Purpose: To explore the predictive role of shape and texture features extracted from MR images for differentiating lipoma from atypical lipomatous tumour (ALT).
Methods and Materials: With Institutional Review Board approval, the PACS databases of multiple institutions were searched for the terms “lipoma”, “liposarcoma” and ”lipomatous”. 3251 subjects were found and screened for pathologically proven surgically resected lipoma or ALT and pre-operative MRI study with axial T1 sequence for lesion segmentation. LifeX software was used to extract 3 shape features (volume, sphericity and compacity) and 32 texture indices from 4 texture matrices: grey-level co-occurrence matrix (GLCM), grey-level run length matrix (GLRLM), grey-level zone length matrix (GLZLM) and neighborhood grey-level different matrix (NGLDM). Receiver-operating characteristic (ROC) analysis was used to calculate the area under the curve (AUC) and 95% confidence interval (CI) for each feature.
Results: The study included 97 patients (median age 58 years [range: 32-92], 55 women), 25 ALTs and 72 lipomas. Several shape and texture features were significantly predictive of pathological diagnosis (p<0.001). AUC values were as follows: compacity 0.83 (0.75-0.91), sphericity 0.63 (0.51 - 0.76), NGLDM busyness 0.77 (0.66-0.88), GLZLM grey-level non-uniformity (GLNU) 0.77 (0.66-0.88), run-length non-uniformity (RLNU) 0.75 (0.63-0.86), GLRLM GLNU 0.74 (0.63-0.85) and RLNU 0.74 (0.63-0.85). In multivariate analysis, compacity, busyness and GLRLM RLNU were independent predictors for pathology with odds ratio (95%CI) of 2.2 (1.5-3.2; p<0.001), 1.7 (1.1-2.7; p=0.018) and 1.0 (1.0-1.0; p=0.025), respectively.
Conclusion: Radiomic features extracted from MR images, specifically compacity and busyness, could help differentiating lipoma from ALT.
Purpose: To provide MRI-based quantitative decision support when to biopsy subfascial lipomatous tumours.
Methods and Materials: MRI of 100 histopathologically proven subfascial lipomas (n=75) and atypical lipomatous tumours/well-differentiated liposarcomas (ALT/WDL) (n=25) were reviewed retrospectively. Exclusion criteria were non-lipomatous soft tissue components, suprafascial and intraabdominal location. Ratios derived from region of interest-based signal intensity (SI) measurements of tumour and adjacent fat on T2 fat-saturated (FS) images were calculated (=mean SItumour/mean SIadjacent-fat). Univariate regression analyses were applied and a p value <0.01 deemed significant. The discriminatory ability was assessed by ROC curve analyses. Interreader agreement was evaluated by calculation of intraclass correlation coefficients (ICC). Tumour size, location, septation and patient demographics were noted.
Results: T2 FS ratios performed well discriminating lipoma from ALT/WDL (p<0.001, area under the ROC curve (AUC) =0.89, cutoff =1.22, specificity =88%, sensitivity =80%). A T2 FS ratio >1.37 indicated ALT/WDL with 95% specificity and 64% sensitivity. Interreader agreement was excellent (ICC =0.924). ALT/WDL was significantly larger than lipoma (p <0.001, AUC=0.84, cutoff =11 cm, 92% specificity, 72% sensitivity). Combination of a T2 FS ratio <1.22 and size <11 cm as decision support would have spared 61% of lipoma patients from biopsy (combined specificity=99%, sensitivity=58%). Qualitative imaging parameters and patient demographics did not differ significantly (each p>0.022).
Conclusion: Lipomatous subfascial tumours with a T2 FS ratio <1.22 and size <11 cm are most likely benign and may not need biopsy. Tumours with a T2 FS ratio >1.37 should be biopsied even when <11 cm.
Purpose: Our aim was to assess the diagnostic value of shear wave elastography (SWE) for the differentiation of benign and malignant soft tissue tumours.
Methods and Materials: Institutional ethics committee approval and informed consent from each participant were obtained for this prospective study. Fifty patients with soft tissue tumours (28 men and 22 women; mean age ± SD: 44.02 ± 19.13 years) were examined with SWE using virtual touch tissue imaging and quantification (VTIQ) method between May 2016 and June 2018. Five region of interest boxes were randomly placed on the lesion and tissue stiffness was measured as shear wave velocity (SWV) in meter per second (m/s). Mean SWV, maximum SWV, minimum SWV, and the difference between maximum and minimum SWV were used to compare the stiffness of the benign and malignant lesions.
Results: Thirty-one benign and 19 malignant soft tissue tumours were diagnosed. Mean SWV, maximum SWV, minimum SWV and the difference between maximum and minimum SWV values for malignant lesions were 3.00 m/s, 3.54 m/s, 2.65 m/s and 0.89 m/s, respectively. Mean SWV, maximum SWV, minimum SWV and the difference between maximum and minimum SWV values for benign lesions were 3.29 m/s, 3.91 m/s, 2.85 m/s and 0.89 m/s, respectively. No statistically significant difference was found between benign and malignant lesions for the measured SWV values (p = 0.353, p = 0.418, p = 0.327, and p = 0.385, respectively).
Conclusion: SWE did not contribute to the differentiation of benign and malignant soft tissue tumours.
Purpose: To study the imaging characteristic signs of aggressive behaving synovial sarcoma (SS) in relation to the pathologic types and grading to look for statistical significance using a radiologic-pathologic correlation.
Methods and Materials: After institutional review board approval, a monocentric retrospective study was undertaken on 15 (9 Males & 6 Females; mean age of 35) histopathologically-proven SS with MRI images between January 2012 and December 2017 at King Fahad Medical City in Riyadh. Data was collected on patient's demographics and the imaging features on multiple imaging modalities, in particular MRIs which were reviewed by 2 musculoskeletal radiologists. Pathology slides were reviewed by a musculoskeletal pathologist. Data were analysed descriptively and statistically.
Results: Among multiple aggressive imaging findings of SS studied (haemorrhagic pockets, peri-lesional oedema, muscle and bone invasion, neurovascular bundle invasion, triple sign, bowl of grapes sign), the only statistically significant imaging finding that correlates with the histopathological high grade (grade 3) of SS is the neurovascular bundle invasion (P <0.05). No statistically significant correlation (P value varies between 0.135 and 0.853) between other imaging characteristics and histopathologic types of SS was found. No statically significant correlation between tumour size and its probability to metastasize (P 0.516). However, large tumour size (>5cm), intralesional haemorrhage, bone invasion are frequently encountered imaging findings in high-grade and metastatic SS. Most common metastatic site is pleuro-pulmonary (81%).
Conclusion: Neuro-vascular bundle invasion could help predict patient histopathological grade and thus aid in disease prognostication.
Purpose: Differentiation between osteosarcomas and other common bone tumours such as chondrosarcoma, Ewing's sarcoma and chordoma can be a significant radiologic-histologic challenge because osteosarcomas represent a highly heterogeneous tumour group. The clinical relevance is that a distinction is important since, depending on the diagnosis, there are different therapeutic approaches with regard to (neo-)adjuvant chemo-/radiotherapy. This study evaluates the performance of MRI-based 3D texture analysis (TA) in comparison to CT-based TA for the differentiation of osteosarcoma.
Methods and Materials: 132 patients were retrospectively evaluated: 64 with contrast-enhanced CT, 28 with non-contrast CT and 40 with contrast-enhanced MRI. All patients had a histopathological diagnosis based on the surgical specimen. TA was performed on CT/MR images obtained for routine purposes using the research software MINT, a spinoff from the DKFZ, which allows and post-process 3D measurements. Kruskal-Wallis test and Dunn-Bonferroni post hoc test were performed to identify the most discriminative texture features (kurtosis, entropy, skewness, MPP, UPP, uniformity).
Results: In MRI, the TA skewness in the T2-weighted sequences significantly differentiated osteosarcoma from chondrosarcoma (p = 0.002), Ewing's sarcoma (p = 0.037), and chordoma (p = 0.006). In contrast-enhanced CT, the TA entropy and MPP could significantly discriminate between osteosarcoma and chondrosarcoma (p = 0.002 and p = 0.001), Ewing's sarcoma (p <0.0005 and p <0.0005) and chordoma (p = 0.038 and p = 0.002). In non-contrast CT, osteosarcoma could not be differentiated.
Conclusion: Tumour heterogeneity quantified by MRI- and contrast-enhanced CT-based 3D texture analysis has the potential to differentiate osteosarcoma from chondrosarcoma, Ewing's sarcoma and chordoma.
Purpose: To investigate the diagnostic value of MR imaging with metal artefact reduction sequences in local recurrence of malignant bone tumour after joint replacement.
Methods and Materials: 94 cases who were pathologically diagnosed with malignant bone tumour underwent clinical and imaging follow-up after joint replacement. All cases received MR scans with metal artefact reducing syngo WARP sequences and FSE sequences. The sensitivity, specificity, and consistency rate of syngo WARP sequences and FSE sequences were compared for diagnosing local recurrence of malignant bone tumours. The Kappa test was used to assess the consistency of syngo WARP and FSE sequences with pathology in diagnosing recurrence respectively. ICC evaluated the consistency of MR images with pathology in measuring the volume of recurrent tumour.
Results: 35 of the 94 cases were pathologically or clinically diagnosed as local recurrence after joint replacement. Local recurrence mainly presented as soft tissue masses and bone destruction on MR images. The sensitivity, specificity, coincidence rate and Kappa value of syngo WARP sequences in the diagnosis of local recurrence of malignant bone tumours were 94.3%, 94.9%, 94.7%, and 0.887, which were higher than FSE sequences. Compared with pathology, the ICC of MR images in measuring recurrent tumour volume was 0.961.
Conclusion: The syngo WARP sequences can significantly reduce metal artefacts in scanning area, and diagnose local recurrence. It also has a good consistency with pathology in evaluating the lesion volume, thus it’s recommended for the standard evaluation of postoperative bone tumour.
Purpose: To assess the response of primary osseous giant cell tumor to denosumab with FDG PET-CT.
Methods and Materials: We present retrospectively evaluated 90 patients of giant cell tumor treated with denosumab in the setting of recurrence, high surgical risk, high grade lesions and monotherapy. Although most of the patients required surgery (extended curettage, en bloc resection and prosthetic reconstruction, arthrodesis, etc.), in two patients with recurrent GCT of the metacarpal and two patients with GCT of pelvis, denosumab was used as monotherapy. Patients were followed with PET-CT with a mean of 2 years (Range: 1-5 years).
Results: PET-CT evaluation showed reduced activity in 81% of the patients and 63% of the Campanacci grade III patients on plain radiographs showed down grading to grade II following denosumab therapy. 77% of the patients showed histological evidence of interval reduction in grade of tumor.
Conclusion: Response to denosumab is noted in the form decrease in metabolic activity, ossification of the lesional soft tissue component and lytic bone. The reduction in metabolic activity was more prominent than regression in size of the primary lesion. Thus PET-CT adds to imaging response assessment in osseous GCT.
Purpose: This study aimed to compare image quality and radiation dose of a combined automatic tube current modulation (ATCM) technique with those of a fixed tube current technique (FTC) in whole-body low-dose multidetector CT (WBLDCT), in the diagnosis of Multiple Myeloma.
Methods and Materials: 28 consecutive patients (15 men, mean age 63,8±11,3 years) underwent unenhanced WBLDCT for evaluation of Multiple Myeloma using a third-generation dual-source CT scanner (192-slice) with dedicated tin filter, from August-2017 to September-2018. All acquisition parameters were identical in both techniques (100 kV, tin filtration) except for tube current: 18 patients were performed with ATCM and 10 patients with FTC (200 mAs). We recorded objective image noise in muscles at different anatomic levels and in the liver parenchyma, radiation dose (DLP, mSv) and body mass index (BMI, kg/m2). The Mann-Whitney U test was used for statistical analysis.
Results: No statistical differences were observed in the BMI estimated in both groups (p=0.94). Signal-to-noise ratio was higher in the ATCM group at scapular and pelvic girdles compared to FTC, with statistically significant differences (p=0.014; p<0.001 respectively). Objective image noise values at the upper neck and in the abdominal region found no significant differences between both techniques (p=0.191 and p=0.408). The effective radiation dose with the ATCM technique was 2.91 vs 1.49 with the FTC (p<0.001).
Conclusion: WBLDCT with ATCM technique achieves diagnostic image quality with better signal-to-noise ratio in some compromised areas as the scapular and pelvic girdles, while maintaining acceptable low radiation dose levels.
Purpose: Collagen-bound water (CBW) and pore water (PW) components of cortical bone reflect bone porosity and organic matrix density. Our study evaluates the longitudinal relaxation time of those two compartments as biomarkers for radiation-induced bone damage.
Methods and Materials: ZTE images of mouse femur were acquired at 4.7T with 14 different inversion times (0-2,600 ms) before radiation and 3 weeks/3months after radiation at 20 Gy by the xrad system 320. Longitudinal relaxation time of the two compartments CBW, PW and CBW fraction (cbwf) for each timepoint was computed by a bi-exponential fitting. µCT imaging were performed at the same timepoints to validate bone porosity.
Results: Baseline measurements displayed a mean T1cbw of 245 ± 110ms, a mean T1pw of 2076 ± 623ms and a cbwf of 6.9 ± 3.4%. 3 weeks after radiation, both T1 relaxation time (CBW 269 ± 89ms, p value = 0.12; PW 2204 ± 575ms, p value=0.09) and cbwf (6.4 ± 3.7%) stayed stabile. After 3 months, T1 relaxation time significantly increases (T1cbw 432 ± 89ms, T1pw 3423 ± 543ms) towards the baseline. Contrarily, cbwf decreased significantly (3.7 ± 2.1%). µCT-derived bone porosity parameter showed good correlation with both T1 relaxation time (pw: r=0.68, cbw:r=0.63) and cbwf (r=0.77).
Conclusion: Longitudinal relaxation time of CBW and PW components are sensitive towards bone porosity changes after radiation and may be useful as non-invasive tool for radiation-induced bone damage.
Purpose: To evaluate the MRI features associated with survival after preoperative radiotherapy in soft tissue sarcoma.
Methods and Materials: A review of a prospectively collected database was performed for patients treated for extremity STS. Pre- and post-radiotherapy MRI were reviewed by two fellowship-trained musculoskeletal radiologists. Maximal tumour diameter and volumes were measured on T2-weighted sequences and the change in high T2 signal was estimated as a percentage. High T2 signal was defined as signal intensity similar to an adjacent vessel or fluid. Tumours were divided into partial response (PR), stable disease (SD) or progressive disease (PD). Local recurrence-free survival (LRFS), metastasis-free survival (MFS) and overall survival (OS) were examined using the Kaplan-Meier method and compared with log-rank test.
Results: 309 studies were available to review. There was an increase in tumour volume after RT with a mean volume increase of 132cm3 (3190cm3-6087cm3). 117 tumours showed a partial response, 86 tumours were classified as stable disease and 106 tumours as progressive disease. Of the 166 patients assessed for T2 change 106 showed no change, 26 showed a reduction in T2 signal and 34 showed an increase in T2 signal. PD by RECIST had a worse MFS (p<0.001) and OS (p<0.001). Tumours that demonstrated an increase in T2 signal post-radiotherapy had an increased MFS (p=0.035) but no difference in LRFS or OS (p=0.079).
Conclusion: PD after radiotherapy is associated with worse MFS and OS. Increase in percentage T2 signal is associated with increase MFS but no change in OS.