Human-level comparable control volume mapping with a deep unsupervised-learning model for image-guided radiation therapy.

PURPOSE: To develop a deep unsupervised learning method with control volume (CV) mapping from patient positioning daily CT (dCT) to planning computed tomography (pCT) for precise patient positioning. METHODS: We propose an unsupervised learning framework, which maps CVs from dCT to pCT to automatically generate the couch shifts, including translation and rotation dimensions. The network inputs are dCT, pCT and CV positions in the pCT. The output is the transformation parameter of the dCT used to setup the head and neck cancer (HNC) patients. The network is trained to maximize image similarity between the CV in the pCT and the CV in the dCT. A total of 554 CT scans from 158 HNC patients were used for the evaluation of the proposed model. At different points in time, each patient had many CT scans. Couch shifts are calculated for the testing by averaging the translation and rotation from the CVs. The ground-truth of the shifts come from bone landmarks determined by an experienced radiation oncologist. RESULTS: The system positioning errors of translation and rotation are less than 0.47 mm and 0.17°, respectively. The random positioning errors of translation and rotation are less than 1.13 mm and 0.29°, respectively. The proposed method enhanced the proportion of cases registered within a preset tolerance (2.0 mm/1.0°) from 66.67% to 90.91% as compared to standard registrations. CONCLUSIONS: We proposed a deep unsupervised learning architecture for patient positioning with inclusion of CVs mapping, which weights the CVs regions differently to mitigate any potential adverse influence of image artifacts on the registration. Our experimental results show that the proposed method achieved efficient and effective HNC patient positioning.

A deep learning framework for prostate localization in cone beam CT-guided radiotherapy.

PURPOSE: To develop a deep learning-based model for prostate planning target volume (PTV) localization on cone beam computed tomography (CBCT) to improve the workflow of CBCT-guided patient setup. METHODS: A two-step task-based residual network (T2 RN) is proposed to automatically identify inherent landmarks in prostate PTV. The input to the T2 RN is the pretreatment CBCT images of the patient, and the output is the deep learning-identified landmarks in the PTV. To ensure robust PTV localization, the T2 RN model is trained by using over thousand sets of CT images with labeled landmarks, each of the CTs corresponds to a different scenario of patient position and/or anatomy distribution generated by synthetically changing the planning CT (pCT) image. The changes, including translation, rotation, and deformation, represent vast possible clinical situations of anatomy variations during a course of radiation therapy (RT). The trained patient-specific T2 RN model is tested by using 240 CBCTs from six patients. The testing CBCTs consists of 120 original CBCTs and 120 synthetic CBCTs. The synthetic CBCTs are generated by applying rotation/translation transformations to each of the original CBCT. RESULTS: The systematic/random setup errors between the model prediction and the reference are found to be <0.25/2.46 mm and 0.14/1.41° in translation and rotation dimensions, respectively. Pearson's correlation coefficient between model prediction and the reference is higher than 0.94 in translation and rotation dimensions. The Bland-Altman plots show good agreement between the two techniques. CONCLUSIONS: A novel T2 RN deep learning technique is established to localize the prostate PTV for RT patient setup. Our results show that highly accurate marker-less prostate setup is achievable by leveraging the state-of-the-art deep learning strategy.

Parametric Response Mapping of Coregistered Positron Emission Tomography and Dynamic Contrast Enhanced Computed Tomography to Identify Radioresistant Subvolumes in Locally Advanced Cervical Cancer.

PURPOSE: To identify subvolumes that may predict treatment response to definitive concurrent chemoradiation therapy using parametric response mapping (PRM) of coregistered positron emission tomography (PET) and dynamic contrast-enhanced (DCE) computed tomography (CT) in locally advanced cervical carcinoma. METHODS AND MATERIALS: Pre- and midtreatment (after 23 ± 4 days of concurrent chemoradiation therapy) DCE CT and PET imaging were performed on 21 patients with cervical cancer who were enrolled in a pilot study to evaluate the prognostic value of CT perfusion for primary cervical cancer (NCT01805141). Three-dimensional coregistered maps of PET/CT standardized uptake value (SUV) and DCE CT blood flow (BF) were generated. PRM was performed using voxel-wise joint histogram analysis to classify voxels within the tumor as highly metabolic and perfused (SUVhiBFhi), highly metabolic and hypoxic (SUVhiBFlo), low metabolic activity and hypoxic (SUVloBFlo), or low metabolic activity and perfused (SUVloBFhi) tissue based on thresholds determined from population means of pretreatment PET SUV and DCE CT BF. Relationships between baseline pretreatment imaging metrics and relative changes in metabolic tumor volume (ΔMTV), calculated from before treatment and during treatment imaging, were determined using univariable and multivariable linear regression models. RESULTS: The relative volume of three PRM subvolumes significantly changed during treatment (SUVhiBFhi: P = .04; SUVhiBFlo: P = .0008; SUVloBFhi: P = .02), whereas SUVloBFlo did not (P = .9). Pretreatment PET SUVmax (r = -.58, P = .006), PET SUVmean (ρ = -.59, P = .005), DCE CT BFmean (r = -.50, P = .02), tumor volume (ρ = -.65, P = .001) and PRM SUVhiBFhi (ρ = -.59, P = .004) were negatively correlated with ΔMTV, whereas PRM SUVloBFlo was positively related to ΔMTV (r = .77, P < .0001). In a multivariable model that predicted ΔMTV, PRM SUVloBFlo, which combines both PET/CT and DCE CT, was the only significant variable (ß = 1.825, P = .03), dominating both imaging modalities independently. CONCLUSIONS: PRM was applied in locally advanced cervical carcinoma treated definitively with chemoradiation, and radioresistant subvolumes were identified that correlated with changes in MTV and predicted treatment response. Identification of these subvolumes may assist in clinical decision making to tailor therapies, such as brachytherapy, in an effort to improve patient outcomes.

Intrafractional Tracking Accuracy of a Transperineal Ultrasound Image Guidance System for Prostate Radiotherapy.

PURPOSE: The aim of this study is to evaluate the tracking accuracy of a commercial ultrasound system under relevant treatment conditions and demonstrate its clinical utility for detecting significant treatment deviations arising from inadvertent intrafractional target motion. METHODS: A multimodality male pelvic phantom was used to simulate prostate image-guided radiotherapy with the system under evaluation. Target motion was simulated by placing the phantom on a motion platform. The tracking accuracy of the ultrasound system was evaluated using an independent optical tracking system under the conditions of beam-on, beam-off, poor image quality with an acoustic shadow introduced, and different phantom motion cycles. The time delay between the ultrasound-detected and actual phantom motion was investigated. A clinical case example of prostate treatment is presented as a demonstration of the utility of the system in practice. RESULTS: Time delay between the motion phantom and ultrasound tracking system is 223 ± 45.2 milliseconds including video and optical tracking system frame rates. The tracking accuracy and precision were better with a longer period. The precision of ultrasound tracking performance in the axial (superior-inferior) direction was better than that in the lateral (left-right) direction (root mean square errors are 0.18 and 0.25 mm, respectively). The accuracy of ultrasound tracking performance in the lateral direction was better than that in the axial direction (the mean position errors are 0.23 and 0.45 mm, respectively). Interference by radiation and image quality do not affect tracking ability significantly. Further, utilizing the tracking system as part of a clinical study for prostate treatment further verified the accuracy and clinical appropriateness. CONCLUSIONS: It is feasible to use transperineal ultrasound daily to monitor prostate motion during treatment. Our results verify the accuracy and precision of an ultrasound system under typical external beam treatment conditions and further demonstrate that the tracking system was able to identify important prostate shifts in a clinical case.

Quality control procedures for dynamic treatment delivery techniques involving couch motion.

In this study, the authors introduce and demonstrate quality control procedures for evaluating the geometric and dosimetric fidelity of dynamic treatment delivery techniques involving treatment couch motion synchronous with gantry and multileaf collimator (MLC). Tests were designed to evaluate positional accuracy, velocity constancy and accuracy for dynamic couch motion under a realistic weight load. A test evaluating the geometric accuracy of the system in delivering treatments over complex dynamic trajectories was also devised. Custom XML scripts that control the Varian TrueBeam™ STx (Serial #3) axes in Developer Mode were written to implement the delivery sequences for the tests. Delivered dose patterns were captured with radiographic film or the electronic portal imaging device. The couch translational accuracy in dynamic treatment mode was 0.01 cm. Rotational accuracy was within 0.3°, with 0.04 cm displacement of the rotational axis. Dose intensity profiles capturing the velocity constancy and accuracy for translations and rotation exhibited standard deviation and maximum deviations below 3%. For complex delivery involving MLC and couch motions, the overall translational accuracy for reproducing programmed patterns was within 0.06 cm. The authors conclude that in Developer Mode, TrueBeam™ is capable of delivering dynamic treatment delivery techniques involving couch motion with good geometric and dosimetric fidelity.

Algorithm for X-ray scatter, beam-hardening, and beam profile correction in diagnostic (kilovoltage) and treatment (megavoltage) cone beam CT.

Quantitative reconstruction of cone beam X-ray computed tomography (CT) datasets requires accurate modeling of scatter, beam-hardening, beam profile, and detector response. Typically, commercial imaging systems use fast empirical corrections that are designed to reduce visible artifacts due to incomplete modeling of the image formation process. In contrast, Monte Carlo (MC) methods are much more accurate but are relatively slow. Scatter kernel superposition (SKS) methods offer a balance between accuracy and computational practicality. We show how a single SKS algorithm can be employed to correct both kilovoltage (kV) energy (diagnostic) and megavoltage (MV) energy (treatment) X-ray images. Using MC models of kV and MV imaging systems, we map intensities recorded on an amorphous silicon flat panel detector to water-equivalent thicknesses (WETs). Scattergrams are derived from acquired projection images using scatter kernels indexed by the local WET values and are then iteratively refined using a scatter magnitude bounding scheme that allows the algorithm to accommodate the very high scatter-to-primary ratios encountered in kV imaging. The algorithm recovers radiological thicknesses to within 9% of the true value at both kV and megavolt energies. Nonuniformity in CT reconstructions of homogeneous phantoms is reduced by an average of 76% over a wide range of beam energies and phantom geometries.

On the selection of optimization parameters for an inverse treatment planning replacement of a forward planning technique for prostate cancer.

The influence of organ volume sampling, lateral scatter inclusion, and the selection of objectives and constraints on the inverse treatment planning process with a commercial treatment planning system is investigated and suitable parameters are identified for an inverse treatment planning replacement of a clinical forward planning technique for prostate cancer. For the beam geometries of the forward technique, a variable set of parameters is used for the calculation of dose from pencil beams. An optimal set is identified after the evaluation of optimized plans that correspond to different sets of pencil-beam parameters. This set along with a single, optimized set of objectives and constraints is used to perform inverse planning on ten randomly selected patients. The acceptability of the resulting plans is verified by comparisons to the clinical ones calculated with the forward techniques. For the particular commercial treatment planning system, the default values of the pencil beam parameters are found adequate for inverse treatment planning. For all ten patients, the optimized, single set of objectives and constraints results in plans with target coverage comparable to that of the forward plans. Furthermore inverse treatment planning reduces the overall mean rectal and bladder doses by 4.8% and 5.8% of the prescription dose respectively. The study indicates that (i) inverse treatment planning results depend implicitly on the sampling of the dose distribution, (ii) inverse treatment planning results depend on the method used by the dose calculation model to account for scatter, and (iii) for certain sites, a single set of optimization parameters can be used for all patient plans.