Introduction

Stereotactic Radiosurgery (SRS) is an efficacy procedure in treatment of brain disease. The radiosurgery was initiated by Leksell gamma procedure [1], and modernized methods have been extended to linear accelerators [2], Cyberknife [3] and so on. The complicated SRS procedure includes simulation, target definition, treatment planning, target localization and dose delivery [4-13]. The external accuracy verification of the whole procedure has been investigated with different quality assurance methodologies [5,16]. However, the final estimation of SRS procedure should be reflected in the disease lesions inside the patient, and this could be done by employing different imaging modalities at different temporal points, but challenges exist in abstracting the weak signal due to radiation in the images. The rationale of this philosophy is to mimic the physics study of a few atomic layers of surface morphology between strong Bragg Peaks [4]. More specifically, the perturbation information could be used to estimate and predict the potential abnormal structure at the phases of pretreatment, during treatment and post treatment. Therefore, in this study, a method was used to estimate the perturbation information in MRIs at different temporal points after a cerebellum target SRS, and the targeting to understand the rationale of radiosurgery in target characteristics, dosimetry accuracy, and radiation dose delivery response and improvement of radiosurgery strategy.

Methods and Materials

This radiosurgery plan was generated in an Eclipse Treatment planning system from Varian medical system. The treatment planning system simulated the x-ray beam, which was generated by a linear accelerator. The simulation included the geometric structure of linear accelerator system and with accurate calibration of dosimetry accuracy. After the dosimetry requirement was given, a dose delivery approach was developed as showed in figure 1. Then patient was treated with high precise setup in the treatment room, and accurate dose was delivered to the intended target. Afterwards, different imaging modalities could be employed to trace the treatment outcome.

Firstly, a cerebellum target was under an SRS with a single ARC small aperture cone on a Linac machine from Varian Medical system. The SRS plan is shown in figure 1.

Figure 1: Single ARC Cone plan for SRS.

In this study, a series of MRIs in different temporal points have been obtained, the temporal range were 0 months, 3 months, 6 months, and 9 months. The volume of interest scans were defined by the isodose volume in the dosimetric plan, which included target volume, and isodose volumes which were at different isodose levels including 100%, 90%, 75%, 60%, 30% and 15% of prescription dose, which is shown in figure 2.

Perturbation information with MRI analysis of Brain Lesion post Radiosurgery

Figure 2: isodose volumes are shown in axial, sagittal and coronal view. Corresponding volumes were 0.04, 0.28, 0.48, 0.77, 2.67, and 10.70cc.

Afterwards, through image fusion method, these volumes of interest were defined in the MRIs through the function of copy structures to registered image. Then structure property function to obtain the structure statistics including minimum Hounsfield Unit (HU), maximum HU, mean HU, and standard deviation (SD) of HU inside the volume of interest (VOI), and this procedure is shown in figure 3.

Figure 3: Pixel information of Volume of interest (VOI). It was directly application of the property function of VOI from contouring module from Eclipse Treatment Planning System from Varian Medical System.

Finally, a relative percentage difference method, which was defined to be the ratio between the differences of SD and mean SD divided by the mean SD to separate the technical variation from imaging procedure. The formula below shows this estimation procedure.

(Note: HSD (HU Standard Deviation in VOI V).
SD (Standard Deviation).
N (Number of VOIs or pixels in a VOI)
P (Pixel in VOI))

Result

For the selected volumes of interest, the mean SD HUs were 20.8, 20.4, 24.1, and 26.21 for T1 MRIs, and was 35.2, 17.4, 31.9, and 37.1 for T1 MRIs with contrast. And the least difference in SD HU vector elements was at 3 months, and the average absolute SD HUs was about three in magnitude. Moreover, the relative percentage difference showed a time-spatial vector pattern with special characteristics. Figure 4 shows some of the analysis. In figure 4(a), the MR T1 with contrast image showed a dip in the 3-month time point, and this implied that this imaging setting is sensitive to the treatment lesion response. And in figure 4(b), the observation perturbation in formula 1 for surface dose volumes at simulation CT and different MRIs were plotted. The comparison showed the obvious variation happen 15% prescription line, which is 3Gy at this treatment. While considering the actual size of the target, this plot showed the optimal sensitivity volume for analysis in this type of study. Then, as another support of this volume of interest selection, figure 4(c) was another support for this range of volume of interest by plotting the T1 and T1C SDHU difference, which is the SDHU of T1 MRI subtracting that of the T1 MRI with contrast.

Figure 4 (a): Average standard deviations of HU of T1 images at different temporal points in different VOIs and corresponding SD SD HU.