SS 716 - Imaging nervous system and musculoskeletal tumours
SS 716 - Imaging nervous system and musculoskeletal tumoursThursday, March 2, 14:00 - 15:30 Room: F1 Session Type: Scientific Session Topics: Imaging Methods, Oncologic Imaging Moderators: F. A. Gallagher (Cambridge/UK), B. M. Schaarschmidt (Düsseldorf/DE) Add session to my schedule In your schedule (remove)
Purpose: Radiomics has recently gained much attention as a tool that promises to extract the maximal amount of information from standard-of-care images. The present study investigates whether a radiomics analysis can predict overall survival from T1-weighted contrast-enhanced baseline MR images in a uniformly treated cohort of glioblastoma patients.
Methods and Materials: This retrospective study was approved by the institutional review board and informed consent was waived. MR images from 66 patients with newly diagnosed GBM from a previous prospective study were analyzed. Tumor segmentation was performed manually on contrast-enhanced 3D T1-weighted images. Using these segmentations, n=100 quantitative image features characterizing tumor shape, signal intensity and texture were calculated by automated high-throughput analysis and prognostic models based on random survival forests were trained on the data. For each patient, mortality was predicted using leave-one-out cross validation and average prediction error was recorded. Association of predicted mortalities with overall survival was assessed using Kaplan-Meier analysis and univariate proportional hazard models.
Results: The final prognostic model predicted a median mortality of 32.3% (range 7.1% - 61%) with a prediction error of 34.5%. Kaplan-Mayer analysis clearly distinguished two patient groups with high and low predicted mortalities (p=8.0e-4). Low predicted mortality was found to be a favorable prognostic factor for overall survival in a univariate Cox proportional hazard model (p=1.7e-04).
Conclusion: Baseline MR imaging in GBM patients contains strong prognostic information, which become accessible by radiomics analysis using random survival forests.
Purpose: Bevacizumab (BVZ) is a monoclonal antibody directed against vascular endothelial growth factor (VEGF) and has been suspected to increase the incident of ischemic stroke (IS) and intracranial hemorrhage (ICH) in GBM patients.
Methods and Materials: In this study 364 MRI of 82 GBM patients were eligible for analyses. All patients were treated with basic treatment. Out of the 82 patients, 40 were treated with BVZ in addition to basic treatment. (BVZ-group) The cohorts matched in age and gender. Vascular risk factors were analyzed retrospectively. The KPS was captured at the beginning and end of the observation, and before and after the event. Chi-square test was used to evaluate the level of significance.
Results: In seven (8 % / 7/82) out of 82 patients a vascular pathology was detected with MRI. 4 (4.8 % / 4/82) of them revealed an IS, while ICH was detected in 3 (3.6 % / 3/82). 4 of them had been treated with BVZ- with a total of 3 IS (7.5% 3/40), and 1 ICH (2.5% / 1/40). Three of them were part of the control-group, with a total of 1 IS (2.3% / 1/42) and 2 ICH (4.7% / 2/42).
Conclusion: The incidence of vascular events did not differ significantly between patients receiving BVZ and the control group. Thus, BVZ treatment does not seem to be associated with an elevated risk for vascular events in GBM patients.
Purpose: Comparison of supratentorial WHO grade II gliomas spectroscopic image.
Methods and Materials: Group of 45 patients with supratentorial WHO grade II gliomas: 26/45 (58%) Astrocytoma fibrillare, 10/45 (22%) Astrocytoma fibrillare partim gemistocyticum, 9/45 (20%) Oligodendroglioma. All patient had their 1HMRS study in Cancer Center Gliwice before surgery. Studies were performed on 1.5 T or 3T scanners. Single voxel PRESS method was used with long (135 ms), and also in 43 patients short (30 ms) echo time (TE). Reference spectra were obtained from healthy tissue of contralateral hemisphere (ref). Spectra were analyzed with LCModel. Following metabolite ratios were calculated for TE135ms: Cho/Cr, NAA/Cr, Cho/NAA, Gly/Cr in lesion and reference and Cho lesion/ Cho ref, NAA lesion/NAA ref and Gly lesion/Gly ref. In TE30ms: mI/Cr, Glx/Cr in lesion and reference and mI lesion/mI ref, Glx lesion/Glx ref. Metabolite integrals between lesion and healthy tissues were compared with statistical methods using STATISTICA 10 software.
Results: Maximal and minimal values in groups: Astrocytoma fibrillare: Cho/Cr 1.17-3.96, NAA/Cr 0.31-2.19, Cho/NAA 0.69-6.21, Gly/Cr 0.04-0.72, mI/Cr 0.19-1.36, Glx/Cr 0.03-5.13, mIlesion/ mIref 0.55-3.09, Glxlesion/Glxref 0.01-2.00, Cholesion /Choref 0,27-3,60, NAAlesion/NAAref 0,06-0,69. Astrocytoma fibrillare partim gemistocyticum: Cho/Cr 1.09-6.40, NAA/Cr 0.50-1.33, Cho/NAA 1.40-5.07, Gly/Cr 0.21-0.98, mI/Cr 0.37-0.92, Glx/Cr 0.29-4.16, mIlesion/mIref 0.70-1.95, Glxlesion/Glxref 0.27-2.00, Cholesion/Choref 0.87-2.80, NAAlesion/NAAref 0.01-0.40.
Oligodendroglioma: Cho/Cr 1.01-7.23, NAA/Cr 0.50-1.54, Cho/NAA 0.68-4.07, Gly/Cr 0.50-1.11, mI/Cr 0.35-0.74, Glx/Cr 0.81-3.35, mIlesion/mIref 0.88-2.34, Glxlesion/Glxref 0.29-1.10, Cholesion /Choref 0.93-5.08, NAAlesion /NAAref 0.30-0.68.
Conclusion: No significant differences in 1HMRS were found between groups: Astrocytoma fibrillare, Astrocytoma fibrillare partim gemistocyticum and Oligodendroglioma.
Purpose: To asses by MRI the dynamic changes of deep brain large metastases after hypofractionation stereotactic radiation therapy.
Methods and Materials: In retrospective analysis, 146 patients (mean age 55 y.o.) with 215 brain metastases, treated by hypofractioning radiotherapy were included (dose 8-10 Gy, mean metastasis volume 1.5 cm3). 1.5 and 3.0T scanners were used with Т13D sequences 1mm before and after contrast enhancing, Т2 WI tra 2mm, flair tra 1-3mm, T2 WI cor 2mm, DWI. In 21 patients SWI was added. Tumour volume control was performed by GammaPlan 10.1 station. Patients were assessed before treatment, after 1 month, then, every 3 months.
Results: We have detected 6 radiologic patterns: tumour dimensions changes (68%), structural necrosis (28%), metastasis contrasting decreasing (64%), contour changes (83%), perifocal swelling area reduction (97%), intratumour haemorrhage (14%). The most variable were dimensions and volume changes: volume reduction or stabilisation, volume increasing by necrosis or disease progression, volume increasing on 1st control, then decreasing on 2nd control (necrosis), then increasing by haemorrhage.
Conclusion: Brain metastases reaction after radiotherapy characterised by heterogeneity. Imaging results (volume and contour changes, haemorrhage) are non-specific and may be signs of local disease progression or postradiation reactions. To clarify these findings, well-designed prospective multi-centre clinical trials are needed in future.
Purpose: To assess gliomas reaction after different types of radiation therapy.
Methods and Materials: In prospective trial 101 patients with brain gliomas were included. Patients were stratified into 2 groups: group A with grade I-II (n=58), group B with grade III-IV (n=43). 1.5 and 3.0T scanners were used with T13D sequences before and after contrast enhancing, Т2 WI tra 2mm, flair tra 1-3mm, T2 WI cor 2mm, DWI. Tumour volume control was performed by GammaPlan 10.1 station. First assessment was before treatment, then after 1 month, next - every 3 months. For disease progression evaluation, stereotactic brain biopsy, PET-CT and CT-perfusion were used.
Results: We have detected 4 postradiation patterns: tumour dimension changes (reduction in 23% in group A and 9% in group B, stabilisation in 46 and 18%, respectively, increasing and then decreasing due to necrosis or haemorrhage - 15 and 7%, respectively, disease progression - 16 and 66%, respectively), contrasting decreasing (17% in group A and 8% in group B), contour changes (54 and 82%, respectively), intratumour haemorrhage (5% in group A and 12% in group B).
Conclusion: Postradiation reaction in patients after gliomas radiotherapy characterised by heterogeneity. Grade III-IV is the serious prognostic risk factor for disease recurrence. MRI-based only disease assessment potentially may be nonspecific and the complex diagnostic tools (included PET-CT and brain biopsy) should be used.
Purpose: To compare Response Evaluation Criteria in Solid Tumours (RECIST 1.1) to volume modifications, in order to evaluate the response in patients with sacral chordoma, not suitable for surgery, treated with carbon ions radiotherapy (CIRT) alone. Secondarily, to detect if baseline ADC values could predict response to treatment.
Methods and Materials: 39 patients were retrospectively studied, considering baseline (03/2013 - 04/2016) and follow-up multiparametric MR exams (mean follow-up 18 months, range 3-37), for a total of 195 examinations. The exams were performed with a 3T MR scanner (Siemens Magnetom Verio). For each exam, lesion maximum diameter and volume were obtained from T2w axial images. Lesion segmentations were consequently aligned to the ADC maps calculated from DWI sequences. The baseline ADC values within the lesion volume were analyzed.
Results: Considering maximum diameter changes between baseline MRIs and the last available follow-ups, according to RECIST 1.1, patients were classified: Partial Response (PR) 0%, Stable Disease (SD) 89.7% and Progressive Disease (PD) 10.3%. Considering lesion volume changes, applying the same response criteria, PR were 53.8%, SD 35.9% and PD 10.3%. The assessment of baseline examinations ADC maps, using Wilcoxon test, demonstrated significantly higher median ADC values of PD vs both PR and SD (p=0.0306 and p=0.0143 respectively).
Conclusion: Our study demonstrates that lesion volume measurement may be more accurate than maximum diameter to better stratify the response of sacral chordoma treated with CIRT. Preliminary results suggest that baseline ADC values could be predictive of response to CIRT, particularly to detect potential non responders.
Purpose: To investigate wether elevated glucose metabolism in neurofibroma determined by [18F]-FDG-PET is correlated with cell density in MRI as expressed through the apparent diffusion coefficient.
Methods and Materials: Maximum and mean standardised uptake values (SUVmax, SUVmean) on [18F]-FDG-PET/CT and [18F]-FDG-PET/MR were compared, and correlated with minimum and mean apparent diffusion coefficients (ADCmin, ADCmean).
Results: 12 (6 male/6 female, mean age was 16,2 ± 5,2 years) patients were prospectively included and analysed on a per-lesion (n=39) basis. The SUVmean of examined neurofibroma loci showed a moderate negative correlation of the ADC mean (r:-.441) and ADCmin (r:-.477), which both proved to be statistically significant (p=.005 and p=.002); the SUVmax of respective lesions however, showed a weaker negative correlation for ADCmean (r: -.311) and ADCmin (r: -.300) and did not reach statistical significance (p=.054 and p=.057).
Conclusion: Our data suggest that the ADCmean and min could possibly supersede the SUVmean as a potential determinant of malignant transformation in neurofibromatosis 1.
Purpose: Differentiation of low-grade chondrosarcoma from enchondroma is both a radiologic and histologic challenge. Benign lesions do not require surgery, whereas curative surgical resection is mandatory in treatment for chondrosarcoma. Texture analysis is a promising tool in oncologic imaging. This study evaluates the performance and accuracy of MRI-based 3D texture analysis for the discrimination of G1-chondrosarcoma from enchondroma.
Methods and Materials: 22 patients were retrospectively evaluated: 11 patients with chondrosarcoma confirmed by histopathological diagnosis; 11 patients with enchondroma confirmed by histopathology or with a follow-up greater 5 years without changes in radiological features. Texture analysis was performed using the commercially available software mint Lesion by Mint Medical, a spinoff from the German Cancer Research Center (DKFZ). It allows 2D-and volumetric measurements and multiparametric texture analysis. Mann-Whitney U test and ROC-analysis were performed to identify the most discriminative texture features (kurtosis, entropy, skewness, MPP, uniformity), and by Youden's index optimal cut-off-values were selected.
Results: Significant differences were found in 6 out of 20 texture parameters (p<0.05). The area-under-the-ROC-curve for this 6 parameters to discriminate chondrosarcoma from enchondroma were 0.876 and 0.826 for kurtosis and skewness in ceT1-fs, respectively; in non-contrast T1 the values were 0.851, 0.793 and 0.822 for entropy, MPP and uniformity, respectively; in STIR it was 0.802 for MPP. Highest discriminatory power had kurtosis in ceT1-fs with an optimal cut-off ≥3.15 (82% sensitivity, 91% specificity, accuracy 86%).
Conclusion: MRI-based 3D texture analysis has the potential to distinguish low-grade chondrosarcoma from enchondroma by kurtosis in ceT1-fs having the highest power of discrimination.
Purpose: The aim of the study is to investigate the role of MRI using conventional and DWI sequences in initial evaluation and follow-up of pathologically proven cases of Ewing’s sarcoma family of tumours.
Methods and Materials: Our study included 35 patients with pathologically proven Ewing’s sarcoma family of tumours. We recorded the conventional and diffusion-weighted MR features of the examined lesions at initial presentation and after neo-adjuvant treatment including tumour location, size, signal characteristics, enhancement and breaking down. The diffusion images with ADC of the minimum and mean values were obtained.
Results: This study included (35) patients, their ages ranges from 2 to 41 years (mean age 17.2 years). Intermediate signal intensity on T1WI was noted in 77.1% of Ewing’s sarcoma cases while high signal intensity on T2WI was noted in 62.9%. Our cases showed homogeneous contrast pattern and avid intensity contrast uptake in 45.7% and 74.3% respectively. Intra-tumoural breaking down was noted in 25.7% of the patients at initial assessment as compared to 65.7 % in patient post neo-adjuvant therapy with statistically significant P=0.005. The initial mean ADC of the Ewing’s sarcoma family of tumours (0.68x10-3 mm²/sec) was significantly different from that after neo-adjuvant treatment (1.60x10-3 mm²/sec) with mean difference (-0.9125) and P<0.001.
Conclusion: MR imaging is the method of choice in the initial evaluation and follow-up of Ewing’s sarcoma family of tumours. Diffusion-weighted MR imaging has the potential of being an additional non-invasive tool offering means for a more detailed analysis of these lesions and assessment of their therapeutic response.
Purpose: MRI biomarkers role in prognosis of newly diagnosed glioblastoma remains unclear. We retrospectively determined the usefulness of dynamic susceptibility contrast (DSC), permeability maps, diffusion parameters, extensive battery of qualitative findings for contrast-enhancing lesion (CEL) and surrounding non-CEL in predicting survival.
Methods and Materials: Before treatment, 72 consecutive patients (45 men; mean age, 64 years) with histologically proven glioblastoma underwent 1.5T MRI (anatomical, first-pass DSC, DWI and post-contrast T1-weighted sequences). Perfusion maps were computed with the Bayesian approach. Volumes of interest for cerebral blood (CB) volume ratio, CB flow ratio, mean transit time (MTT), time-to-peak, delay time (DT), permeability constant (k2), and apparent diffusion coefficient (ADC) in CEL, non-CEL, and contralateral tissue using Olea Sphere V.3.0 software (Olea Medical, La Ciotat, France) we determined. We evaluated 26 VASARI descriptors. Patients were classified by survival:<1year and >1year. Surgery, radiotherapy and chemotherapy was considered complete treatment.
Results: Forty-nine patients (68%) survived<1year. Thirty-nine (54.16%) underwent complete treatment. Survival groups differed in age (62.21±12.53 vs 50.63±15.1years;P<0.001), ADC-CEL (0.98±0.97 vs 0.78±0.05mm2/s;P=0.032), delay-CEL (-0.41±0.82 vs 0.07±0.54sec;P=0.043), DT-nonCEL (-0.26±0.58 vs 0.18±0.47sec;P=0.01) and treatment (P<0.001), for <1 and >1year survival, respectively. In univariate analysis, complete treatment best predicted survival at 1 year (AUC=0.774, 60.6% sensitivity, 94.3% specificity, 95.6% positive predictive value, 54.1% negative predictive value). However, DT-CEL and age yielded the best combined prediction of survival (AUC=0.859, 86.7% sensitivity, 73.3% specificity, 86.7% positive predictive value, 73.3% negative predictive value).
Conclusion: Beyond well-known survival factors, our data suggests perfusion parameter DT might help predict survival in newly diagnosed glioblastoma.