Gregory Arthur Szalkowski

All Publications

• NTCP model guided whole brain radiation re-planning to reduce risk of acute xerostomia and dry eye. Journal of applied clinical medical physics Szalkowski, G., Fenoli, J., Oakey, M., Tan, X., Pearlstein, K. A., Royce, T. J., Chera, B. S., Das, S. K., Wang, K., Mavroidis, P. 2025; 26 (12): e70344

  Abstract

  Previous work has shown that whole brain radiation (WBRT) can lead to acute xerostomia and dry eye from dose delivered to the parotid and lacrimal glands, respectively. We performed a retrospective study to assess whether a previously developed normal tissue complication probability (NTCP) model could guide planning to reduce the risk of these complications. We also evaluate if the use of VMAT/IMRT instead of 3D planning can reduce the risk of side effects while maintaining dose to the brain.We identified 11 patients who had previously received WBRT to 30 Gy in 10 fractions using 3D-conformal radiation therapy without prospective delineation of the parotid or lacrimal glands. For each patient, these structures were contoured and new 3D and IMRT plans were created to limit the V20 to the parotid glands and the V15 to the lacrimal glands while maintaining the dose to the brain. A previously developed relative seriality (RS) NTCP model was used to assess the reduction in xerostomia and dry eye risk relative to the original plan that was achieved with the new plans.The 3D re-plans significantly (p < 0.001) reduced the estimated risk of xerostomia, by 12.2 ± 4.5%, but did not significantly (p > 0.025) reduce the risk of dry eye. The IMRT re-plans significantly (p < 0.001) reduced the risk of xerostomia, by 20.6 ± 7.2%, and dry eye by 11.0 ± 3.8%. Both re-plans maintained target coverage.By using parameter values obtained from NTCP models, we were able to create whole brain plans that lowered the estimated risk of xerostomia and dry eye while maintaining target coverage.

  View details for DOI 10.1002/acm2.70344

  View details for PubMedID 41253707

• Single- versus multi-fraction spine stereotactic radiosurgery (ALL-STAR) for patients with spinal metastases: a randomized phase III trial protocol. BMC cancer Pratapneni, A., Klebaner, D., Soltys, S. G., Rahimy, E., Gibbs, I. C., Chang, S. D., Li, G., Hayden Gephart, M., Veeravagu, A., Szalkowski, G. A., Gu, X., Wang, L., Chuang, C., Liu, L., Jackson, S., Lu, R., Skerchak, J. A., Huang, K. Z., Wong, S., Brown, E., Pollom, E. L. 2025; 25 (1): 323

  Abstract

  For patients with spine metastases, stereotactic radiosurgery (SRS) provides excellent local control and pain response. Despite increasing use of this treatment modality, there is no consensus on the optimal dose and fractionation of spine SRS for efficacy and toxicity. We have initiated a single-center phase III randomized trial that compares two dose regimens with similar biological equivalent dose (BED) to determine the isolated effect of SRS fractionation on local control.Patients with one to three cervical, thoracic, or lumbar spine metastases spanning no more than two contiguous vertebral levels in need of radiation will be eligible for enrollment. Patients will be assigned 1:1 to receive either 22 Gy in 1 fraction or 28 Gy in 2 fractions. Biased coin randomization will be used to randomly assign patients while balancing the following stratifying variables between the two treatment arms at baseline: gastrointestinal histology (yes/no), paraspinal tissue extension (yes/no), epidural compression (low-/high-grade), and number of sites treated (one to three). The primary endpoint is one-year local control, defined per Spine Response Assessment in Neuro-Oncology (SPINO) criteria. The secondary endpoints include patient-reported health-related quality of life (HRQOL), pain associated with the treated site, vertebral compression fracture (VCF), and two-year local control. Patients will be followed for these outcomes at one to two weeks, one month, three months, and six months after treatment, and every six months thereafter until 24 months after treatment. While on the study, patients will receive routine co-interventions as clinically indicated.The studies published thus far comparing the single- and multi-fraction SRS are lacking long-term local control outcomes and are limited by selection bias as well as single-fraction arms with higher BED, which is correlated with improved local control. Our study will isolate the effect of fractionation by comparing one-year local control in patients treated with single- and multi-fraction SRS with equivalent BED. We anticipate that the results of this, as well as secondary endpoints such as pain response, adverse effects, and quality of life will provide much-needed guidance regarding optimal dose and fractionation for both maximizing local control and minimizing toxicity.NCT#06173401. Approved by Stanford Scientific Review Committee (study ID: BRN0060) on 9/12/2023 and Stanford Institutional Review Board (study ID: IRB-72248) on 11/14/2023.

  View details for DOI 10.1186/s12885-025-13655-6

  View details for PubMedID 39984889

  View details for PubMedCentralID PMC11846292

• Use of Carbon Fiber Implants to Improve the Safety and Efficacy of Radiation Therapy for Spine Tumor Patients. Brain sciences Lam, F. C., Guru, S., AbuReesh, D., Hori, Y. S., Chuang, C., Liu, L., Wang, L., Gu, X., Szalkowski, G. A., Wang, Z., Wohlers, C., Tayag, A., Emrich, S. C., Ustrzynski, L., Zygourakis, C. C., Desai, A., Hayden Gephart, M., Byun, J., Pollom, E. L., Rahimy, E., Soltys, S., Park, D. J., Chang, S. D. 2025; 15 (2)

  Abstract

  Current standard of care treatment for patients with spine tumors includes multidisciplinary approaches, including the following: (1) surgical tumor debulking, epidural spinal cord decompression, and spine stabilization techniques; (2) systemic chemo/targeted therapies; (3) radiation therapy; and (4) surveillance imaging for local disease control and recurrence. Titanium pedicle screw and rod fixation have become commonplace in the spine surgeon's armamentarium for the stabilization of the spine following tumor resection and separation surgery. However, the high degree of imaging artifacts seen with titanium implants on postoperative CT and MRI scans can significantly hinder the accurate delineation of vertebral anatomy and adjacent neurovascular structures to allow for the safe and effective planning of downstream radiation therapies and detection of disease recurrence. Carbon fiber-reinforced polyetheretherketone (CFR-PEEK) spine implants have emerged as a promising alternative to titanium due to the lack of artifact signals on CT and MRI, allowing for more accurate and safe postoperative radiation planning. In this article, we review the tenants of the surgical and radiation management of spine tumors and discuss the safety, efficacy, and current limitations of CFR-PEEK spine implants in the multidisciplinary management of spine oncology patients.

  View details for DOI 10.3390/brainsci15020199

  View details for PubMedID 40002531

  View details for PubMedCentralID PMC11852773

• Efficient and accurate commissioning and quality assurance of radiosurgery beam via prior-embedded implicit neural representation learning. Medical physics Liu, L., Chang, C., Wang, L., Gu, X., Szalkowski, G., Xing, L. 2025

  Abstract

  Dosimetric commissioning and quality assurance (QA) for linear accelerators (LINACs) present a significant challenge for clinical physicists due to the high measurement workload and stringent precision standards. This challenge is exacerbated for radiosurgery LINACs because of increased measurement uncertainty and more demanding setup accuracy for small-field beams. Optimizing physicists' effort during beam measurements while ensuring the quality of the measured data is crucial for clinical efficiency and patient safety.To develop a radiosurgery LINAC beam model that embeds prior knowledge of beam data through implicit neural representation (NeRP) learning and to evaluate the model's effectiveness in guiding beam data sampling, predicting complete beam dataset from sparse samples, and verifying detector choice and setup during commissioning and QA.Beam data including lateral profile and tissue-phantom-ratio (TPR), collected from CyberKnife LINACs, were investigated. Multi-layer perceptron (MLP) neural networks were optimized to parameterize a continuous function of the beam data, implicitly defined by the mapping from measurement coordinates to measured dose values. Beam priors were embedded into network weights by first training the network to learn the NeRP of a vendor-provided reference dataset. The prior-embedded network was further fine-tuned with sparse clinical measurements and used to predict unacquired beam data. Prospective and retrospective evaluations of different beam data samples in finetuning the model were performed using the reference beam dataset and clinical testing datasets, respectively. Model prediction accuracy was evaluated over 10 clinical datasets collected from various LINACs with different manufacturing modes and collimation systems. Model sensitivity in detecting beam data acquisition errors including inaccurate detector positioning and inappropriate detector choice was evaluated using two additional datasets with intentionally introduced erroneous samples.Prospective and retrospective evaluations identified consistent beam data samples that are most effective in fine-tuning the model for complete beam data prediction. Despite of discrepancies between clinical beam and the reference beam, fine-tuning the model with sparse beam profile measured at a single depth or with beam TPR measured at a single collimator size predicted beam data that closely match ground truth water tank measurements. Across the 10 clinical beam datasets, the averaged mean absolute error (MAE) in percentage dose was lower than 0.5% and the averaged 1D Gamma passing rate (1%/0.5  mm for profile and 1%/1  mm for TPR) was higher than 99%. In contrast, the MAE and Gamma passing rates were above 1% and below 95% between the reference beam dataset and clinical beam datasets. Model sensitivity to beam data acquisition errors was demonstrated by significant model prediction changes when fine-tuned with erroneous versus correct beam data samples, as quantified by a Gamma passing rate as low as 18.16% between model predictions.A model for small-field radiosurgery beam was proposed that embeds prior knowledge of beam properties and predicts the entire beam data from sparse measurements. The model can serve as a valuable tool for clinical physicists to verify the accuracy of beam data acquisition and promises to improve commissioning and QA reliability and efficiency with substantially reduced number of beam measurements.

  View details for DOI 10.1002/mp.17617

  View details for PubMedID 39812551

• Automatic Treatment Planning for Radiation Therapy: A Cross-Modality and Protocol Study ADVANCES IN RADIATION ONCOLOGY Szalkowski, G., Xu, X., Das, S., Yap, P., Lian, J. 2024; 9 (12)

  View details for DOI 10.1016/j.adro.2024.101649

  View details for Web of Science ID 001351355300001

• Automatic Treatment Planning for Radiation Therapy: A Cross-Modality and Protocol Study. Advances in radiation oncology Szalkowski, G., Xu, X., Das, S., Yap, P. T., Lian, J. 2024; 9 (12): 101649

  Abstract

  This study investigated the applicability of 3-dimensional dose predictions from a model trained on one modality to a cross-modality automated planning workflow. Additionally, we explore the impact of integrating a multicriteria optimizer (MCO) on adapting predictions to different clinical preferences.Using a previously created 3-stage U-Net in-house model trained on the 2020 American Association of Physicists in Medicine OpenKBP challenge data set (340 head and neck plans, all planned using 9-field static intensity modulated radiation therapy [IMRT]), we retrospectively generated dose predictions for 20 patients. These dose predictions were, in turn, used to generate deliverable IMRT, VMAT, and tomotherapy plans using the fallback plan functionality in Raystation. The deliverable plans were evaluated against the dose predictions based on primary clinical goals. A new set of plans was also generated using MCO-based optimization with predicted dose values as constraints. Delivery QA was performed on a subset of the plans to assure clinical deliverability.The mimicking approach accurately replicated the predicted dose distributions across different modalities, with slight deviations in the spinal cord and external contour maximum doses. MCO optimization significantly reduced doses to organs at risk, which were prioritized by our institution while maintaining target coverage. All tested plans met clinical deliverability standards, evidenced by a gamma analysis passing rate >98%.Our findings show that a model trained only on IMRT plans can effectively contribute to planning across various modalities. Additionally, integrating predictions as constraints in an MCO-based workflow, rather than direct dose mimicking, enables a flexible, warm-start approach for treatment planning, although the benefit is reduced when the training set differs significantly from an institution's preference. Together, these approaches have the potential to significantly decrease plan turnaround time and quality variance, both at high-resource medical centers that can train in-house models and smaller centers that can adapt a model from another institution with minimal effort.

  View details for DOI 10.1016/j.adro.2024.101649

  View details for PubMedID 39553397

  View details for PubMedCentralID PMC11566342

• Where Does Auto-Segmentation for Brain Metastases Radiosurgery Stand Today? Bioengineering (Basel, Switzerland) Kim, M., Wang, J. Y., Lu, W., Jiang, H., Stojadinovic, S., Wardak, Z., Dan, T., Timmerman, R., Wang, L., Chuang, C., Szalkowski, G., Liu, L., Pollom, E., Rahimy, E., Soltys, S., Chen, M., Gu, X. 2024; 11 (5)

  Abstract

  Detection and segmentation of brain metastases (BMs) play a pivotal role in diagnosis, treatment planning, and follow-up evaluations for effective BM management. Given the rising prevalence of BM cases and its predominantly multiple onsets, automated segmentation is becoming necessary in stereotactic radiosurgery. It not only alleviates the clinician's manual workload and improves clinical workflow efficiency but also ensures treatment safety, ultimately improving patient care. Recent strides in machine learning, particularly in deep learning (DL), have revolutionized medical image segmentation, achieving state-of-the-art results. This review aims to analyze auto-segmentation strategies, characterize the utilized data, and assess the performance of cutting-edge BM segmentation methodologies. Additionally, we delve into the challenges confronting BM segmentation and share insights gleaned from our algorithmic and clinical implementation experiences.

  View details for DOI 10.3390/bioengineering11050454

  View details for PubMedID 38790322

• Stereotactic radiosurgery for sarcoma metastases to the brain: a single-institution experience. Neurosurgical focus Zamarud, A., Park, D. J., Dadey, D. Y., Yoo, K. H., Marianayagam, N. J., Yener, U., Szalkowski, G. A., Pollom, E., Soltys, S., Chang, S. D., Meola, A. 2023; 55 (2): E7

  Abstract

  Brain metastases (BMs) secondary to sarcoma are rare, and their incidence ranges from 1% to 8% of all bone and soft tissue sarcomas. Although stereotactic radiosurgery (SRS) is widely used for BMs, only a few papers have reported on SRS for sarcoma metastasizing to the brain. The purpose of this study was to evaluate the safety and effectiveness of SRS for sarcoma BM.The authors retrospectively reviewed the clinical and radiological outcomes of patients with BM secondary to histopathologically confirmed sarcoma treated with SRS, either as primary treatment or as adjuvant therapy after surgery, at their institution between January 2005 and September 2022. They also compared the outcomes of patients with hemorrhagic lesions and of those without.Twenty-three patients (9 females) with 150 BMs secondary to sarcoma were treated with CyberKnife SRS. Median age at the time of treatment was 48.22 years (range 4-76 years). The most common primary tumor sites were the heart, lungs, uterus, upper extremities, chest wall, and head and neck. The median Karnofsky Performance Status on presentation was 73.28 (range 40-100). Eight patients underwent SRS as a primary treatment and 15 as adjuvant therapy to the resection cavity. The median tumor volume was 24.1 cm3 (range 0.1-150.3 cm3), the median marginal dose was 24 Gy (range 18-30 Gy) delivered in a median of 1 fraction (range 1-5) to a median isodose line of 76%. The median follow-up was 8 months (range 2-40 months). Median progression-free survival and overall survival were 5.3 months (range 0.4-32 months) and 8.2 months (range 0.1-40), respectively. The 3-, 6-, and 12-month local tumor control (LTC) rates for all lesions were respectively 78%, 52%, and 30%. There were no radiation-induced adverse effects. LTC at the 3-, 6-, and 12-month follow-ups was better in patients without hemorrhagic lesions (100%, 70%, and 40%, respectively) than in those with hemorrhagic lesions (68%, 38%, and 23%, respectively).SRS, both as a primary treatment and as adjuvant therapy to the resection cavity after surgery, is a safe and relatively effective treatment modality for sarcoma BMs. Nonhemorrhagic lesions show better LTC than hemorrhagic lesions. Larger studies aiming to validate these results are encouraged.

  View details for DOI 10.3171/2023.5.FOCUS23168

  View details for PubMedID 37527671

• Stereotactic body radiotherapy optimization to reduce the risk of carotid blowout syndrome using normal tissue complication probability objectives JOURNAL OF APPLIED CLINICAL MEDICAL PHYSICS Szalkowski, G., Karakas, Z., Cengiz, M., Schreiber, E., Das, S., Yazici, G., Ozyigit, G., Mavroidis, P. 2022; 23 (5): e13563

  Abstract

  To determine the possibility of further improving clinical stereotactic body radiotherapy (SBRT) plans using normal tissue complication probability (NTCP) objectives in order to minimize the risk for carotid blowout syndrome (CBOS).10 patients with inoperable locally recurrent head and neck cancer, who underwent SBRT using CyberKnife were analyzed. For each patient, three treatment plans were examined: (1) cone-based without delineation of the ipsilateral internal carotid (clinical plan used to treat the patients); (2) cone-based with the carotid retrospectively delineated and spared; and (3) Iris-based with carotid sparing. The dose-volume histograms of the target and primary organs at risk were calculated. The three sets of plans were compared based on dosimetric and TCP/NTCP (tumor control and normal tissue complication probabilities) metrics. For the NTCP values of carotid, the relative seriality model was used with the following parameters: D50 = 40 Gy, γ = 0.75, and s = 1.0.Across the 10 patient plans, the average TCP did not significantly change when the plans were re-optimized to spare the carotid. The estimated risk of CBOS was significantly decreased in the re-optimized plans, by 14.9% ± 7.4% for the cone-based plans and 17.7% ± 7.1% for the iris-based plans (p = 0.002 for both). The iris-based plans had significant (p = 0.02) reduced CBOS risk and delivery time (20.1% ± 7.4% time reduction, p = 0.002) compared to the cone-based plans.A significant improvement in the quality of the clinical plans could be achieved through the delineation of the internal carotids and the use of more modern treatment delivery modalities. In this way, for the same target coverage, a significant reduction in the risk of CBOS could be achieved. The range of risk reduction varied depending on the proximity of carotid artery to the target.

  View details for DOI 10.1002/acm2.13563

  View details for Web of Science ID 000759473900001

  View details for PubMedID 35194924

  View details for PubMedCentralID PMC9121056

• Synthetic digital reconstructed radiographs for MR-only robotic stereotactic radiation therapy: A proof of concept COMPUTERS IN BIOLOGY AND MEDICINE Szalkowski, G., Nie, D., Zhu, T., Yap, P., Lian, J. 2021; 138: 104917

  Abstract

  To create synthetic CTs and digital reconstructed radiographs (DRRs) from MR images that allow for fiducial visualization and accurate dose calculation for MR-only radiosurgery.We developed a machine learning model to create synthetic CTs from pelvic MRs for prostate treatments. This model has been previously proven to generate synthetic CTs with accuracy on par or better than alternate methods, such as atlas-based registration. Our dataset consisted of 11 paired CT and conventional MR (T2) images used for previous CyberKnife (Accuray, Inc) radiotherapy treatments. The MR images were pre-processed to mimic the appearance of fiducial-enhancing images. Two models were trained for each parameter case, using a sub-set of the available image pairs, with the remaining images set aside for testing and validation of the model to identify the optimal patch size and number of image pairs used for training. Four models were then trained using the identified parameters and used to generate synthetic CTs, which in turn were used to generate DRRs at angles 45° and 315°, as would be used for a CyberKnife treatment. The synthetic CTs and DRRs were compared visually and using the mean squared error and peak signal-to-noise ratio against the ground-truth images to evaluate their similarity.The synthetic CTs, as well as the DRRs generated from them, gave similar visualization of the fiducial markers in the prostate as the true counterparts. There was no significant difference found for the fiducial localization for the CTs and DRRs. Across the 8 DRRs analyzed, the mean MSE between the normalized true and synthetic DRRs was 0.66 ± 0.42% and the mean PSNR for this region was 22.9 ± 3.7 dB. For the full CTs, the mean MAE was 72.9 ± 88.1 HU and the mean PSNR was 31.2 ± 2.2 dB.Our machine learning-based method provides a proof of concept of a way to generate synthetic CTs and DRRs for accurate dose calculation and fiducial localization for use in radiation treatment of the prostate.

  View details for DOI 10.1016/j.compbiomed.2021.104917

  View details for Web of Science ID 000710608100007

  View details for PubMedID 34688037

  View details for PubMedCentralID PMC8627784

• Feasibility Study of Cross-Modality IMRT Auto-Planning Guided by a Deep Learning Model Szalkowski, G., Xu, X., Das, S., Yap, P., Lian, J. WILEY. 2021

  View details for Web of Science ID 000673145402246

• Image Synthesis for Planning and Target Tracking of MR-Based Stereotactic Radiation Therapy Szalkowski, G., Nie, D., Zhu, T., Yap, P., Lian, J. WILEY. 2021

  View details for Web of Science ID 000673145400106

• Optimization of hexagonal-pattern minibeams for spatially fractionated radiotherapy using proton beam scanning MEDICAL PHYSICS Charyyev, S., Artz, M., Szalkowski, G., Chang Chih-Wei, Stanforth, A., Lin Liyong, Zhang Rongxiao, Wang, C. 2020; 47 (8): 3485-3495

  Abstract

  In this study, we investigated computationally and experimentally a hexagonal-pattern array of spatially fractionated proton minibeams produced by proton pencil beam scanning (PBS) technique. Spatial fractionation of dose delivery with millimeter or submillimeter beam size has proven to be a promising approach to significantly increase the normal tissue tolerance. Our goals are to obtain an optimized minibeam design and to show that it is feasible to implement the optimized minibeams at the existing proton clinics.An optimized minibeam arrangement is one that would produce high peak-to-valley dose ratios (PVDRs) in normal tissues and a PVDR approaching unity at the Bragg peak. Using Monte Carlo (MC) code TOPAS we simulated proton pencil beams that mimic those available at the existing proton therapy facilities and obtained a hexagonal-pattern array of minibeams by collimating the proton pencil beams through the 1-3 mm diameter pinholes of a collimator. We optimized the minibeam design by considering different combinations of parameters including collimator material and thickness (t), center-to-center (c-t-c) distance, and beam size. The optimized minibeam design was then evaluated for normal tissue sparing against the uniform pencil beam scanning (PBS) by calculating the therapeutic advantage (TA) in terms of cell survival fraction. Verification measurements using radiochromic films were performed at the Emory proton therapy center (EPTC).Optimized hexagonal-pattern minibeams having PVDRs of >10 at phantom surface and of >3 at depths up to 6 cm were achieved with 2 mm diameter modulated proton minibeams (with proton energies between 120 and 140 MeV) corresponding to a spread-out-Bragg-peak (SOBP) over the depth of 10-14 cm. The results of the film measurements agree with the MC results within 10%. The TA of the 2 mm minibeams against the uniform PBS is >3 from phantom surface to the depth of 5 cm and then smoothly drops to ~1.5 as it approaches the proximal edge of the SOBP. For 2 mm minibeams and 6 mm c-t-c distance, we delivered 1.72 Gy at SOBP for 7.2 × 7.2 × 4 cm3 volume in 48 s.We conclude that it is feasible to implement the optimized hexagonal-pattern 2 mm proton minibeam radiotherapy at the existing proton clinics, because desirable PVDRs and TAs are achievable and the treatment time is reasonable.

  View details for DOI 10.1002/mp.14192

  View details for Web of Science ID 000531381400001

  View details for PubMedID 32319098

• Computer-Aided Star Shot Analysis for Linac Quality Assurance Testing Szalkowski, G. A., Roper, J. TAYLOR & FRANCIS INC. 2019: 905-911

  View details for DOI 10.1080/00295450.2018.1533349

  View details for Web of Science ID 000472556500004

• Monte Carlo Study of Photon Minibeams Szalkowski, G., Wang, C., Charyyev, S. WILEY. 2018: E614

  View details for Web of Science ID 000434978004348

• Development of Proton Minibeams as New Form of GRID Radiotherapy Charyyev, S., Wang, C., Szalkowski, G. WILEY. 2018: E488

  View details for Web of Science ID 000434978003264

• Design of Faraday cup ion detectors built by thin film deposition NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH SECTION A-ACCELERATORS SPECTROMETERS DETECTORS AND ASSOCIATED EQUIPMENT Szalkowski, G. A., Darrow, D. S., Cecil, F. E. 2017; 848: 87-90

  View details for DOI 10.1016/j.nima.2016.12.007

  View details for Web of Science ID 000394627600012