The expansion of machine learning to high-stakes application domains such as medicine, finance, and criminal justice, where making informed decisions requires clear understanding of the model, has ...increased the interest in interpretable machine learning. The widely used Classification and Regression Trees (CART) have played a major role in health sciences, due to their simple and intuitive explanation of predictions. Ensemble methods like gradient boosting can improve the accuracy of decision trees, but at the expense of the interpretability of the generated model. Additive models, such as those produced by gradient boosting, and full interaction models, such as CART, have been investigated largely in isolation. We show that these models exist along a spectrum, revealing previously unseen connections between these approaches. This paper introduces a rigorous formalization for the additive tree, an empirically validated learning technique for creating a single decision tree, and shows that this method can produce models equivalent to CART or gradient boosted stumps at the extremes by varying a single parameter. Although the additive tree is designed primarily to provide both the model interpretability and predictive performance needed for high-stakes applications like medicine, it also can produce decision trees represented by hybrid models between CART and boosted stumps that can outperform either of these approaches.
Purpose:
To measure the acoustic signal generated by a pulsed proton spill from a hospital‐based clinical cyclotron.
Methods:
An electronic function generator modulated the IBA C230 isochronous ...cyclotron to create a pulsed proton beam. The acoustic emissions generated by the proton beam were measured in water using a hydrophone. The acoustic measurements were repeated with increasing proton current and increasing distance between detector and beam.
Results:
The cyclotron generated proton spills with rise times of 18 μs and a maximum measured instantaneous proton current of 790 nA. Acoustic emissions generated by the proton energy deposition were measured to be on the order of mPa. The origin of the acoustic wave was identified as the proton beam based on the correlation between acoustic emission arrival time and distance between the hydrophone and proton beam. The acoustic frequency spectrum peaked at 10 kHz, and the acoustic pressure amplitude increased monotonically with increasing proton current.
Conclusions:
The authors report the first observation of acoustic emissions generated by a proton beam from a hospital‐based clinical cyclotron. When modulated by an electronic function generator, the cyclotron is capable of creating proton spills with fast rise times (18 μs) and high instantaneous currents (790 nA). Measurements of the proton‐generated acoustic emissions in a clinical setting may provide a method for in vivo proton range verification and patient monitoring.
Purpose
This study suggests a lifelong learning‐based convolutional neural network (LL‐CNN) algorithm as a superior alternative to single‐task learning approaches for automatic segmentation of head ...and neck (OARs) organs at risk.
Methods and materials
Lifelong learning‐based convolutional neural network was trained on twelve head and neck OARs simultaneously using a multitask learning framework. Once the weights of the shared network were established, the final multitask convolutional layer was replaced by a single‐task convolutional layer. The single‐task transfer learning network was trained on each OAR separately with early stoppage. The accuracy of LL‐CNN was assessed based on Dice score and root‐mean‐square error (RMSE) compared to manually delineated contours set as the gold standard. LL‐CNN was compared with 2D‐UNet, 3D‐UNet, a single‐task CNN (ST‐CNN), and a pure multitask CNN (MT‐CNN). Training, validation, and testing followed Kaggle competition rules, where 160 patients were used for training, 20 were used for internal validation, and 20 in a separate test set were used to report final prediction accuracies.
Results
On average contours generated with LL‐CNN had higher Dice coefficients and lower RMSE than 2D‐UNet, 3D‐Unet, ST‐ CNN, and MT‐CNN. LL‐CNN required ~72 hrs to train using a distributed learning framework on 2 Nvidia 1080Ti graphics processing units. LL‐CNN required 20 s to predict all 12 OARs, which was approximately as fast as the fastest alternative methods with the exception of MT‐CNN.
Conclusions
This study demonstrated that for head and neck organs at risk, LL‐CNN achieves a prediction accuracy superior to all alternative algorithms.
Purpose
To determine whether individual liver tumor patients can be safely treated with pencil beam scanning proton therapy. This study reports a planning preparation workflow that can be used for ...beam angle selection and the evaluation of the efficacy of abdominal compression (AC) to mitigate motion.
Methods
Four‐dimensional computed tomography scans (4DCT) with and without AC were available from 10 liver tumor patients with fluoroscopy‐proven motion reduction by AC, previously treated using photons. For each scan, the motion amplitudes and the motion‐induced variation of water‐equivalent thickness (ΔWET) in each voxel of the target volume were evaluated during treatment plan preparation. Optimal proton beam angles were selected after volume analysis of the respective beam‐specific planning target volume (BSPTV). M⊥80 and ΔWET80 derived from the 80th percentiles of perpendicular motion amplitude (M⊥) and ΔWET were compared with and without AC. Proton plans were created on the average CT to achieve target coverage similar to that of the conventional photon treatments. 4D dynamic dose calculation was performed postplan by synchronizing proton beam delivery timing patterns to the 4DCT phases to assess interplay and fractionation effects, and to determine motion criteria for subsequent patient treatment.
Results
Selected coplanar beam angles ranged between 180° and 39°, primarily from right lateral (oblique) and posterior (oblique) directions. While AC produced a significant reduction in mean Liver‐GTV dose, any reduction in mean heart dose was patient dependent and not significant. Similarly, AC resulted in reductions in M⊥, ΔWET, and BSPTV volumes and improved dose degradation (ΔD95 and ΔD1) within the CTV. For small motion (M⊥80 < 7 mm and ΔWET80 < 5 mm), motion mitigation was not needed. For moderate motion (M⊥80 7–10 mm or ΔWET80 5–7 mm), AC produced a modest improvement. For large motion (M⊥80 > 10 mm or ΔWET80 > 7 mm), AC and/or some other form of mitigation strategies were required.
Conclusion
A workflow for screening patients’ motion characteristics and optimizing beam angle selection was established for the pencil beam scanning proton therapy treatment of liver tumors. Abdominal compression was found to be useful at mitigation of moderate and large motion.
For lung tumors with large motion amplitudes, the use of proton pencil beam scanning (PBS) can produce large dose errors. In this study, we assess under what circumstances PBS can be used to treat ...lung cancer patients who exhibit large tumor motion, based on the quantification of tumor motion and the dose interplay.
PBS plans were optimized on average 4DCT datasets using a beam-specific PTV method for 10 consecutive patients with locally advanced non-small-cell-lung-cancer (NSCLC) treated with proton therapy to 6660/180 cGy. End inhalation (CT0) and end exhalation (CT50) were selected as the two extreme scenarios to acquire the relative stopping power ratio difference (Δrsp) for a respiration cycle. The water equivalent difference (ΔWET) per radiological path was calculated from the surface of patient to the iCTV by integrating the Δrsp of each voxel. The magnitude of motion of voxels within the target follows a quasi-Gaussian distribution. A motion index (MI (>5mm WET)), defined as the percentage of target voxels with an absolute integral ΔWET larger than 5 mm, was adopted as a metric to characterize interplay. To simulate the treatment process, 4D dose was calculated by accumulating the spot dose on the corresponding respiration phase to the reference phase CT50 by deformable image registration based on spot timing and patient breathing phase.
The study indicated that the magnitude of target underdose in a single fraction plan is proportional to the MI (p < .001), with larger motion equating to greater dose degradation and standard deviations. The target homogeneity, minimum, maximum and mean dose in the 4D dose accumulations of 37 fractions varied as a function of MI.
This study demonstrated that MI can predict the level of dose degradation, which potentially serves as a clinical decision tool to assess whether lung cancer patients are potentially suitable to receive PBS treatment.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, UILJ, UKNU, UL, UM, UPUK
The goal of this study is to demonstrate the feasibility of a novel non-coplanar-arc optimization algorithm (CyberArc). This method aims to reduce the delivery time of conventional CyberKnife ...treatments by allowing for continuous beam delivery. CyberArc uses a 4 step optimization strategy, in which nodes, beams, and collimator sizes are determined, source trajectories are calculated, intermediate radiation models are generated, and final monitor units are calculated, for the continuous radiation source model. The dosimetric results as well as the time reduction factors for CyberArc are presented for 7 prostate and 2 brain cases. The dosimetric quality of the CyberArc plans are evaluated using conformity index, heterogeneity index, local confined normalized-mutual-information, and various clinically relevant dosimetric parameters. The results indicate that the CyberArc algorithm dramatically reduces the treatment time of CyberKnife plans while simultaneously preserving the dosimetric quality of the original plans.
Purpose
This study aims to reduce the delivery time of CyberKnife m6 treatments by allowing for noncoplanar continuous arc delivery. To achieve this, a novel noncoplanar continuous arc delivery ...optimization algorithm was developed for the CyberKnife m6 treatment system (CyberArc‐m6).
Methods and Materials
CyberArc‐m6 uses a five‐step overarching strategy, in which an initial set of beam geometries is determined, the robotic delivery path is calculated, direct aperture optimization is conducted, intermediate MLC configurations are extracted, and the final beam weights are computed for the continuous arc radiation source model. This algorithm was implemented on five prostate and three brain patients, previously planned using a conventional step‐and‐shoot CyberKnife m6 delivery technique. The dosimetric quality of the CyberArc‐m6 plans was assessed using locally confined mutual information (LCMI), conformity index (CI), heterogeneity index (HI), and a variety of common clinical dosimetric objectives.
Results
Using conservative optimization tuning parameters, CyberArc‐m6 plans were able to achieve an average CI difference of 0.036 ± 0.025, an average HI difference of 0.046 ± 0.038, and an average LCMI of 0.920 ± 0.030 compared with the original CyberKnife m6 plans. Including a 5 s per minute image alignment time and a 5‐min setup time, conservative CyberArc‐m6 plans achieved an average treatment delivery speed up of 1.545x ± 0.305x compared with step‐and‐shoot plans.
Conclusions
The CyberArc‐m6 algorithm was able to achieve dosimetrically similar plans compared to their step‐and‐shoot CyberKnife m6 counterparts, while simultaneously reducing treatment delivery times.
Dose calculation for pencil beam scanning proton therapy requires accurate measurement of the broad tails of the proton spot profiles for every nozzle in clinical use. By applying a pair ...magnification method and merging film data, 200 mm × 240 mm dose kernels extending to 10−4 of the central spot dose are generated for six selected energies of the IBA dedicated and universal nozzles (DN and UN). One-dimensional, circular profiles up to 100 mm in radius are generated from the asymmetric profiles to facilitate spot profile comparison. For the highest energy, 225 MeV, the output of both the DN and the UN for field sizes from 40 to 200 mm increases in parallel, slowest at the surface (∼1%) and fastest at a depth of 150 mm (∼9%). In contrast, at the lowest energy, 100 MeV, the output of the DN across the same range of field sizes increases 3-4% versus 6-7% for the UN throughout all the depths. The charge deficits in the measured depth-dose of Bragg peaks are similar between the UN and the DN. At 100 MeV, the field size factor difference at the surface between two orientations of a rectangular 40 mm × 200 mm field is 1.4% at isocentre for the DN versus 2% for the UN. Though the one-dimensional distributions are similar for the primary and tail components at different positions, the primary components of the DN spots are more elliptical 270 mm upstream than at isocentre.
We investigated prognostic models based on clinical, radiologic, and radiomic feature to preoperatively identify meningiomas at risk for poor outcomes.
Retrospective review was performed for 303 ...patients who underwent resection of 314 meningiomas (57% World Health Organization grade I, 35% grade II, and 8% grade III) at two independent institutions, which comprised primary and external datasets. For each patient in the primary dataset, 16 radiologic and 172 radiomic features were extracted from preoperative magnetic resonance images, and prognostic features for grade, local failure (LF) or overall survival (OS) were identified using the Kaplan-Meier method, log-rank tests and recursive partitioning analysis. Regressions and random forests were used to generate and test prognostic models, which were validated using the external dataset.
Multivariate analysis revealed that apparent diffusion coefficient hypointensity (HR 5.56, 95% CI 2.01-16.7,
= .002) was associated with high grade meningioma, and low sphericity was associated both with increased LF (HR 2.0, 95% CI 1.1-3.5,
= .02) and worse OS (HR 2.94, 95% CI 1.47-5.56,
= .002). Both radiologic and radiomic predictors of adverse meningioma outcomes were significantly associated with molecular markers of aggressive meningioma biology, such as somatic mutation burden, DNA methylation status, and
expression. Integrated prognostic models combining clinical, radiologic, and radiomic features demonstrated improved accuracy for meningioma grade, LF, and OS (area under the curve 0.78, 0.75, and 0.78, respectively) compared to models based on clinical features alone.
Preoperative radiologic and radiomic features such as apparent diffusion coefficient and sphericity can predict tumor grade, LF, and OS in patients with meningioma.