Impact of magnetic resonance imaging-related geometric distortion of dose distribution in fractionated stereotactic radiotherapy in patients with brain metastases.

PURPOSE: The geometric distortion related to magnetic resonance (MR) imaging in a diagnostic radiology (MRDR) and radiotherapy (MRRT) setup is evaluated, and the dosimetric impact of MR distortion on fractionated stereotactic radiotherapy (FSRT) in patients with brain metastases is simulated. MATERIALS AND METHODS: An anthropomorphic skull phantom was scanned using a 1.5­T MR scanner, and the magnitude of MR distortion was calculated with (MRDR-DC and MRRT-DC) and without (MRDR-nDC and MRRT-nDC) distortion-correction algorithms. Automated noncoplanar volumetric modulated arc therapy (HyperArc, HA; Varian Medical Systems, Palo Alto, CA, USA) plans were generated for 53 patients with 186 brain metastases. The MR distortion at each gross tumor volume (GTV) was calculated using the distance between the center of the GTV and the MR image isocenter (MIC) and the quadratic regression curve derived from the phantom study (MRRT-DC and MRRT-nDC). Subsequently, the radiation isocenter of the HA plans was shifted according to the MR distortion at each GTV (HADC and HAnDC). RESULTS: The median MR distortions were approximately 0.1 mm when the distance from the MIC was < 30 mm, whereas the median distortion varied widely when the distance was > 60 mm (0.23, 0.47, 0.37, and 0.57 mm in MRDR-DC, MRDR-nDC, MRRT-DC, and MRRT-nDC, respectively). The dose to the 98% of the GTV volume (D98%) decreased as the distance from the MIC increased. In the HADC plans, the relative dose difference of D98% was less than 5% when the GTV was located within 70 mm from the MIC, whereas the underdose of GTV exceeded 5% when it was 48 mm (-26.5% at maximum) away from the MIC in the HAnDC plans. CONCLUSION: Use of a distortion-correction algorithm in the studied MR diagnoses is essential, and the dosimetric impact of MR distortion is not negligible, particularly for tumors located far away from the MIC.

Neurite orientation dispersion and density imaging and diffusion tensor imaging to facilitate distinction between infiltrating tumors and edemas in glioblastoma.

BACKGROUND: Glioblastomas are highly infiltrative tumors, and differentiating between non-enhancing tumors (NETs) and vasogenic edema (Edemas) occurring in the non-enhancing T2-weighted hyperintense area is challenging. Here, we differentiated between NETs and Edemas in glioblastomas using neurite orientation dispersion and density imaging (NODDI) and diffusion tensor imaging (DTI). MATERIALS AND METHODS: Data were collected retrospectively from 21 patients with primary glioblastomas, three with metastasis, and two with meningioma as controls. MRI data included T2 weighted images and contrast enhanced T1 weighted images, NODDI, and DTI. Three neurosurgeons manually assigned volumes of interest (VOIs) to the NETs and Edemas. The DTI and NODDI-derived parameters calculated for each VOI were fractional anisotropy (FA), apparent diffusion coefficient (ADC), intracellular volume fraction (ICVF), isotropic volume fraction (ISOVF), and orientation dispersion index. RESULTS: Sixteen and 14 VOIs were placed on NETs and Edemas, respectively. The ICVF, ISOVF, FA, and ADC values of NETs and Edemas differed significantly (p < 0.01). Receiver operating characteristic curve analysis revealed that using all parameters allowed for improved differentiation of NETs from Edemas (area under the curve = 0.918) from the use of NODDI parameters (0.910) or DTI parameters (0.899). Multiple logistic regression was performed with all parameters, and a predictive formula to differentiate between NETs and Edemas could be created and applied to the edematous regions of the negative control-group images; the tumor prediction degree was well below 0.5, confirming differentiation as edema. CONCLUSIONS: Using NODDI and DTI may prove useful in differentiating NETs from Edemas in the non-contrast T2 hyperintensity region of glioblastomas.