Browsing by Subject "Radiosurgery"

Purpose: Radiosurgery has become a widely used procedure in the treatment of both solid tumors and secondary metastases in the brain. In cases with multiple brain lesions, isocenters are typically set up for each target, a process which can take hours and become very uncomfortable for the patient. Recently, multifocal treatments with a single isocenter have emerged as a solution. With the high doses delivered to small regions during radiosurgery, the importance of treatment verification is paramount, especially when delivering high doses to regions off isocenter.

Methods: A 5-arc RapidArc radiosurgery plan with a single isocenter and 5 targets was used to treat a dosimeter placed within a RPC-type head and neck phantom. The treatment was delivered five times at varying prescription doses, depending on the sensitivity of the PRESAGE dosimeter used. The delivered dose distribution was measured using an in-house optical-CT system and compared to the Eclipse-planned dose distribution using dose volume histograms and Gamma analysis.

Results: Reasonable dose agreement was measured between the majority of the dosimeters and the Eclipse plan (80-85% pass rate at 5%/3 mm Gamma critera). The failing voxels were located on the periphery of the dosimeter at regions of extremely high or low dose, suggesting a dose dependent stability of the PRESAGE formulation. The formulation with the best temporal stability had a much higher Gamma pass rate of 98% at 3%/2mm criteria.

Conclusions: The potential of accurate delivery of the complex radiosurgery plan was demonstrated with one of the three formulations of PRESAGE. While agreement was worse in the other formulations, the problem seemed to be an easily-fixable stability issue, resulting in improper scaling of doses. Replication of the most stable formulation would provide an excellent tool for verification of radiosurgery treatment delivery and other complex procedures.

Background: Radiomics is a process of converting medical images into high-dimensional quantitative features and the subsequent mining these features for providing decision support. It is conducted as a potential noninvasive, low-cost, and patient-specific routine clinical tool. Building a predictive model which is reliable, efficient, and accurate is a vital part for the success of radiomics. Machine learning method is a powerful tool to achieve this. Feature extraction strongly affects the performance. Delta-feature is one way of feature extraction methods to reflect the spatial variation in tumor phenotype, hence it could provide better treatment-specific assessment.

Purpose: To compare the performance of using pre-treatment features and delta-features for assessing the brain radiosurgery treatment response, and to investigate the performance of different combinations of machine learning methods for feature selection and for feature classification.

Materials and Methods: A cohort of 12 patients with brain treated by radiosurgery was included in this research. The pre-treatment, one-week post-treatment, and two-month post-treatment T1 and T2 FLAIR MR images were acquired. 61 radiomic features were extracted from the gross tumor volume (GTV) of each image. The delta-features from pre-treatment to two post-treatment time points were calculated. With leave-one-out sampling, pre-treatment features and the two sets of delta-features were separately input into a univariate Cox regression model and a machine learning model (L1-regularized logistic regression [L1-LR], random forest [RF] or neural network [NN]) for feature selection. Then a machine learning method (L1-LR, L2-regularized logistic regression [L2-LR], RF, NN, kernel support vector machine [Kernel-SVM], linear-SVM, or naïve bayes [NB]) was used to build a classification model to predict overall survival. The performance of each model combination and feature type was estimated by the area under receiver operating characteristic (ROC) curve (AUC).

Results: The AUC of one-week delta-features was significantly higher than that of pre-treatment features (p-values < 0.0001) and two-month delta-features (p-value= 0.000). The model combinations of L1-LR for feature selection and RF for classification as well as RF for feature selection and NB for classification based on one-week delta-features presented the highest AUC values (both AUC=0.944).

Conclusions: This work potentially implied that the delta-features could be better in predicting treatment response than pre-treatment features, and the time point of computing the delta-features was a vital factor in assessment performance. Analyzing delta-features using a suitable machine learning approach is potentially a powerful tool for assessing treatment response.

Purpose: Linear accelerator-based Stereotactic Radiosurgery (SRS) is often performed using either dynamic conformal arcs or VMAT. For multifocal disease, multifocal conformal arc techniques can struggle to deliver the desired dose with high conformity for all targets simultaneously. While VMAT may improve coverage and conformity, and can offer the planner more flexibility, it can result in highly modulated treatment plans with non-intuitive MLC trajectories. The complex MLC modulation trajectories can often struggle to shield healthy areas between targets, thus leaving open gaps being irradiated between targets. The purpose of this research is to overcome these limitations by developing a technique for SRS of multifocal targets that combines the intuitive MLC trajectories of dynamic conformal arcs with the flexibility of VMAT.

Methods: A Conformal Arc Informed VMAT (CAVMAT) planning technique was developed in which arcs are assigned subgroups of targets, for which the MLCs are able to effectively conform to all targets in the subgroup. Arc weights are optimized to achieve desired dose per target while minimizing the variation in MU per arc. The optimized conformal arc plan then serves as the starting point in a VMAT inverse optimization to fine tune the dose to each target, optimize conformity, and meet any plan specific objectives. To demonstrate feasibility, ten multifocal VMAT cases were re-planned using the CAVMAT technique. The following metrics of plan quality were used to compare VMAT with CAVMAT: volume of healthy brain receiving 6Gy, 12Gy, and 16Gy, conformity index, and total number of monitor units.

Results: The V6Gy of the healthy brain was 10±13% lower in CAVMAT than in VMAT (range 25% lower to 15% higher for CAVMAT plans than VMAT plans). V12Gy of healthy brain tissue showed 5±14% lower in CAVMAT than in VMAT (range 16% lower to 24% higher in CAVMAT plans than VMAT plans). The V16Gy of healthy brain tissue was 3±16% lower in CAVMAT than in VMAT (range 16% to 4% lower and 41% higher in CAVMAT plans than VMAT plans in one case). The MU (Monitor Units) for the CAVMAT plans were 6156.4MU with a standard deviation of 878.41MU compared to 7031.3MU with a deviation of 1788.89MUs for VMAT. The CI (Conformity Index) for CAVMAT are 1.31 with a standard deviation of 0.13, the VMAT plan has a mean conformity of 1.28 with a standard deviation of 0.18. The mean maximum dose of the CAVMAT plan is 2445.37cGy with a standard deviation of 107.22cGy compared to 2309.28cGy with a standard deviation of 114.72cGy for VMAT.

Conclusion: CAVMAT plans succeeded in lessening low dose spill with lower MUs on average compared to VMAT plans. The conformity indexes are comparable to VMAT plans and maximum doses to patients are higher in the CAVMAT plans than in the VMAT plans.

Brain metastases can occur in 20%-40% of cancer patients. Single isocenter multi-target (SIMT) radiosurgery planned with volumetric modulated arc therapy (VMAT) is a method of treating multiple brain metastases simultaneously. These treatments deliver high doses which emphasizes the need for accuracy and avoidance of critical neurological structures to maintaining a patient’s quality of life (QOL). QOL is highly correlated to patient memory, suggesting the hippocampus as a critical structure to avoid in treatment due to its significant role in episodic memory development. Radiosurgery treatments commonly use specialized MLC leaves (HD) that have a narrower width to improve dose conformity around metastases. This work serves to evaluate the feasibility of avoiding the hippocampus in SIMT treatments using VMAT based on MLC leaf width, as well as the effects of hippocampal avoidance on plan quality.40 patients, each previously treated with SIMT planned with VMAT for between 4-10 brain metastases using HD-MLCs, enrolled in an IRB approved protocol. The plans were evaluated for meeting RTOG 0933 recommended dose constraints to the hippocampus. If constraints were not met, then treatments were replanned with optimization objectives added to the hippocampus and/or arc orientation adjustments to meet constraints without compromising target coverage or other organ-at-risk (OAR) dose constraints. Afterwards, the treatments both with and without hippocampal objectives were replanned using standard width MLC leaves (SD-MLCs). All 4 plan types were then evaluated for plan quality using conformity index, V20%[cc], V50%[cc], V75%[cc] and D50%[cGy]. 8 total hippocampi across 7 patient plans exceeded constraints. 4 of these hippocampi could be spared through optimization objective adjustment, and 1 more with an arc orientation adjustment. Biological effective dose (BED) to the 8 hippocampi decreased by 29% ± 23% (p= 0.007) with no significant changes in dose to normal tissue when planned with the addition of hippocampal objectives. SD-MLCs showed similar results, sparing the same 5 of the 8 hippocampi exceeding constraints, but also increased dose to healthy tissue, most substantially in V20%[cc] which increased 59.56% ± 53% (p= 0.015) and 48.22% ± 32.17% (p= 0.0056) in plans with and without hippocampal objectives respectively. Also, in plans both with and without hippocampal optimization objectives, there was no significant change in conformity index when switching MLCs from HD to SD leaves. Meeting recommended hippocampal constraints was possible in 57% of patients that initially exceeded constraints. The ability to spare this structure was independent of MLC width, and correlated to the distance between the hippocampus and the nearest target. From this data set, the smallest distance avoidable was 0.45cm. All un-avoided hippocampi were at least touching a target, if not overlapped. The larger MLC leaves resulted in higher doses to larger volumes of normal tissue, however the planning technique of VMAT was able to meet target coverage without compromising treatment conformity when larger MLC leaves were used.

Purpose: To build a model that will predict the survival time for patients that were treated with stereotactic radiosurgery for brain metastases using support vector machine (SVM) regression.

Methods and Materials: This study utilized data from 481 patients, which were equally divided into training and validation datasets randomly. The SVM model used a Gaussian RBF function, along with various parameters, such as the size of the epsilon insensitive region and the cost parameter (C) that are used to control the amount of error tolerated by the model. The predictor variables for the SVM model consisted of the actual survival time of the patient, the number of brain metastases, the graded prognostic assessment (GPA) and Karnofsky Performance Scale (KPS) scores, prescription dose, and the largest planning target volume (PTV). The response of the model is the survival time of the patient. The resulting survival time predictions were analyzed against the actual survival times by single parameter classification and two-parameter classification. The predicted mean survival times within each classification were compared with the actual values to obtain the confidence interval associated with the model’s predictions. In addition to visualizing the data on plots using the means and error bars, the correlation coefficients between the actual and predicted means of the survival times were calculated during each step of the classification.

Results: The number of metastases and KPS scores, were consistently shown to be the strongest predictors in the single parameter classification, and were subsequently used as first classifiers in the two-parameter classification. When the survival times were analyzed with the number of metastases as the first classifier, the best correlation was obtained for patients with 3 metastases, while patients with 4 or 5 metastases had significantly worse results. When the KPS score was used as the first classifier, patients with a KPS score of 60 and 90/100 had similar strong correlation results. These mixed results are likely due to the limited data available for patients with more than 3 metastases or KPS scores of 60 or less.

Conclusions: The number of metastases and the KPS score both showed to be strong predictors of patient survival time. The model was less accurate for patients with more metastases and certain KPS scores due to the lack of training data.

Purpose:

A new development in linac-based intracranial stereotactic radiosurgery (SRS) and extracranial stereotactic body radiation therapy (SBRT) is treatment of multiple targets using single isocenter volumetric modulated arc therapy (VMAT) technique, dramatically reducing treatment time while maintaining high target conformality and steep dose gradients between targets and surrounding organs at risk (OAR). In VMAT, the gantry rotates around the patient while continuously delivering radiation. Throughout the VMAT arc, the beam is modulated based on an inverse optimization algorithm in order to spare organs at risk. Single isocenter multi-target VMAT has already been implemented for intracranial SRS and is increasingly used for extracranial SBRT treatments. Despite the increasing popularity of this technique, certain inherent clinically meaningful challenges warrant further investigation. Specifically, single isocenter, multifocal SRS and SBRT can result in small volumes targets with a large off-axis distance from the treatment isocenter. Consequently, angular errors in the collimator, patient support assembly (PSA), or gantry could have an increased impact on target coverage, warranting a re-evaluation of routine linear accelerator QA tolerance recommendations in TG-142. Also, questions have arisen regarding the ability of clinical dose calculation algorithms to calculate dose accurately for these cases at large off-axis distances. Specifically, it is questionable whether or not the MLC model used is sophisticated enough to accurately model the dose off axis. This is of concern because the MLC leaves have different dimensions outside of the HD region, but the dosimetric-leaf-gap model used is the same for both regions. Applying the single isocenter technique to extracranial SBRT of oligometastases introduces additional unique challenges that must be addressed. These include greater intra and inter-fractional setup uncertainties, and dosimetric interplay since immobilization is more difficult and internal motion is non-negligible. The purpose of this thesis is to explore these specific physics and treatment planning considerations for single isocenter multi-target intracranial radiosurgery and extracranial SBRT.

Materials and Methods:

Intracranial SRS

For single isocenter multifocal stereotactic radiosurgery, we evaluated potential dose deviations from mechanical errors in PSA, collimator, and gantry angle within the tolerance recommended by TG-142 for radiosurgery machines. Systematic errors in PSA, collimator, and gantry angle were introduced at the recommended tolerance levels into both multifocal SRS plans and traditional single target SRS using dynamic conformal arcs, and the resulting dosimetric effect were quantified within the treatment planning system. In addition, we quantified the accuracy of the treatment planning system dose calculation algorithm for targets located at large off-axis distances with 3D Slicer analysis software. The dose distribution from the treatment planning system was compared to the distribution measured using a high-resolution 3D dosimetry system (PRESAGE®/Optical-CT). Comparisons were made using DVH and gamma analysis (criteria: 2mm/3%, 3mm/3%, and 3mm/5%). DVH comparisons included a shell analysis in which we compared the dose from 2mm within the target to the target’s surface, the surface to 2mm outside the target, and 2mm to 4mm outside the target for both the plan and the dosimeter.

Extracranial SBRT

For applying the single isocenter multifocal technique to extracranial oligometastases, we propose a treatment method that addresses intra and inter-fractional motion as well as dosimetric interplay. The developed technique uses a Single Isocenter with Distinct Optimizations (SIDO) in which all Volumetric Modulated Arc Therapy (VMAT) fields share an isocenter but each field treats only one target. When necessary, setup uncertainties from rotations and deformations are mitigated by applying a couch translation between VMAT arcs, and interplay is minimized using dynamic conformal arcs (DCAs) as the starting point for inverse optimization. We evaluated this planning technique using relevant dose indices including conformity index, gradient index and modulation factor.

Results:

Intracranial SRS

Induced errors at TG-142 tolerance levels showed the greatest change in multifocal SRS target coverage for collimator and gantry rotations, while minimal change in coverage was noted for errors in PSA rotation. For single isocenter cases, the largest dose discrepancies were a result of 1° errors in the collimator and gantry angles, specifically with respect to the volume of the PTV receiving the prescription dose. These errors caused up to 33% and 18% deviations, respectively, to the volume of the PTV receiving the prescription dose with mean deviations of 5% and 2%, respectively. When the collimator and gantry errors in single isocenter plans were reduced to 0.5°, the discrepancies in the volume of the PTV receiving the prescription dose were reduced to a max value of <5% for the gantry and collimator with mean values of approximately 1%. For 1° errors in DCA plans, however, deviations to the volume of the PTV receiving the prescription dose did not exceed 5% for the collimator, couch, or gantry and similar results were seen in all other dosimetric indices investigated.

A preliminary analysis of the Eclipse dose calculation algorithm in comparison to actual dose delivered to targets shows agreement with 89.46%, 94.87%, and 96.39% of voxels having a passing gamma index with criteria of 2mm/3%, 3mm/3%, and 3mm/5% respectively (distance from isocenter ranged from 0-10cm). Targets within 8cm of the isocenter showed less than 2% discrepancy between the plan and measurement with respect to the percent of the target receiving the prescription dose. The target at 10cm from the isocenter, however, had a 15% discrepancy between the plan and measurement with respect to the percent of the target receiving the prescription dose and therefore warrants further investigation.

Extracranial SBRT

As the distance between targets increases, the probability for requiring a second translation between treatment arcs also increases. Assuming a margin of 5mm and considering six lung and five liver patient cases, a second translational shift would be required 0%, ~10%, and ~25% of the time for target separations of 5cm, 10cm, and 15cm respectively.

For greater than 3cm separation between targets in extracranial SBRT, SIDO and SIDO with DCA have an average conformity index of 0.862, and 0.864 respectively, which is comparable to the average conformity of traditional multifocal treatment techniques at these target separations of 0.901. When separation between PTVs is less than 3cm, however, traditional single isocenter VMAT has superior conformity with a mean value of 0.875, as opposed to 0.772 and 0.782 for SIDO and SIDO with DCA respectively; and decreasing conformity with decreasing target separation. SIDO with DCA had superior GI over all other planning techniques for almost all cases with a mean value of 7.31 across all target separations. SIDO with DCA even performed better than the DCA technique with a mean GI of 7.43 across all target separations, which was assumed to be the best method for obtaining a desirable GI. SIDO with DCA had a comparable MF to the DCA plans and was closer to 1 than all other planning techniques. The mean MF values across all target separations for SIDO with DCA and DCA were 1.17 and 0.83 respectively.

Conclusions:

Intracranial SRS

Institutions utilizing a single isocenter VMAT technique for multifocal disease should pay careful attention to the angular mechanical tolerances in designing a robust and complete QA program, especially with respect to the collimator and gantry recommended tolerances. We recommend reducing collimator and gantry tolerances from 1.0° to 0.5° to decrease the potential magnitude of deviations between the planning and delivered dose distributions. The PRESAGE®/Optical-CT 3D dosimetry system verified the accuracy of the Eclipse dose calculation algorithm to within 2% for targets located up to 8cm from the isocenter. Further investigation is required for more distal targets, as they did not have sufficient agreement.

Extracranial SBRT

A single isocenter approach for SBRT treatment of extracranial oligometastases may be feasible using the proposed SIDO and SIDO with DCA treatment planning techniques. SIDO for extracranial oligometastases allows flexibility to mitigate spatial uncertainties from rotation and deformation, and has comparable dosimetry to traditional VMAT with low modulation when inverse optimization begins with DCAs. These advantages make SIDO beneficial for target separations of greater than 3cm, however, for target separations less than 3cm a traditional single isocenter technique is more appropriate.