Logo: to the web site of Uppsala University

  Background

Daily adaptive radiotherapy, as performed with the Elekta Unity MR-Linac, requires choosing between different adaptation methods, namely ATP (Adapt to Position) and ATS (Adapt to Shape), where the latter requires daily re-contouring to obtain a dose plan tailored to the daily anatomy. These steps are inherently resource-intensive, and quickly predicting the dose distribution and the dosimetric evaluation criteria while the patient is on the table could facilitate a fast selection of adaptation method and decrease the treatment times.

Purpose

In this work, we aimed to develop a deep-learning-based dose-prediction pipeline for prostate MR-Linac treatments.

Methods

Two hundred twelve MR-images, structure sets, and dose distributions from 35 prostate patients treated with 6.1 Gy for 7 or 6 fractions at our MR-Linac were included, split into train/test partitions of 152/60 images, respectively. A deep-learning segmentation network was trained to segment the CTV (prostate), bladder, and rectum. A second network was trained to predict the dose distribution based on manually delineated structures. At inference, the predicted segmentations acted as input to the dose prediction network, and the predicted dose was compared to the true (optimized in the treatment planning system) dose distribution.

Results

Median DSC values from the segmentation network were 0.90/0.94/0.87 for CTV/bladder/rectum. Predicted segmentations as input to the dose prediction resulted in mean differences between predicted and true doses of 0.7%/0.7%/1.7% (relative to the prescription dose) for D_98%/D_95%/D_2% for the CTV. For the bladder, the difference was 0.7%/0.3% for D_mean/D_2% and for the rectum 0.1/0.2/0.2 pp (percentage points) for V_33Gy/V_38Gy/V_41Gy. In comparison, true segmentations as input resulted in differences of 1.1%/0.9%/1.6% for CTV, 0.5%/0.4% for bladder, and 0.7/0.4/0.3 pp for the rectum. Only D_2% for CTV and D_mean/D_2% for bladder were found to be statistically significantly better when using true structures instead of predicted structures as input to the dose prediction.

Conclusions

Small differences in the fulfillment of clinical dose-volume constraints are seen between utilizing deep-learning predicted structures as input to a dose prediction network and manual structures. Overall mean differences <2% indicate that the dose-prediction pipeline is useful as a decision support tool where differences are >2%.

The recent introduction of a commercial 1.5 T MR-linac system has considerably improved the image quality of the patient acquired in the treatment unit as well as enabling online new technology requires new methodology that allows for the high field MR in a linac environment. The presence of the magnetic field requires special attention to the phantoms, detectors, and tools to perform QA. Due to the design of the system, the integrated megavoltage imager (MVI) is essential for radiation beam calibrations and QA. Additionally, the alignment between the MR image system and the radiation isocenter must be checked. The MR-linac system has vendor-supplied phantoms for calibration and QA tests. However, users have developed their own routine QA systems to independently check that the machine is performing as required, as to ensure we are able to deliver the intended dose with sufficient certainty. The aim of this work is therefore to review the MR-linac specific QA procedures reported in the literature.Semin Radiat Oncol 34:120-128

Purpose: The precision of the dose delivery in radiation therapy with high-field MR-linacs is challenging due to the sub-stantial variation in the beam attenuation of the patient positioning system (PPS) (the couch and coils) as a function of the gantry angle. This work aimed to compare the attenuation of two PPSs located at two different MR-linac sites through measurements and calculations in the treatment planning system (TPS).

Methods: Attenuation measurements were performed at every 1 degrees gantry angle at the two sites with a cylindrical water phantom with a Farmer chamber inserted along the rotational axis of the phantom. The phantom was positioned with the chamber reference point (CRP) at the MR-linac isocentre. A compensation strategy was applied to minimise sinusoidal measurement errors due to, e.g. air cavity or setup. A series of tests were performed to assess the sensitivity to measurement uncertainties. The dose to a model of the cylindrical water phantom with the PPS added was calculated in the TPS (Monaco v5.4 as well as in a development version Dev of an upcoming release), for the same gantry angles as for the measurements. The TPS PPS model dependency of the dose calculation voxelisation resolution was also investigated.

Results: A comparison of the measured attenuation of the two PPSs yielded differences of less than 0.5% for most gantry angles. The maximum deviation between the attenuation measurements for the two different PPSs exceeded +/- 1% at two specific gantry angles 115 degrees and 245 degrees, where the beam traverses the most complex PPS structures. The attenuation increases from 0% to 25% in 15 degrees intervals around these angles. The measured and calculated attenuation, as calculated in v5.4, was generally within 1-2% with a systematic overestimation of the attenuation for gantry angles around 180 degrees, as well as a maximum error of 4-5% for a few discrete angles in 10 degrees gantry angle intervals around the complex PPS structures. The PPS modelling was improved compared to v5.4 in Dev, especially around 180 degrees, and the results of those calculations were within +/- 1%, but with a similar 4% maximum deviation for the most complex PPS structures.

Conclusions: Generally, the two tested PPS structures exhibit very similar attenuation as a function of the gantry angle, including the angles with a steep change in attenuation. Both TPS versions, v5.4 and Dev delivered clinically acceptable accuracy of the calculated dose, as the differences in the measurements were overall better than +/- 2%. Additionally, Dev improved the accuracy of the dose calculation to +/- 1% for gantry angles around 180 degrees.

Purpose: Dose painting (DP) is a radiation therapy (RT) strategy for patients with heterogeneous tumors delivering higher dose to radiation resistant regions and less to sensitive ones, thus aiming to maximize tumor control with limited side effects. The success of DP treatments is influenced by the spatial accuracy in dose delivery. Adaptive RT (ART) workflows can reduce the overall geometric dose delivery uncertainty. The purpose of this study is to dosimetrically compare ART and non-adaptive conventional RT workflows for delivery of DP prescriptions in the treatment of prostate cancer (PCa).

Materials and methods: We performed a planning and treatment simulation study of four study arms. Adaptive and conventional workflows were tested in combination with DP and Homogeneous dose. We used image data from 5 PCa patients that had been treated on the Elekta Unity MR linac; the patients had been imaged in treatment position before each treatment fraction (7 in total). The local radiation sensitivity from apparent diffusion coefficient maps of 15 high-risk PCa patients was modelled in a previous study. these maps were used as input for optimization of DP plans aiming for maximization of tumor control probability (TCP) under rectum dose constraints. A range of prostate doses were planned for the homogeneous arms. Adaptive plans were replanned based on the anatomy-of-the-day, whereas conventional plans were planned using a pre-treatment image and subsequently recalculated on the anatomy-of-the-day. The dose from 7 fractions was accumulated using dose mapping. The endpoints studied were the TCP and dose-volume histogram metrics for organs at risk.

Results: Accumulated DP doses (adaptive and conventional) resulted in high TCP, between 96-99%. The largest difference between adaptive and conventional DP was 2.6 percentage points (in favor of adaptive DP). An analysis of the dose per fraction revealed substantial target misses for one patient in the conventional workflow that-if systematic-could jeopardize the TCP. Compared to homogeneous prescriptions with equal mean prostate dose, DP resulted in slightly higher TCP.

Conclusion: Compared to homogeneous dose, DP maintains or marginally increases the TCP. Adaptive DP workflows could avoid target misses compared to conventional workflows.

Background and Purpose: Treatments on combined Magnetic Resonance (MR) scanners and Linear Accelerators (Linacs) for radiotherapy, called MR-Linacs, often require daily contouring. Currently, deformable image registration (DIR) algorithms propagate contours from reference scans, however large shape and size changes can be troublesome. Artificial neural network (ANN) based contouring may alleviate this issue, however generally requires large datasets for training. Mitigating the problem of scarcity of data, we propose patient specific networks trained on a single dataset for each patient, for contouring onto the following datasets in an adaptive MR-Linacworkflow. Materials and Methods: MR-scans from 17 prostate patients treated on an MR-Linac with contours of Clinical Target Volume (CTV), bladder and rectum were utilized. U-net shaped models were trained based on the image from the first fraction of each patient, and subsequently applied onto the following treatment images. Results were compared with manual contours in terms of the Dice coefficient and Added Path Length (APL). As benchmark, contours propagated through the clinical DIR algorithm were similarly evaluated. Results: In Dice coefficient the ANN output was 0.92 +/- 0.03, 0.93 +/- 0.07 and 0.84 +/- 0.10 while for DIR 0.95 +/- 0.03, 0.93 +/- 0.08, 0.88 +/- 0.06 for CTV, bladder and rectum respectively. Similarly, APL where 3109 +/- 1642, 7250 +/- 4234 and 5041 +/- 2666 for ANN and 1835 +/- 1621, 7236 +/- 4287 and 4170 +/- 2920 voxels for DIR. Conclusions: Patient specific ANN models trained on images from the first fraction of a prostate MR-Linac treatment showed similar accuracy when applied to the subsequent fraction images as a clinically implemented DIR method.

The clinical introduction of hybrid magnetic resonance (MR) guided radiotherapy (RT) delivery systems has led to the need to validate the end-to-end dose delivery performance on such machines. In the current study, an MR visible phantom was developed and used to test the spatial deviation between planned and delivered dose at two 1.5 T MR linear accelerator (MR linac) systems, including pre-treatment imaging, dose planning, online imaging, image registration, plan adaptation, and dose delivery. The phantom consisted of 3D printed plastic and MR visible silicone rubber. It was designed to minimise air gaps close to the radiochromic film used as a dosimeter. Furthermore, the phantom was designed to allow submillimetre, reproducible positioning of the film in the phantom. At both MR linac systems, 54 complete adaptive, MR guided RT workflow sessions were performed. To test the dose delivery performance of the MR linac systems in various adaptive RT (ART) scenarios, the sessions comprised a range of systematic positional shifts of the phantom and imaging or plan adaptation conditions. In each workflow session, the positional translation between the film and the adaptive planned dose was determined. The results showed that the accuracy of the MR linac systems was between 0.1 and 0.9 mm depending on direction. The highest mean deviance observed was in the posterior-anterior direction, and the direction of the error was consistent between centres. The precision of the systems was related to whether the workflow utilized the internal image registration algorithm of the MR linac. Workflows using the internal registration algorithm led to a worse precision (0.2-0.7 mm) compared to workflows where the algorithm was decoupled (0.2 mm). In summary, the spatial deviation between planned and delivered dose of MR-guided ART at the two MR linac systems was well below 1 mm and thus acceptable for clinical use.

Background and purpose:

Devices that combine an MR-scanner with a Linac for radiotherapy, referred to as MR-Linac systems, introduce the possibility to acquire high resolution images prior and during treatment. Hence, there is a possibility to acquire individualised learning sets for motion models for each fraction and the construction of intrafractional motion models. We investigated the feasibility for a principal component analysis (PCA) based, intrafractional motion model of the male pelvic region.

Materials and methods:

4D-scans of nine healthy male volunteers were utilized, FOV covering the entire pelvic region including prostate, bladder and rectum with manual segmentation of each organ at each time frame. Deformable image registration with an optical flow algorithm was performed for each subject with the first time frame as reference. PCA was performed on a subset of the resulting displacement vector fields to construct individualised motion models evaluated on the remaining fields.

Results:

The registration algorithm produced accurate registration result, in general DICE overlap >0.95 across all time frames. Cumulative variance of the eigen values from the PCA showed that 50% or more of the motion is explained in the first component for all subjects. However, the size and direction for the components differed between subjects. Adding more than two components did not improve the accuracy significantly and the model was able to explain motion down to about 1 mm.ConclusionsAn individualised intrafractional male pelvic motion model is feasible. Geometric accuracy was about 1 mm based on 1–2 principal components.

Background and purpose: Probabilistic optimization is an alternative to margins for handling geometrical uncertainties in treatment planning of radiotherapy where uncertainties are explicitly incorporated in the optimization. We present a novel probabilistic method based on the same statistical measures as those behind conventional margin based planning. Material and methods: Percentile Dosage (PD) was defined as the dose coverage that a treatment plan meet or exceed to a given probability. For optimization, we used the convex measure Expected Percentile Dosage (EPD) defined as the average dose coverage below a given PD. An iterative method gradually adjusted the constraint tolerance associated with the EPD until the desired target PD was met. It was applied to planning of cervical cancer patients focusing on systematic uncertainty caused by organ deformation. The resulting plans were compared to margin based plans using target and organ at risk PDs. Results: The EPD tolerance converged in less than ten iterations to produce a PD within 0.1 Gy of the requested. The PD was on average within 0.5% of the requested PD when validated versus independent scenarios. The rectum volume, extracted from the PDs, receiving 90% of the intended target dose was decreased with 16% for the same target PD in comparison to margin based plans. Conclusions: The proposed probabilistic optimization method enabled prescription of a dose volume histogram metric to a chosen confidence. The probabilistic plans showed improved target dose homogeneity and decreased rectum dose for the same target dose coverage compared to margin based plans.

A comprehensive methodology for treatment simulation and evaluation of dose coverage probabilities is presented where a population based statistical shape model (SSM) provide samples of fraction specific patient geometry deformations.The learning data consists of vector fields from deformable image registration of repeated imaging giving intra-patient deformations which are mapped to an average patient serving as a common frame of reference. The SSM is created by extracting the most dominating eigenmodes through principal component analysis of the deformations from all patients. The sampling of a deformation is thus reduced to sampling weights for enough of the most dominating eigenmodes that describe the deformations.For the cervical cancer patient datasets in this work, we found seven eigenmodes to be sufficient to capture 90% of the variance in the deformations of the, and only three eigenmodes for stability in the simulated dose coverage probabilities. The normality assumption of the eigenmode weights was tested and found relevant for the 20 most dominating eigenmodes except for the first. Individualization of the SSM is demonstrated to be improved using two deformation samples from a new patient. The probabilistic evaluation provided additional information about the trade-offs compared to the conventional single dataset treatment planning.