Building more accurate decision trees with the additive tree.

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.

FLASH Radiotherapy: What Can FLASH's Ultra High Dose Rate Offer to the Treatment of Patients With Sarcoma?

FLASH is an emerging treatment paradigm in radiotherapy (RT) that utilizes ultra-high dose rates (UHDR; >40 Gy)/s) of radiation delivery. Developing advances in technology support the delivery of UHDR using electron and proton systems, as well as some ion beam units (eg, carbon ions), while methods to achieve UHDR with photons are under investigation. The major advantage of FLASH RT is its ability to increase the therapeutic index for RT by shifting the dose response curve for normal tissue toxicity to higher doses. Numerous preclinical studies have been conducted to date on FLASH RT for murine sarcomas, alongside the investigation of its effects on relevant normal tissues of skin, muscle, and bone. The tumor control achieved by FLASH RT of sarcoma models is indistinguishable from that attained by treatment with standard RT to the same total dose. FLASH's high dose rates are able to mitigate the severity or incidence of RT side effects on normal tissues as evaluated by endpoints ranging from functional sparing to histological damage. Large animal studies and clinical trials of canine patients show evidence of skin sparing by FLASH vs. standard RT, but also caution against delivery of high single doses with FLASH that exceed those safely applied with standard RT. Also, a human clinical trial has shown that FLASH RT can be delivered safely to bone metastasis. Thus, data to date support continued investigations of clinical translation of FLASH RT for the treatment of patients with sarcoma. Toward this purpose, hypofractionated irradiation schemes are being investigated for FLASH effects on sarcoma and relevant normal tissues.

FLASH Proton Radiation Therapy Mitigates Inflammatory and Fibrotic Pathways and Preserves Cardiac Function in a Preclinical Mouse Model of Radiation-Induced Heart Disease.

PURPOSE: Studies during the past 9 years suggest that delivering radiation at dose rates exceeding 40 Gy/s, known as "FLASH" radiation therapy, enhances the therapeutic index of radiation therapy (RT) by decreasing normal tissue damage while maintaining tumor response compared with conventional (or standard) RT. This study demonstrates the cardioprotective benefits of FLASH proton RT (F-PRT) compared with standard (conventional) proton RT (S-PRT), as evidenced by reduced acute and chronic cardiac toxicities. METHODS AND MATERIALS: Mice were imaged using cone beam computed tomography to precisely determine the heart's apex as the beam isocenter. Irradiation was conducted using a shoot-through technique with a 5-mm diameter circular collimator. Bulk RNA-sequencing was performed on nonirradiated samples, as well as apexes treated with F-PRT or S-PRT, at 2 weeks after a single 40 Gy dose. Inflammatory responses were assessed through multiplex cytokine/chemokine microbead assay and immunofluorescence analyses. Levels of perivascular fibrosis were quantified using Masson's Trichrome and Picrosirius red staining. Additionally, cardiac tissue functionality was evaluated by 2-dimensional echocardiograms at 8- and 30-weeks post-PRT. RESULTS: Radiation damage was specifically localized to the heart's apex. RNA profiling of cardiac tissues treated with PRT revealed that S-PRT uniquely upregulated pathways associated with DNA damage response, induction of tumor necrosis factor superfamily, and inflammatory response, and F-PRT primarily affected cytoplasmic translation, mitochondrion organization, and adenosine triphosphate synthesis. Notably, F-PRT led to a milder inflammatory response, accompanied by significantly attenuated changes in transforming growth factor ß1 and α smooth muscle actin levels. Critically, F-PRT decreased collagen deposition and better preserved cardiac functionality compared with S-PRT. CONCLUSIONS: This study demonstrated that F-PRT reduces the induction of an inflammatory environment with lower expression of inflammatory cytokines and profibrotic factors. Importantly, the results indicate that F-PRT better preserves cardiac functionality, as confirmed by echocardiography analysis, while also mitigating the development of long-term fibrosis.

Proton FLASH Radiotherapy Ameliorates Radiation-induced Salivary Gland Dysfunction and Oral Mucositis and Increases Survival in a Mouse Model of Head and Neck Cancer.

Head and neck cancer radiotherapy often damages salivary glands and oral mucosa, severely negatively impacting patients' quality of life. The ability of FLASH proton radiotherapy (F-PRT) to decrease normal tissue toxicity while maintaining tumor control compared with standard proton radiotherapy (S-PRT) has been previously demonstrated for several tissues. However, its potential in ameliorating radiation-induced salivary gland dysfunction and oral mucositis and controlling orthotopic head and neck tumor growth has not been reported. The head and neck area of C57BL/6 mice was irradiated with a single dose of radiotherapy (ranging from 14-18 Gy) or a fractionated dose of 8 Gy × 3 of F-PRT (128 Gy/second) or S-PRT (0.95 Gy/second). Following irradiation, the mice were studied for radiation-induced xerostomia by measuring their salivary flow. Oral mucositis was analyzed by histopathologic examination. To determine the ability of F-PRT to control orthotopic head and neck tumors, tongue tumors were generated in the mice and then irradiated with either F-PRT or S-PRT. Mice treated with either a single dose or fractionated dose of F-PRT showed significantly improved survival than those irradiated with S-PRT. F-PRT-treated mice showed improvement in their salivary flow. S-PRT-irradiated mice demonstrated increased fibrosis in their tongue epithelium. F-PRT significantly increased the overall survival of the mice with orthotopic tumors compared with the S-PRT-treated mice. The demonstration that F-PRT decreases radiation-induced normal tissue toxicity without compromising tumor control, suggests that this modality could be useful for the clinical management of patients with head and neck cancer.

Unedited allogeneic iNKT cells show extended persistence in MHC-mismatched canine recipients.

Allogeneic invariant natural killer T cells (allo-iNKTs) induce clinical remission in patients with otherwise incurable cancers and COVID-19-related acute respiratory failure. However, their functionality is inconsistent among individuals, and they become rapidly undetectable after infusion, raising concerns over rejection and limited therapeutic potential. We validate a strategy to promote allo-iNKT persistence in dogs, an established large-animal model for novel cellular therapies. We identify donor-specific iNKT biomarkers of survival and sustained functionality, conserved in dogs and humans and retained upon chimeric antigen receptor engineering. We reason that infusing optimal allo-iNKTs enriched in these biomarkers will prolong their persistence without requiring MHC ablation, high-intensity chemotherapy, or cytokine supplementation. Optimal allo-iNKTs transferred into MHC-mismatched dogs remain detectable for at least 78 days, exhibiting sustained immunomodulatory effects. Our canine model will accelerate biomarker discovery of optimal allo-iNKT products, furthering application of MHC-unedited allo-iNKTs as a readily accessible universal platform to treat incurable conditions worldwide.

Framework for Quality Assurance of Ultrahigh Dose Rate Clinical Trials Investigating FLASH Effects and Current Technology Gaps.

FLASH radiation therapy (FLASH-RT), delivered with ultrahigh dose rate (UHDR), may allow patients to be treated with less normal tissue toxicity for a given tumor dose compared with currently used conventional dose rate. Clinical trials are being carried out and are needed to test whether this improved therapeutic ratio can be achieved clinically. During the clinical trials, quality assurance and credentialing of equipment and participating sites, particularly pertaining to UHDR-specific aspects, will be crucial for the validity of the outcomes of such trials. This report represents an initial framework proposed by the NRG Oncology Center for Innovation in Radiation Oncology FLASH working group on quality assurance of potential UHDR clinical trials and reviews current technology gaps to overcome. An important but separate consideration is the appropriate design of trials to most effectively answer clinical and scientific questions about FLASH. This paper begins with an overview of UHDR RT delivery methods. UHDR beam delivery parameters are then covered, with a focus on electron and proton modalities. The definition and control of safe UHDR beam delivery and current and needed dosimetry technologies are reviewed and discussed. System and site credentialing for large, multi-institution trials are reviewed. Quality assurance is then discussed, and new requirements are presented for treatment system standard analysis, patient positioning, and treatment planning. The tables and figures in this paper are meant to serve as reference points as we move toward FLASH-RT clinical trial performance. Some major questions regarding FLASH-RT are discussed, and next steps in this field are proposed. FLASH-RT has potential but is associated with significant risks and complexities. We need to redefine optimization to focus not only on the dose but also on the dose rate in a manner that is robust and understandable and that can be prescribed, validated, and confirmed in real time. Robust patient safety systems and access to treatment data will be critical as FLASH-RT moves into the clinical trials.

The current status of preclinical proton FLASH radiation and future directions.

We review the current status of proton FLASH experimental systems, including preclinical physical and biological results. Technological limitations on preclinical investigation of FLASH biological mechanisms and determination of clinically relevant parameters are discussed. A review of the biological data reveals no reproduced proton FLASH effect in vitro and a significant in vivo FLASH sparing effect of normal tissue toxicity observed with multiple proton FLASH irradiation systems. Importantly, multiple studies suggest little or no difference in tumor growth delay for proton FLASH when compared to conventional dose rate proton radiation. A discussion follows on future areas of development with a focus on the determination of the optimal parameters for maximizing the therapeutic ratio between tumor and normal tissue response and ultimately clinical translation of proton FLASH radiation.

Radiomic Phenotypes for Improving Early Prediction of Survival in Stage III Non-Small Cell Lung Cancer Adenocarcinoma after Chemoradiation.

We evaluate radiomic phenotypes derived from CT scans as early predictors of overall survival (OS) after chemoradiation in stage III primary lung adenocarcinoma. We retrospectively analyzed 110 thoracic CT scans acquired between April 2012-October 2018. Patients received a median radiation dose of 66.6 Gy at 1.8 Gy/fraction delivered with proton (55.5%) and photon (44.5%) beam treatment, as well as concurrent chemotherapy (89%) with carboplatin-based (55.5%) and cisplatin-based (36.4%) doublets. A total of 56 death events were recorded. Using manual tumor segmentations, 107 radiomic features were extracted. Feature harmonization using ComBat was performed to mitigate image heterogeneity due to the presence or lack of intravenous contrast material and variability in CT scanner vendors. A binary radiomic phenotype to predict OS was derived through the unsupervised hierarchical clustering of the first principal components explaining 85% of the variance of the radiomic features. C-scores and likelihood ratio tests (LRT) were used to compare the performance of a baseline Cox model based on ECOG status and age, with a model integrating the radiomic phenotype with such clinical predictors. The model integrating the radiomic phenotype (C-score = 0.69, 95% CI = (0.62, 0.77)) significantly improved (p<0.005) upon the baseline model (C-score = 0.65, CI = (0.57, 0.73)). Our results suggest that harmonized radiomic phenotypes can significantly improve OS prediction in stage III NSCLC after chemoradiation.

A phenomenological model of proton FLASH oxygen depletion effects depending on tissue vasculature and oxygen supply.

Introduction: Radiation-induced oxygen depletion in tissue is assumed as a contributor to the FLASH sparing effects. In this study, we simulated the heterogeneous oxygen depletion in the tissue surrounding the vessels and calculated the proton FLASH effective-dose-modifying factor (FEDMF), which could be used for biology-based treatment planning. Methods: The dose and dose-weighted linear energy transfer (LET) of a small animal proton irradiator was simulated with Monte Carlo simulation. We deployed a parabolic partial differential equation to account for the generalized radiation oxygen depletion, tissue oxygen diffusion, and metabolic processes to investigate oxygen distribution in 1D, 2D, and 3D solution space. Dose and dose rates, particle LET, vasculature spacing, and blood oxygen supplies were considered. Using a similar framework for the hypoxic reduction factor (HRF) developed previously, the FEDMF was derived as the ratio of the cumulative normoxic-equivalent dose (CNED) between CONV and UHDR deliveries. Results: Dynamic equilibrium between oxygen diffusion and tissue metabolism can result in tissue hypoxia. The hypoxic region displayed enhanced radio-resistance and resulted in lower CNED under UHDR deliveries. In 1D solution, comparing 15 Gy proton dose delivered at CONV 0.5 and UHDR 125 Gy/s, 61.5% of the tissue exhibited ≥20% FEDMF at 175 µm vasculature spacing and 18.9 µM boundary condition. This percentage reduced to 34.5% and 0% for 8 and 2 Gy deliveries, respectively. Similar trends were observed in the 3D solution space. The FLASH versus CONV differential effect remained at larger vasculature spacings. A higher FLASH dose rate showed an increased region with ≥20% FEDMF. A higher LET near the proton Bragg peak region did not appear to alter the FLASH effect. Conclusion: We developed 1D, 2D, and 3D oxygen depletion simulation process to obtain the dynamic HRF and derive the proton FEDMF related to the dose delivery parameters and the local tissue vasculature information. The phenomenological model can be used to simulate or predict FLASH effects based on tissue vasculature and oxygen concentration data obtained from other experiments.

Comparison of secondary neutron dose in proton therapy resulting from the use of a tungsten alloy MLC or a brass collimator system.

PURPOSE: To apply the dual ionization chamber method for mixed radiation fields to an accurate comparison of the secondary neutron dose arising from the use of a tungsten alloy multileaf collimator (MLC) as opposed to a brass collimator system for defining the shape of a therapeutic proton field. METHODS: Hydrogenous and nonhydrogenous ionization chambers were constructed with large volumes to enable measurements of absorbed doses below 10(-4) Gy in mixed radiation fields using the dual ionization chamber method for mixed-field dosimetry. Neutron dose measurements were made with a nominal 230 MeV proton beam incident on a closed tungsten alloy MLC and a solid brass block. The chambers were cross-calibrated against a (60)Co-calibrated Farmer chamber in water using a 6 MV x-ray beam and Monte Carlo simulations were performed to account for variations in ionization chamber response due to differences in secondary neutron energy spectra. RESULTS: The neutron and combined proton plus γ-ray absorbed doses are shown to be nearly equivalent downstream from either a closed tungsten alloy MLC or a solid brass block. At 10 cm downstream from the distal edge of the collimating material the neutron dose from the closed MLC was (5.3 ± 0.4) × 10(- 5) Gy/Gy. The neutron dose with brass was (6.4 ± 0.7) × 10(- 5) Gy/Gy. Further from the secondary neutron source, at 50 cm, the neutron doses remain close for both the MLC and brass block at (6.9 ± 0.6) × 10(- 6) Gy/Gy and (6.3 ± 0.7) × 10(- 6) Gy/Gy, respectively. CONCLUSIONS: The dual ionization chamber method is suitable for measuring secondary neutron doses resulting from proton irradiation. The results of measurements downstream from a closed tungsten alloy MLC and a brass block indicate that, even in an overly pessimistic worst-case scenario, secondary neutron production in a tungsten alloy MLC leads to absorbed doses that are nearly equivalent to those seen from brass collimators. Therefore, the choice of tungsten alloy in constructing the leaves of a proton MLC is appropriate, and does not lead to a substantial increase in the secondary neutron dose to the patient compared to that generated in a brass collimator.

Comparison of FLASH Proton Entrance and the Spread-Out Bragg Peak Dose Regions in the Sparing of Mouse Intestinal Crypts and in a Pancreatic Tumor Model.

Ultra-high dose rate FLASH proton radiotherapy (F-PRT) has been shown to reduce normal tissue toxicity compared to standard dose rate proton radiotherapy (S-PRT) in experiments using the entrance portion of the proton depth dose profile, while proton therapy uses a spread-out Bragg peak (SOBP) with unknown effects on FLASH toxicity sparing. To investigate, the biological effects of F-PRT using an SOBP and the entrance region were compared to S-PRT in mouse intestine. In this study, 8-10-week-old C57BL/6J mice underwent 15 Gy (absorbed dose) whole abdomen irradiation in four groups: (1) SOBP F-PRT, (2) SOBP S-PRT, (3) entrance F-PRT, and (4) entrance S-PRT. Mice were injected with EdU 3.5 days after irradiation, and jejunum segments were harvested and preserved. EdU-positive proliferating cells and regenerated intestinal crypts were quantified. The SOBP had a modulation (width) of 2.5 cm from the proximal to distal 90%. Dose rates with a SOBP for F-PRT or S-PRT were 108.2 ± 8.3 Gy/s or 0.82 ± 0.14 Gy/s, respectively. In the entrance region, dose rates were 107.1 ± 15.2 Gy/s and 0.83 ± 0.19 Gy/s, respectively. Both entrance and SOBP F-PRT preserved a significantly higher number of EdU + /crypt cells and percentage of regenerated crypts compared to S-PRT. Moreover, tumor growth studies showed no difference between SOBP and entrance for either of the treatment modalities.

Characterization of a high-resolution 2D transmission ion chamber for independent validation of proton pencil beam scanning of conventional and FLASH dose delivery.

INTRODUCTION: Ultra-high dose rate (FLASH) radiotherapy has become a popular research topic with the potential to reduce normal tissue toxicities without losing the benefit of tumor control. The development of FLASH proton pencil beam scanning (PBS) delivery requires accurate dosimetry despite high beam currents with correspondingly high ionization densities in the monitoring chamber. In this study, we characterized a newly designed high-resolution position sensing transmission ionization chamber with a purpose-built multichannel electrometer for both conventional and FLASH dose rate proton radiotherapy. METHODS: The dosimetry and positioning accuracies of the ion chamber were fully characterized with a clinical scanning beam. On the FLASH proton beamline, the cyclotron output current reached up to 350 nA with a maximum energy of 226.2 MeV, with 210 ± 3 nA nozzle pencil beam current. The ion recombination effect was characterized under various bias voltages up to 1000 V and different beam intensities. The charge collected by the transmission ion chamber was compared with the measurements from a Faraday cup. RESULTS: Cross-calibrated with an Advanced Markus chamber (PTW, Freiburg, Germany) in a uniform PBS proton beam field at clinical beam setting, the ion chamber calibration was 38.0 and 36.7 GyE·mm2 /nC at 100 and 226.2 MeV, respectively. The ion recombination effect increased with larger cyclotron current at lower bias voltage while remaining ≤0.5 ± 0.5% with ≥200 V of bias voltage. Above 200 V, the normalized ion chamber readings demonstrated good linearity with the mass stopping power in air for both clinical and FLASH beam intensities. The spot positioning accuracy was measured to be 0.10 ± 0.08 mm in two orthogonal directions. CONCLUSION: We characterized a transmission ion chamber system under both conventional and FLASH beam current densities and demonstrated its suitability for use as a proton pencil beam dose and spot position delivery monitor under FLASH dose rate conditions.

Current delivery limitations of proton PBS for FLASH.

PURPOSE: Proton Pencil Beam Scanning (PBS) is an attractive solution to realize the advantageous normal tissue sparing elucidated from FLASH high dose rates. The mechanics of PBS spot delivery will impose limitations on the effective field dose rate for PBS. METHODS: This study incorporates measurements from clinical and FLASH research beams on uniform single energy and the spread-out Bragg Peak PBS fields to extrapolate the PBS dose rate to high cyclotron beam currents 350, 500, and 800 nA. The impact of the effective field dose rate from cyclotron current, spot spacing, slew time and field size were studied. RESULTS: When scanning magnet slew time and energy switching time are not considered, single energy effective field FLASH dose rate (≥40 Gy/s) can only be achieved with less than 4 × 4 cm2 fields when the cyclotron output current is above 500 nA. Slew time and energy switching time remain the limiting factors for achieving high effective dose rate of the field. The dose rate-time structures were obtained. The amount of the total dose delivered at the FLASH dose rate in single energy layer and volumetric field was also studied. CONCLUSION: It is demonstrated that while it is difficult to achieve FLASH dose rate for a large field or in a volume, local FLASH delivery to certain percentage of the total dose is possible. With further understanding of the FLASH radiobiological mechanism, this study could provide guidance to adapt current clinical multi-field proton PBS delivery practice for FLASH proton radiotherapy.

Patient-derived organoids as a platform for modeling a patient's response to chemoradiotherapy in esophageal cancer.

3D patient-derived organoids (PDOs) have been utilized to evaluate potential therapies for patients with different cancers. However, the use of PDOs created from treatment-naive patient biopsies for prediction of clinical outcomes in patients with esophageal cancer has not yet been reported. Herein we describe a pilot prospective observational study with the goal of determining whether esophageal cancer PDOs created from treatment naive patients can model or predict clinical outcomes. Endoscopic biopsies of treatment-naive patients at a single tertiary care center were used to generate esophageal cancer PDOs, which were treated with standard-of-care chemotherapy, gamma-irradiation, and newer non-standard approaches, such as proton beam therapy or two small molecule inhibitors. Clinical outcomes of patients following neoadjuvant treatment were compared to their in vitro PDO responses, demonstrating the PDO's ability to mirror clinical response, suggesting the value of PDOs in prediction of clinical response to new therapeutic approaches. Future prospective clinical trials should test the use of pre-treatment PDOs to identify specific, targeted therapies for individual patients with esophageal adenocarcinoma.

Using electron beam radiation to simulate the dose distribution for whole body solar particle event proton exposure.

As a part of the near solar system exploration program, astronauts may receive significant total body proton radiation exposures during a solar particle event (SPE). In the Center for Acute Radiation Research (CARR), symptoms of the acute radiation sickness syndrome induced by conventional radiation are being compared to those induced by SPE-like proton radiation, to determine the relative biological effectiveness (RBE) of SPE protons. In an SPE, the astronaut's whole body will be exposed to radiation consisting mainly of protons with energies below 50 MeV. In addition to providing for a potentially higher RBE than conventional radiation, the energy distribution for an SPE will produce a relatively inhomogeneous total body dose distribution, with a significantly higher dose delivered to the skin and subcutaneous tissues than to the internal organs. These factors make it difficult to use a (60)Co standard for RBE comparisons in our experiments. Here, the novel concept of using megavoltage electron beam radiation to more accurately reproduce both the total dose and the dose distribution of SPE protons and make meaningful RBE comparisons between protons and conventional radiation is described. In these studies, Monte Carlo simulation was used to determine the dose distribution of electron beam radiation in small mammals such as mice and ferrets as well as large mammals such as pigs. These studies will help to better define the topography of the time-dose-fractionation versus biological response landscape for astronaut exposure to an SPE.

Machine learning highlights the deficiency of conventional dosimetric constraints for prevention of high-grade radiation esophagitis in non-small cell lung cancer treated with chemoradiation.

BACKGROUND AND PURPOSE: Radiation esophagitis is a clinically important toxicity seen with treatment for locally-advanced non-small cell lung cancer. There is considerable disagreement among prior studies in identifying predictors of radiation esophagitis. We apply machine learning algorithms to identify factors contributing to the development of radiation esophagitis to uncover previously unidentified criteria and more robust dosimetric factors. MATERIALS AND METHODS: We used machine learning approaches to identify predictors of grade ≥ 3 radiation esophagitis in a cohort of 202 consecutive locally-advanced non-small cell lung cancer patients treated with definitive chemoradiation from 2008 to 2016. We evaluated 35 clinical features per patient grouped into risk factors, comorbidities, imaging, stage, histology, radiotherapy, chemotherapy and dosimetry. Univariate and multivariate analyses were performed using a panel of 11 machine learning algorithms combined with predictive power assessments. RESULTS: All patients were treated to a median dose of 66.6 Gy at 1.8 Gy per fraction using photon (89.6%) and proton (10.4%) beam therapy, most often with concurrent chemotherapy (86.6%). 11.4% of patients developed grade ≥ 3 radiation esophagitis. On univariate analysis, no individual feature was found to predict radiation esophagitis (AUC range 0.45-0.55, p ≥ 0.07). In multivariate analysis, all machine learning algorithms exhibited poor predictive performance (AUC range 0.46-0.56, p ≥ 0.07). CONCLUSIONS: Contemporary machine learning algorithms applied to our modern, relatively large institutional cohort could not identify any reliable predictors of grade ≥ 3 radiation esophagitis. Additional patients are needed, and novel patient-specific and treatment characteristics should be investigated to develop clinically meaningful methods to mitigate this survival altering toxicity.

Design, Implementation, and in Vivo Validation of a Novel Proton FLASH Radiation Therapy System.

PURPOSE: Recent studies suggest that ultrahigh-dose-rate, "FLASH," electron radiation therapy (RT) decreases normal tissue damage while maintaining tumor response compared with conventional dose rate RT. Here, we describe a novel RT apparatus that delivers FLASH proton RT (PRT) using double scattered protons with computed tomography guidance and provide the first report of proton FLASH RT-mediated normal tissue radioprotection. METHODS AND MATERIALS: Absolute dose was measured at multiple depths in solid water and validated against an absolute integral charge measurement using a Faraday cup. Real-time dose rate was obtained using a NaI detector to measure prompt gamma rays. The effect of FLASH versus standard dose rate PRT on tumors and normal tissues was measured using pancreatic flank tumors (MH641905) derived from the KPC autochthonous PanCa model in syngeneic C57BL/6J mice with analysis of fibrosis and stem cell repopulation in small intestine after abdominal irradiation. RESULTS: The double scattering and collimation apparatus was dosimetrically validated with dose rates of 78 ± 9 Gy per second and 0.9 ± 0.08 Gy per second for the FLASH and standard PRT. Whole abdominal FLASH PRT at 15 Gy significantly reduced the loss of proliferating cells in intestinal crypts compared with standard PRT. Studies with local intestinal irradiation at 18 Gy revealed a reduction to near baseline levels of intestinal fibrosis for FLASH-PRT compared with standard PRT. Despite this difference, FLASH-PRT did not demonstrate tumor radioprotection in MH641905 pancreatic cancer flank tumors after 12 or 18 Gy irradiation. CONCLUSIONS: We have designed and dosimetrically validated a FLASH-PRT system with accurate control of beam flux on a millisecond time scale and online monitoring of the integral and dose delivery time structure. Using this system, we found that FLASH-PRT decreases acute cell loss and late fibrosis after whole-abdomen and focal intestinal RT, whereas tumor growth inhibition is preserved between the 2 modalities.

Predicting radiation pneumonitis in locally advanced stage II-III non-small cell lung cancer using machine learning.

BACKGROUND AND PURPOSE: Radiation pneumonitis (RP) is a radiotherapy dose-limiting toxicity for locally advanced non-small cell lung cancer (LA-NSCLC). Prior studies have proposed relevant dosimetric constraints to limit this toxicity. Using machine learning algorithms, we performed analyses of contributing factors in the development of RP to uncover previously unidentified criteria and elucidate the relative importance of individual factors. MATERIALS AND METHODS: We evaluated 32 clinical features per patient in a cohort of 203 stage II-III LA-NSCLC patients treated with definitive chemoradiation to a median dose of 66.6 Gy in 1.8 Gy daily fractions at our institution from 2008 to 2016. Of this cohort, 17.7% of patients developed grade ≥2 RP. Univariate analysis was performed using trained decision stumps to individually analyze statistically significant predictors of RP and perform feature selection. Applying Random Forest, we performed multivariate analysis to assess the combined performance of important predictors of RP. RESULTS: On univariate analysis, lung V20, lung mean, lung V10 and lung V5 were found to be significant RP predictors with the greatest balance of specificity and sensitivity. On multivariate analysis, Random Forest (AUC = 0.66, p = 0.0005) identified esophagus max (20.5%), lung V20 (16.4%), lung mean (15.7%) and pack-year (14.9%) as the most common primary differentiators of RP. CONCLUSIONS: We highlight Random Forest as an accurate machine learning method to identify known and new predictors of symptomatic RP. Furthermore, this analysis confirms the importance of lung V20, lung mean and pack-year as predictors of RP while also introducing esophagus max as an important RP predictor.

Design and commissioning of an image-guided small animal radiation platform and quality assurance protocol for integrated proton and x-ray radiobiology research.

Small animal x-ray irradiation platforms are expanding the capabilities and future pathways for radiobiology research. Meanwhile, proton radiotherapy is transitioning to a standard treatment modality in the clinician's precision radiotherapy toolbox, highlighting a gap between state-of-the-art clinical radiotherapy and small animal radiobiology research. Comparative research of the biological differences between proton and x-ray beams could benefit from an integrated small animal irradiation system for in vivo experiments and corresponding quality assurance (QA) protocols to ensure rigor and reproducibility. The objective of this study is to incorporate a proton beam into a small animal radiotherapy platform while implementing QA modelled after clinical protocols. A 225 kV x-ray small animal radiation research platform (SARRP) was installed on rails to align with a modified proton experimental beamline from a 230 MeV cyclotron-based clinical system. Collimated spread out Bragg peaks (SOBP) were produced with beam parameters compatible with small animal irradiation. Proton beam characteristics were measured and alignment reproducibility with the x-ray system isocenter was evaluated. A QA protocol was designed to ensure consistent proton beam quality and alignment. As a preliminary study, cellular damage via γ-H2AX immunofluorescence staining in an irradiated mouse tumor model was used to verify the beam range in vivo. The beam line was commissioned to deliver Bragg peaks with range 4-30 mm in water at 2 Gy min-1. SOBPs were delivered with width up to 25 mm. Proton beam alignment with the x-ray system agreed within 0.5 mm. A QA phantom was created to ensure reproducible alignment of the platform and verify beam delivery. γ-H2AX staining verified expected proton range in vivo. An image-guided small animal proton/x-ray research system was developed to enable in vivo investigations of radiobiological effects of proton beams, comparative studies between proton and x-ray beams, and investigations into novel proton treatment methods.