RPS 2410 - Imaging of musculoskeletal tumours

Author Block: J. Li¹, J. Liu²; ¹Fujian/CN, ²Xiamen/CN
Purpose: To evaluate the value of dual energy computed tomography (DECT) quantitative parameters for the differentiation of small osteolytic metastases (SOMs) and SOM-mimics such as osteopenia, osteoporosis, and Schmorl's nodule.
Methods or Background: Fat(HAP), fat(calcium), hydroxyapatite(fat), and calcium(fat) densities [Dfat(HAP), Dfat(calcium), DHAP(fat), and Dcalcium(fat)], as well as CT value were collected. Comparisons were made using the independent sample T test. Diagnostic performance was assessed in terms of area under the receiver operating characteristic curve (AUC). The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of each parameter was assessed as well.
Results or Findings: A total of 106 patients were included, of whom 24 had SOMs (lesion, n = 48), while 82 had SOM-mimics (lesion, n = 202). SOMs associated with significantly higher CT value, Dfat(calcium), and Dfat(HAP) compared to SOM-mimics (P < .001). The AUCs were 0.674, 0.879, and 0.887, respectively. The sensitivity, specificity, PPV, NPV, and accuracy of fat(calcium) were 77.1%, 85.1%, 55.2%, 94.0%, 83.6%, respectively; while those for Dfat(HAP) were 83.3%, 80.7%, 50.6%, 95.3%, 81.2%, respectively. The optimal diagnostic cutoffs for Dfat(calcium) and Dfat(HAP) were ≥ 1000.0 mg/cm3 and ≥ 966.9 mg/cm3, respectively, which achieved consistent diagnostic results among 89.6% lesions (n = 224). The combined use of Dfat(HAP) and Dfat(calcium) achieved significantly better diagnostic performance, with AUC, sensitivity, specificity, PPV, NPV, and accuracy of 0.910, 82.2%, 87.2%, 61.7%, 95.1%, and 86.2%, respectively.
Conclusion: Dfat(calcium) and Dfat(HAP) on DECT carry the potential as parameters for the discrimination of SOMs from SOM-mimics
Limitations: This was a retrospective study with a relatively small sample size. In addition, the focus on thoracolumbar lesions limits the generalizability of our results.
Funding for this study: Natural Science Foundation of Fujian Province, China (grant numbers: 2023J01181)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Fujian Cancer Hospital Ethics Committee (K2023-198-01)

Author Block: S. Lusi, R. Romanelli, A. Marzullo, A. Laffi, A. Santoro, L. Balzarini, M. Francone, A. F. Bertuzzi; Milan/IT
Purpose: This study relies on the use of Artificial Intelligence (AI) to develop and validate a multimodal machine learning method that could provide a prognostic model in AYA patients affected by sarcomas, exploring the clinical, radiomic and pathological features that may be predictive of disease outcome.
Methods or Background: The study is a monocentric retrospective cohort study involving 245 patients with sarcomas. Our preliminary and full results were performed on a smaller cohort of 22 patients with soft tissue sarcoma of the extremities (13 non-AYA and 9 AYA) for whom clinical data (using Excel form), radiomic features (from a pre-treatment MRI) and histopathological features (extracted using a foundation model) were collected. All this data was then used to match similar patient profiles in the two groups using logistic regression propensity scores. Disease-free survival of matched patients was described using a Kaplan-Meier curve.
Results or Findings: Statistical analysis didn't identify any correlation between clinical and radiological features that could explain the differences in prognosis between the two groups, probably due to the small cohort size. However, AI analysis using a Kaplan-Meier curve showed that AYA patients had a worse prognosis than non-AYA patients (p < 0.05), confirming for the first time, to our knowledge, by deep machine learning what is observed in clinical practice.
Conclusion: Despite the limitations of these preliminary results based on a small cohort of patients, our findings provide valuable insights into the differential prognosis that characterises these two groups. AI holds promise for uncovering hidden characteristics, with future research potentially incorporating biological markers to further explore therapeutic targets.
Limitations: The small cohort size of the study limits its statistical power. The retrospective design may introduce selection bias. Future studies in larger populations are needed.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: The study is retrospective.

Author Block: Q. Bui; Paris/FR
Purpose: In long bones, distinguishing between enchondromas and chondrosarcomas before surgery is often challenging and may require invasive biopsy for accurate diagnosis. The purpose of this work was to assess the performance of MRI radiomics-based machine learning in classifying enchondromas and chondrosarcomas in long bones.
Methods or Background: Ninety-eight patients with pathology-proven cartilaginous tumours of long bones were retrospectively included from a tertiary bone tumour centre. The training set consisted of 81 MRI (n = 33 enchondromas; n = 48 chondrosarcomas). The internal test set consisted of 17 MRI (n = 7 enchondromas; n = 10 chondrosarcomas). 3D segmentation was performed on T1-weighted and fat-suppressed T2-weighted MRI images and radiomics features were extracted. Dimensionality reduction was performed based on reproducibility, redundancy and feature importance. Different models were tested, including multiparametric, single sequence and sequential. A Random Forest classifier was tuned on the training set using five-fold cross-validation and tested on the internal test set.
Results or Findings: The Random Forest classifier with the T2 then T1 sequential model, which was the best-performing model, achieved an AUC of 0.943 [0.832 – 1.000] on the internal test set. Its accuracy in correctly classifying enchondromas and chondrosarcomas was 71% (5/7) and 100% (10/10), respectively.
Conclusion: This work shows that MRI radiomics can accurately differentiate between benign and malignant cartilaginous tumours. Although comparisons among various models did not achieve statistical significance, the data suggest that a sequential approach using single sequence models might outperform a multiparametric model. Further studies with larger sample sizes are needed to integrate these findings into clinical practice and improve preoperative diagnosis of cartilaginous tumours within the context of personalised medicine.
Limitations: Limitations of the study were the small sample size and the lack of an external test set.
Funding for this study: The author received a research grant from Societé Française de Radiologie and Assistance Publique - Hôpitaux de Paris.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: This study involved a retrospective analysis of anonymised data collected as part of routine care.

Author Block: M. E. Chevasco Hanze, L. Ponsa Cobas, J. A. Narváez García, D. A. Sandoval Díaz, J. Hernández Gañan, J. C. Sardiñas Barrero; Barcelona/ES
Purpose: Desmoid tumors (DT) are locally aggressive, infiltrative neoplasms with a high risk of local recurrence. Recently, percutaneous cryoablation has emerged as an alternative therapy, though its role as a salvage treatment remains unclear. This study aimed to evaluate disease progression after cryoablation at a 1-year follow-up and develop a predictive model using clinical and radiomic variables.
Methods or Background: We performed a retrospective analysis of patients treated with cryoablation for extra-abdominal DT from January 2018 to September 2023. Pre- and post-cryoablation contrast-enhanced MRIs were reviewed, and disease progression was defined as less than 90% necrosis or non-enhancement at follow-up. Radiomics features were extracted from T2-weighted pre-cryoablation MRIs, and data were filtered based on correlation matrices and statistical tests (T-Student, Mann-Whitney U, Chi-square). Prediction models, including LASSO, Random Forest, XGBoost, SVM, and KNN, were evaluated using ROC analysis and 5-fold cross-validation to determine the optimal approach.
Results or Findings: Twenty-eight patients were included (median age 43; 67% women), with a no disease progression rate of 60.71%, significantly associated with partial response on mRECIST criteria (p = 0.022). The Random Forest model showed the best performance (AUC = 0.77). Key predictive features included tumor diameter, sphericity, major axis length, minimum intensity, kurtosis, interquartile range, and tumor location. Tumors >61 mm, with ellipsoid shape (major axis length >79 mm) and regular form (sphericity <0.7), predicted no disease progression. Similarly, tumors <61 mm with a regular shape (sphericity <0.5) and fibrous matrix (minimum <175) predicted favorable outcomes.
Conclusion: Cryoablation therapy has demonstrated a good therapy for DT treatment. Radiomics shape and first order features have shown their relevance in cryoablation therapy planning as it serve as a patient selection tool.
Limitations: Small sample size.
No split train-test approach.
Funding for this study: No funding was received
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: No intervention, just an observational study

Author Block: C. Chabot; Brussels/BE
Purpose: To assess the diagnostic accuracy, repeatability, and reproducibility of pseudo-CT MRI sequences (ZTE, BB) in detecting osteolytic lesions in MM using WB-CT as the reference standard.
Methods or Background: In this prospective study, consecutive patients were enrolled in our academic hospital. Inclusion criteria were newly diagnosed MM, monoclonal gammopathy of undetermined significance at high risk for MM, or suspicion of progressive MM. Participants underwent ZTE and BB sequences covering the lumbar spine, pelvis, and proximal femurs as part of 3T WB-MRI examinations, as well as clinically indicated 18F-FDG PET/CT examination that included optimized WB-CT. Ten bone regions and two scores (categorical score/semiquantitative score) were assessed by three radiologists on the ZTE, BB, and WB-CT images. The accuracy, repeatability, and reproducibility of categorical scores (Gwet agreement coefficients AC1 and AC2) and differences in semiquantitative scores were assessed at per-sequence, per-region, and per-patient levels.
Results or Findings: 47 participants were included. In experienced readers, BB and ZTE showed 98% accuracy per-patient, while BB accuracy ranged from 83%–100% and ZTE from 74%–94% per-region. Increased false-negative findings in the spine ranging from 17%-23%, according to the lumbar vertebra, was observed using ZTE(P<.013). Regardless of the region (except coxal bones), differences in the BB score minus the ZTE score were positively skewed(P<.021). Repeatability was very good(AC1≥0.87), while reproducibility was at least good(AC2≥0.63).
Conclusion: Both MRI-based ZTE and BB pseudo-CT sequences of the lumbar spine, pelvis, and femurs demonstrated high diagnostic accuracy in detecting osteolytic lesions in MM. Compared with BB, the ZTE sequence yielded more FN findings in the spine.
Limitations: Pseudo-CT sequences were limited to the lumbar spine, pelvis, and femurs; the reference CT required optimization from 18F-FDG PET/CT; focus was on detecting osteolytic lesions, including nonactive ones that may persist post-treatment.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This prospective study was approved by the institutional ethics committee (2020/27JUL/380) and is registered on ClinicalTrials.gov (no. NCT05381077). Written informed consent was obtained from all participants.

Author Block: T. Marth, C. Stern, R. Sutter; Zürich/CH
Purpose: Several studies have demonstrated the potential of tin prefiltration to reduce the radiation dose while maintaining diagnostic quality for musculoskeletal imaging. Still, no study has reported data on the impact of radiation dose reduction on clinical routine scanning.
Methods or Background: Retrospective inclusion of 300 patients who underwent clinically indicated CT scans of the pelvis, knee, or ankle on a single CT scanner (SOMATOM Definition AS, Siemens Healthineers) before January 2020 (without tin filter) and after December 2020 (with tin filter). For each joint, 50 patients with tin prefiltration and 50 patients without tin prefiltration were selected.
Results or Findings: The CTDIvol, DLP, and effective dose were reduced significantly in all tin-prefiltered examinations compared to the conventional, non-tin-prefiltered examinations (p<.001). Tin-prefiltered CT scans had a significantly lower CTDIvol and mean effective dose (all p<.001): CTDIvol was 65% lower in the pelvis, 73% in the knee, and 54% in the ankle. This resulted in a reduction of the effective dose of 61%, 71%, and 60%, respectively. In absolute numbers, the reduction of the median effective dose delivered in a single scan of the pelvis was 2.29 mSv, 0.15 mSv for the knee, and 0.03 mSv in the ankle.
Conclusion: The implementation of tin prefiltration in clinical routine scan protocols significantly reduced the effective radiation dose for unenhanced CT scans of the pelvis (61% reduction, 2.29 mSv), the knee (71% reduction, 0.15 mSv), and the ankle (60% reduction, 0.03 mSv).
Limitations: It would be possible to reduce radiation dose even more by applying deep learning-based denoising algorithms, however this was not yet available in clinical routine during the data acquisition.
Funding for this study: No specific funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: BASEC-ID 2024-01094 Kantonale Ethikkommission Zürich

Author Block: Y. Beer¹, N. Shabshin², L. Copel¹, Y. Kimmel³, R. Ophir¹, Y. S. Schiffenbauer³, S. Tal¹; ¹Zrifin/IL, ²Afula/IL, ³Petach Tikva/IL
Purpose: Digital tomosynthesis (DTS) is a well-established technology that has become the gold standard for breast mammography. In recent years its benefits in musculoskeletal (MSK) imaging have been acknowledged, leading to a rapid increase in its utilization. It improves detection, localization and characterization of subtle fractures. In some patients DTS can alleviate the need for CT with lower complexity. Recently a new technology based on Cold Cathode Xray tube with a multi tube set-up has made this technology more affordable and accessible. The purpose of this study is to evaluate the diagnostic potential of the cold-cathode multi-tube DTS
Methods or Background: The study included 19 patients with suspected fractures who underwent CT and radiographs (XR). Patients were scanned using the cold-cathode DTS . 15 patients had imaging performed with Cast or metal. Images were evaluated by an MSK radiologist and orthopedic surgeon in consensus with CT as the gold standard. Studies were evaluated for presence, location, intraarticular involvement, displaced fragments and incidental lesions. The surgeon was asked whether DTS provided valuable information and whether it increased the confidence of the final diagnosis.
Results or Findings: In 17/19 studies DTS added value to the XR. In 7, DTS found fractures occult in XR. In 3, DTS was able to clear a suspected fracture. in 5, DTS was able to better localize the fracture. In 1, DTS was able to determine fracture age and in 1 study DTS found a sclerotic lesion obscured in XR. In addition, on XR the cast limited evaluation of fine bony details, however there was no such limitation with DTS
Conclusion: Cold cathode DTS provides high quality tomography of anatomies enabling depiction of occult pathologies, localization, characterization and resolution of questionable findings
Limitations: Initial study results
Funding for this study: Study was funded by Nanox-x Imaging Ltd.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Study was approved by the local Ethics committee in the institution and each patient underwent informed consent