not detected by RECIST or volume, 2) from Pre-Rx to Rx3, the average mChoi decreased by 66% over 20 months (-3.3%/month), and 3) from Pre-Rx to Rx3, the average volume reduced by 49% over 20 months (-2.45%/month). Sorafenib group: 1) RECIST, mChoi, and volume detected two progressions. mChoi detected one response not detected by RECIST or volume, 2) from Pre-Rx to Rx3, the average mChoi decreased by 49% over 34 months (-1.44%/month), and 3) from Pre-Rx to Rx3, the average volume decreased by 4% over 34 months (-0.09%/month).
Conclusion: Patients treated with nirogacestat exhibited larger volumetric and mChoi changes, with the fastest response and similar sustained positive treatment effects compared to those treated with sorafenib, indicating the potential of nirogacestat as an effective treatment option for desmoid tumors.
Limitations: Small sample size.
Funding for this study: The John S. Dunn, Sr. Distinguished Chair in Diagnostic imaging and M.R Evelyn Hudson Foundation Endowed Professorship.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: R. Perraut-Gattegno, J. Stadelmann, P-A. A. Poletti, S. Boudabbous; Geneva/CH
Purpose: The aim of the study was to assess if some features of radiomics are common to extra-skeletal chondrosarcoma and skeletal chondrosarcoma, and to distinguish this rare entity from the most frequent encountered soft tissue sarcoma (myxoid liposarcoma and pleomorphic sarcoma) in MRI imaging.
Methods or Background: Retrospective study on our radiological data base (PACS archive) confronted to data from multidisciplinary sarcoma Concilium, with the evaluation of 150 radiomics features were extracted from volume-of-interest on T1 and T2-weighted spin-echo MRI, from 19 tumor cases (4 ESC, 5 CS, 5 PS and 5 MLS, based on histopathological diagnosis). These radiomics features were then used to train Random Forests to select the best 10 features for the classification of the 4 types of tumors. Then we used a logistic regression to evaluate the best pair of features for dichotomy classification: EOC and OC versus the other types (MLS, PS).
Results or Findings: The visual evaluation of the feature scatter plot confirms the separability of the ESC and SC against the other types of tumours. The evaluation of the logistic regression resulted of a sensitivity of 0.93 specificity of 0.92 for an overall accuracy of 0.92 with the following features: GLCM inverse difference moment normalized (on T2 sequences) and GLCM correlation (on T1 sequences). We consider this preliminary study promising for the use of radiomic features in the differentiation of extra-skeletal and skeletal chondrosarcoma from other soft tissue tumors.
Conclusion: Despite being developed on few cases, 2 radiomics features accurately distinguishes the chondrosarcoma group (ESC + SC) from the MLS and PS on MRI with an overall accuracy of 0.92. We as well evaluated that the chondrosarcoma group had identical profiles on all the radiomic features tested.
Limitations: Small number of cases
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Geneva ethics committee.

Author Block: R. F. Valenzuela, E. Duran-Sierra, M. Antony, J. Espinoza, C. M. Costelloe, J. E. Madewell, W. Murphy, D. Wells, B. Amini; Houston, TX/US
Purpose: Undifferentiated pleomorphic sarcoma (UPS) patients are typically treated with either consecutive chemotherapy and radiotherapy (CT-RT) or RT alone. This study aimed to determine the impact of the treatment strategy on imaging and pathology in UPS response assessment based on multiparametric MRI (mp-MRI).
Methods or Background: This retrospective study included 30 extremity UPS patients with preoperative mp-MRI, including contrast-enhanced susceptibility-weighted-imaging (CE-SWI) and perfusion-weighted-imaging with dynamic-contrast-enhancement (PWI/DCE), with surgery performed from February 2021-May 2023. Lesions were visually classified on CE-SWI into one of six morphology patterns. On PWI/DCE, lesions were classified into one of six enhancement patterns. Time-intensity curves (TICs) were classified as types I-V. Patients demonstrating >=90% pathology-assessed-treatment-effect (PATE) on the surgical specimen were labeled as responders (n=15), and those with <90% PATE were labeled as partial/non-responders (n=15).
Results or Findings: 19 patients were treated with presurgical consecutive CT-RT, and 11 with presurgical RT. CT-RT group displayed a higher number of responders (53%) vs. RT (45%). Consecutive CT-RT demonstrated a higher average PATE (78%) compared to the RT group (63%). CT-RT group: 80% of responders displayed a CE-SWI complete-ring pattern (P=1.77x10-5), 90% a PWI/DCE capsular-enhancement (P=3.06x10-6), and 100% a TIC II (P=4.54x10-5). A predictive model combining the PWI/DCE TIC II, PWI/DCE capsular-enhancement, and CE-SWI complete-ring yielded the highest classification performance (AUC = 0.98), discriminating responders from partial/non-responders, outperforming RECIST (AUC = 0.49). RT group: 100% of responders displayed a TIC II (P=0.0253). Perfect classification performance was achieved by a model based on the PWI/DCE TIC II (AUC=1.0).
Conclusion: Predictive models based on PWI/DCE and CE-SWI features can outperform RECIST, predicting pathology-assessed response. Furthermore, classification models that best separate responders from partial/non-responders can vary not only by tumor type but also by treatment regimen.
Limitations: Small sample size.
Funding for this study: The John S. Dunn, Sr. Distinguished Chair in Diagnostic imaging and M.R Evelyn Hudson Foundation Endowed Professorship.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: F. P. Papa¹, R. Cuocolo², F. Molea¹, L. D'Ambrosio³, G. Grignani⁴, T. Robba¹; ¹Turin/IT, ²Naples/IT, ³Orbassano/IT, ⁴Candiolo/IT
Purpose: Synovial sarcoma is an aggressive malignancy that necessitates innovative diagnostic and therapeutic strategies. Predicting chemotherapy response is a pivotal step in the diagnostic-therapeutic pathway, enabling personalized treatment approaches. Radiomics, combined with machine learning, offers a promising avenue to identify imaging biomarkers that predict treatment response.
Methods or Background: In this retrospective, monocentric study, 51 patients diagnosed with synovial sarcoma underwent peri-treatment MRI acquisition. T1-weighted sequences were analyzed to extract 1116 radiomic features, which were subjected to feature selection processes. Machine learning models were developed to predict chemotherapy response. Data were intentionally gathered from heterogeneous MRI scanners to ensure the robustness of the models against clinical variability. Performance metrics included area under the curve (AUC) and accuracy assessments on both training and validation sets.
Results or Findings: The final ExtraTrees model, trained on 20 selected features, showed promising performance on the training set with an accuracy of 0.81 (±0.34), precision of 0.88 (±0.37), recall of 0.87 (±0.42), F1-score of 0.84 (±0.28), and AUC of 0.89 (±0.35), despite high variability.
On the test set (n=17), accuracy was 64.7%, with good sensitivity for non-responders (80%) but lower for responders (58%). Precision was high for responders (0.88) but lower for non-responders (0.44). The test AUC was modest (0.60), while the Precision-Recall Curve (0.72) indicated reasonable precision for the positive class. The Brier score (0.250) suggested moderate model calibration.
Conclusion: This study highlights the potential of radiomics to predict chemotherapy response in synovial sarcoma, despite challenges posed by scanner heterogeneity and intra-reader variability. Further research is required to refine these methods and integrate them into routine clinical practice.
Limitations: This study is limited by single-reader segmentation, small sample size, and lack of external validation.
Funding for this study: No fundings
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: Y. ABDELOUAHAB, F. Mihoubi Bouvier, J-L. Drape, A. Feydy, P. Anract; Paris/FR
Purpose: Massive knee prostheses are increasingly used after tumor resections or complex revisions. While they allow limb preservation, their imaging follow-up remains poorly standardized, and no dedicated imaging study had previously focused on these implants.
Objectives : (i) To describe physiological and pathological imaging features of massive knee prostheses and propose a surveillance scheme; (ii) to estimate complication rates, implant survival, and associated prognostic factors.
Methods or Background: A retrospective monocentric study was conducted at Cochin Hospital (2010–2022), including 88 patients with distal femoral (n = 71) or proximal tibial (n = 17) prostheses. Over 1,000 imaging examinations were analyzed (radiographs, CT, MRI, EOS, ultrasound). Complications were classified using Henderson’s system, expanded with a type VI neuro-functional category.
Results or Findings: Median follow-up was 37 months. Implant survival was 50% at 37 months (mean 58). The complication rate was 73%: infections (30%), aseptic loosening (13.2%), neuro-functional complications (13.2%), and recurrences (15%). Radiography remained central for mechanical issues. Peri-prosthetic ossifications developed around the prosthetic collar: ossification without osteointegration was strongly associated with loosening (61.9% vs 6.9% and 18.2%), while ossification with osteointegration was protective. New calcifications beyond six months, especially in osteosarcoma patients, were highly suggestive of recurrence. Ultrasound proved particularly effective in the popliteal fossa (recurrences, thrombosis, peri-prosthetic changes), while MRI was more useful for soft tissues outside this area.
Conclusion: This first dedicated imaging study on massive knee prostheses establishes peri-collar ossification without osteointegration as a prognostic marker. Above all, it highlights the novel role of ultrasound as a non-invasive, accessible, and highly sensitive modality for surveillance—particularly in the popliteal fossa—offering a decisive new tool to improve early detection and long-term outcomes.
Limitations: Retrospective, single-center design and heterogeneous imaging protocols may limit generalizability.
Funding for this study: This study received no specific funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The Ethical Review Committee for publications of the Cochin
university Hospital (CLEP)

Author Block: P. Spinnato¹, M. Colangeli¹, G. Bilancia¹, A. Crombé², M. Simonetti¹; ¹Bologna/IT, ²Talence/FR
Purpose: Adamantinoma is a primary malignant bone tumor that is extremely uncommon. The categorization of metastatic risk at diagnosis is difficult since there is currently no histological grading. Imaging may therefore be useful in predicting prognosis.
Our goal was to assess baseline imaging characteristics and how they relate to the emergence of metastatic disease.
Methods or Background: We retrospectively collected baseline radiological (CT, MRI, and conventional radiography) and clinical (metastatic disease) data from all consecutive patients with a histological diagnosis of adamantinoma between 2006 and 2022 at our single Sarcoma Center. Lodwick-Madewell grading, periosteal reaction, multifocality, soft-tissue extraskeletal component, tumor location, size, main radiological pattern (lytic, sclerotic, mixed), peritumoral edema, and vascular invasion were all examined. At diagnosis or during follow-up, correlations between the radiological characteristics listed above and metastatic disease were evaluated.
Results or Findings: Twenty-two patients were included (15 female, 7 male - mean age 28.8±13.3 years, range 7-58 years old). Six out of 22 patients (27.3%) developed distant metastases at diagnosis or during follow-up controls (minimum follow-up 24 months). The following radiological features were significantly correlated with the development of metastases at baseline or during follow-up controls: presence of an
extra-skeletal component (p=0.0004), vascular invasion (p=0.0002), and diffuse peritumoral edema (p=0.0004).
A risk model including two of the above mentioned imaging features, provided a sensibility of 100% (54.07%-100%), a specificity of 93.75% (69.67%-99.84%), with an accuracy of 95.45% (77.16%-99.88%) in predicting metastatic disease.
Conclusion: An accurate evaluation of baseline imaging studies in patients affected by adamantinoma may significantly aid in prognosis prediction and the selection of high metastatic-risk patients. For these patients, strict follow-up controls and more aggressive treatments should be suggested after multidisciplinary discussions in sarcoma centers.
Limitations: Retrospective, single-center analysis.
Funding for this study: No funding received.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: CE-AVEC code: RX.RARE.SARC (2025)

Author Block: T. B. Rodrigues, C. Rodrigues, A. M. Gaspar, M. Ramalho; Lisbon/PT
Purpose: To quantitatively compare the contrast enhancement dynamics of lipid-poor and lipid-rich vertebral hemangiomas on MRI and assess whether enhancement parameters can reliably distinguish these subtypes.
Methods or Background: Lipid-poor vertebral hemangiomas are rare and diagnostically challenging due to their high T2 signal intensity on fat-suppressed MRI sequences, often mimicking more aggressive lesions. A word-search retrospective analysis included 43 vertebral hemangiomas: 21 lipid-poor (12 females, 9 males; mean age 58.6 ± 13.7 years; mean lesion size 15.5 ± 7.4 mm) and 22 lipid-rich (15 females, 7 males; mean age 57.0 ± 12.6 years; mean size 11.5 ± 5.7 mm). Lesions were confirmed via CT attenuation or MRI follow-up stability. Signal intensity was measured using a fixed region of interest across pre-contrast T1-weighted fat-suppressed, arterial, venous, and delayed (interstitial) phases. Percentage enhancement was calculated relative to pre-contrast signal. Statistical comparisons employed Student’s t-test, Mann-Whitney U test, and Chi-square test.
Results or Findings: Lipid-poor hemangiomas exhibited significantly greater enhancement across all post-contrast phases compared to lipid-rich counterparts. In the arterial phase, lipid-poor lesions showed a mean enhancement of 206.8 ± 131.8%, versus 118.0 ± 82% for lipid-rich lesions (p< 0.05). Venous phase enhancement was also higher in lipid-poor hemangiomas (278.1 ± 117.0%) than in lipid-rich ones (166.3 ± 97.7%, p < 0.05). In the delayed (interstitial) phase, lipid-poor enhancement (267.0 ± 100.2%) remained markedly superior to lipid-rich enhancement (139.4 ± 90.9%, p < 0.05). No significant differences were found in age or sex (p > 0.05), though lipid-poor lesions trended toward larger size (p = 0.051).
Conclusion: Lipid-poor vertebral hemangiomas demonstrate significantly stronger and more persistent enhancement than lipid-rich lesions, suggesting that quantitative enhancement metrics may serve as a valuable tool for subtype differentiation on MRI.
Limitations: Retrospective; Small sample size
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: R. Marcello¹, A. Ciabattoni¹, M. Cuccarelli², G. Marcello², S. Vitale¹; ¹Roma/IT, ²Rome/IT
Purpose: Pathological fractures due to metastatic bone lesions are associated with significant pain and functional impairment. Minimally invasive treatments such as kyphoplasty and osteoplasty offer symptomatic relief, but the risk of cement leakage remains a major concern. This study evaluates the safety and efficacy of bipolar radiofrequency ablation (RFA) prior to cement injection to improve procedural control and patient outcomes.
Methods or Background: We retrospectively reviewed 37 patients with symptomatic malignant pathological fractures who underwent bipolar RFA-assisted kyphoplasty (n=31) or osteoplasty (n=6) between January 2011 and May 2019. Bipolar RFA was performed using the Osteocool RF Ablation System (Medtronic), achieving a constant temperature of 70°C over 7–15 minutes. Cement (PMMA) was then injected into the thermally ablated cavity. Clinical outcomes were assessed via VAS pain scores, and radiological results were evaluated by MR to determine cement distribution and leakage.
Results or Findings: All procedures were technically successful. No peri-procedural morbidity or mortality occurred. Post-procedural VAS scores demonstrated a significant decrease in pain for all patients (p < 0.05). MR imaging showed no significant venous or cortical cement leakage. The thermally created cavity facilitated controlled cement deposition.
Conclusion: Bipolar RFA-assisted cementoplasty appears to be a safe and effective technique for managing malignant pathological fractures. This approach reduces the risk of cement leakage while improving pain control and procedural precision. It may serve as a valuable adjunct in interventional strategies for metastatic bone disease. Prospective studies are needed to confirm these findings.
Limitations: Any
Funding for this study: Any
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: Y. Li¹, W. Zhang¹, L. Ma¹, L. Liangjie², Y. Zhang¹; ¹Zhengzhou/CN, ²Beijing/CN
Purpose: Amide proton transfer weighted imaging (APTWI) has been widely explored for tumor evaluation. This study attempted to evaluate the added value of APTWI to multiple model diffusion imaging for differentiating benign and malignant bone tumors.
Methods or Background: This prospective study included 68 patients with pathologically confirmed bone tumors (25 benign and 43 malignant) . All patients underwent MRI examinations including conventional DWI, intravoxel incoherent motion (IVIM), diffusion kurtosis imaging (DKI), and APTWI. The apparent diffusion coefficient (ADC), pure molecular diffusion (D), perfusion fraction (f), pseudo-diffusion (D*), mean kurtosis (MK), and mean diffusivity (MD) were calculated from multiple diffusion models. The magnetization transfer ratio asymmetry at a chemical shift of 3.5 ppm [MTRasym(3.5 ppm)] were derived from APTWI. Mann-Whitney U test or independent samples t-test was used for group comparison between benign and malignant tumors. Logistic regression modeling was used to identify independent predictors and performed combination diagnosis. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance.
Results or Findings: Malignant tumors showed significantly lower ADC, D, and MD, while significantly higher MK and MTRasym(3.5 ppm) values than those of benign tumors. ROC analysis yielded AUC values of 0.838 (ADC), 0.826 (D), 0.768 (MD), 0.820 (MK), and 0.761 [MTRasym(3.5 ppm)], respectively, for differentiation between benign and malignant bone tumors. Multivariate logistic regression revealed that MTRasym(3.5 ppm) and MK were independent factors for prediction of benign and malignant tumors. The combination of MTRasym(3.5 ppm) and MK showed significant improved diagnostic performance with the AUC of 0.901.
Conclusion: The integration of APTWI and advanced diffusion models provides additional insights into tumor metabolism and microstructure of bone tumors, thus significantly enhancing the diagnostic efficacy.
Limitations: Single-center study, small tumor cohort, and lack of external validation.
Funding for this study: We don't have any funding support.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective investigation was permitted by the Review Committee of the First Affiliated Hospital of Zhengzhou University (2023-KY-0888-003).

Author Block: B. Bogner, A. Runkel, M. Reisert, T. D. Diallo, P. M. Jungmann, F. Bamberg, M. Jung; Freiburg Im Breisgau/DE
Purpose: To identify distinct treatment response patterns in soft tissue sarcomas (STS) during neoadjuvant radiation therapy (NRT) using multiparametric MRI (mpMRI).
Methods or Background: In this prospective study, 25 STS patients (intermediate or high grade) underwent 3T mpMRI at baseline, mid-treatment (week 3), and post-NRT (week 5). Quantitative parameters were extracted from T2-weighted, diffusion-weighted imaging (DWI), intravoxel incoherent motion (IVIM), and dynamic contrast-enhanced (DCE) sequences following 3D tumor segmentation. Principal component analysis (PCA) and unsupervised k-means clustering identified response patterns based on percentage changes from baseline to post-treatment. Representative whole-mount histology sections were obtained using spatially co-registered in vivo and ex vivo MRI, with necrosis quantified from three representative slices per tumor. Clusters were compared using Mann-Whitney U tests and validated against histopathological necrosis.
Results or Findings: Among 25 patients (mean age 68±14 years, BMI 25.6±5.5 kg/m², 35% female), k-means clustering identified two distinct response patterns. Cluster 1 (n=13) demonstrated variable changes in vascularity/permeability (Ktrans: +32.8±83.5%, Kep: +23.2±27.5%) and perfusion (f: +13.7±35.8%, fD: +10.0±40.1%) with stable cellularity (ADC: +6.2±7.7%, D: +8.4±10.0%). In contrast, cluster 2 (n=12) showed reduced vascularity/permeability (Ktrans: -42.9±30.3%, Kep: -29.8±25.0%, both p<0.01 vs. cluster 1) and perfusion (f: -11.4±21.0%, fD: -24.1±25.0%, both p=0.024) with reduced cellularity (ADC: +31.7±11.6%, D: +33.9±14.0%, both p<0.001). Cluster 2 demonstrated significantly higher post-surgical necrosis (60±30% vs. 30±30%, p=0.021), indicating superior treatment response. Notably, volumetric changes did not differ significantly between clusters (p=0.568), despite distinct functional imaging patterns and different necrosis levels.
Conclusion: Unsupervised clustering identified two distinct functional response patterns during NRT in STS, independent of volumetric changes. These patterns corresponded to different levels of histopathological necrosis, suggesting that mpMRI could non-invasively identify treatment responders during neoadjuvant therapy.
Limitations: Small sample size.
Funding for this study: None.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: University of Freiburg.