The emerging role of Artificial Intelligence in proton therapy: A review.

Artificial intelligence (AI) has made a tremendous impact in the space of healthcare, and proton therapy is not an exception. Proton therapy has witnessed growing popularity in oncology over recent decades, and researchers are increasingly looking to develop AI and machine learning tools to aid in various steps of the treatment planning and delivery processes. This review delves into the emergent role of AI in proton therapy, evaluating its development, advantages, intended clinical contexts, and areas of application. Through the analysis of 76 studies, we aim to underscore the importance of AI applications in advancing proton therapy and to highlight their prospective influence on clinical practices.

Late-xerostomia prediction model based on 18F-FDG PET image biomarkers of the main salivary glands.

BACKGROUND AND PURPOSE: Recently, a comprehensive xerostomia prediction model was published, based on baseline xerostomia, mean dose to parotid glands (PG) and submandibular glands (SMG). Previously, PET imaging biomarkers (IBMs) of PG were shown to improve xerostomia prediction. Therefore, this study aimed to explore the potential improvement of the additional PET-IBMs from both PG and SMG to the recent comprehensive xerostomia prediction model (i.e., the reference model). MATERIALS AND METHODS: Totally, 540 head and neck cancer patients were split into training and validation cohorts. PET-IBMs from the PG and SMG, were selected using bootstrapped forward selection based on the reference model. The IBMs from both the PG and SMG with the highest selection frequency were added to the reference model, resulting in a PG-IBM model and a SMG-IBM model which were combined into a composite model. Model performance was assessed using the area under the curve (AUC). Likelihood ratio test compared the predictive performance between the reference model and models including IBMs. RESULTS: The final selected PET-IBMs were 90th percentile of the PG SUV and total energy of the SMG SUV. The additional two PET-IBMs in the composite model improved the predictive performance of the reference model significantly. The AUC of the reference model and the composite model were 0.67 and 0.69 in the training cohort, and 0.71 and 0.73 in the validation cohort, respectively. CONCLUSION: The composite model including two additional PET-IBMs from PG and SMG improved the predictive performance of the reference xerostomia model significantly, facilitating a more personalized prediction approach.

Feasibility of Monte Carlo dropout-based uncertainty maps to evaluate deep learning-based synthetic CTs for adaptive proton therapy.

BACKGROUND: Deep learning has shown promising results to generate MRI-based synthetic CTs and to enable accurate proton dose calculations on MRIs. For clinical implementation of synthetic CTs, quality assurance tools that verify their quality and reliability are required but still lacking. PURPOSE: This study aims to evaluate the predictive value of uncertainty maps generated with Monte Carlo dropout (MCD) for verifying proton dose calculations on deep-learning-based synthetic CTs (sCTs) derived from MRIs in online adaptive proton therapy. METHODS: Two deep-learning models (DCNN and cycleGAN) were trained for CT image synthesis using 101 paired CT-MR images. sCT images were generated using MCD for each model by performing 10 inferences with activated dropout layers. The final sCT was obtained by averaging the inferred sCTs, while the uncertainty map was obtained from the HU variance corresponding to each voxel of 10 sCTs. The resulting uncertainty maps were compared to the observed HU-, range-, WET-, and dose-error maps between the sCT and planning CT. For range and WET errors, the generated uncertainty maps were projected along the 90-degree angle. To evaluate the dose distribution, a mask based on the 5%-isodose curve was applied to only include voxels along the beam paths. Pearson's correlation coefficients were calculated to determine the correlation between the uncertainty maps and HUs, range, WET, and dose errors. To evaluate the dosimetric accuracy of synthetic CTs, clinical proton treatment plans were recalculated and compared to the pCTs RESULTS: Evaluation of the correlation showed an average of r = 0.92 ± 0.03 and r = 0.92 ± 0.03 for errors between uncertainty-HU, r = 0.66 ± 0.09 and r = 0.62 ± 0.06 between uncertainty-range, r = 0.64 ± 0.06 and r = 0.58 ± 0.07 between uncertainty-WET, and r = 0.65 ± 0.09 and r = 0.67 ± 0.07 between uncertainty and dose difference for DCNN and cycleGAN model, respectively. Dosimetric comparison for target volumes showed an average 3%/3 mm gamma pass rate of 99.76 ± 0.43 (DCNN) and 99.10 ± 1.27 (cycleGAN). CONCLUSION: The observed correlations between uncertainty maps and the various metrics (HU, range, WET, and dose errors) demonstrated the potential of MCD-based uncertainty maps as a reliable QA tool to evaluate the accuracy of deep learning-based sCTs.

Clinical outcome after pencil beam scanning proton therapy and dysphagia/xerostomia NTCP calculations of proton and photon radiotherapy delivered to patients with cancer of the major salivary glands.

OBJECTIVES: The purpose of this study is to report the oncological outcome, observed toxicities and normal tissue complication probability (NTCP) calculation for pencil beam scanning (PBS) PT delivered to salivary gland tumour (SGT) patients. METHODS: We retrospectively reviewed 26 SGT patients treated with PBSPT (median dose, 67.5 Gy(RBE)) between 2005 and 2020 at our institute. Toxicities were recorded according to CTCAEv.4.1. Overall survival (OS), local control (LC), locoregional control (LRC) and distant control (DC) were estimated. For all patients, a photon plan was re-calculated in order to assess the photon/proton NTCP. RESULTS: With a median follow-up time of 46 months (range, 3-118), 5 (19%), 2 (8%), 3 (12%) and 2 (8%) patients presented after PT with distant, local, locoregional failures and death, respectively. The estimated 4 year OS, LC, LCR and DC were 90%, 90%, 87 and 77%, respectively. Grade 3 late toxicity was observed in 2 (8%) patients. The estimated 4 year late high-grade (≥3) toxicity-free survival was 78.4%. The calculated mean difference of NTCP-values after PBSPT and VMAT plans for developing Grade 2 or 3 xerostomia were 3.8 and 2.9%, respectively. For Grade 2-3 dysphagia, the grade corresponding percentages were 8.6 and 1.9%. Not using an up-front model-based approach to select patients for PT, only 40% of our patients met the Dutch eligibility criteria. CONCLUSION: Our data suggest excellent oncological outcome and low late toxicity rates for patients with SGT treated with PBSPT. NTCP calculation showed a substantial risk reduction for Grade 2 or 3 xerostomia and dysphagia in some SGT patients, while for others, no clear benefit was seen with protons, suggesting that comparative planning should be performed routinely for these patients. ADVANCES IN KNOWLEDGE: We have reported that the clinical outcome of SGT patients treated with PT and compared IMPT to VMAT for the treatment of salivary gland tumour and have observed that protons delivered significantly less dose to organs at risks and were associated with less NTCP for xerostomia and dysphagia. Noteworthy, not using an up-front model-based approach, only 40% of our patients met the Dutch eligibility criteria.

Validation of the 18F-FDG PET image biomarker model predicting late xerostomia after head and neck cancer radiotherapy.

BACKGROUND AND PURPOSE: Previously, PET image biomarkers (PET-IBMs) - the 90th percentile standardized uptake value (P90-SUV) and the Mean SUV (Mean-SUV) of the contralateral parotid gland (cPG) - were identified as predictors for late-xerostomia following head and neck cancer (HNC) radiotherapy. The aim of the current study was to assess in an independent validation cohort whether these pre-treatment PET-IBM can improve late-xerostomia prediction compared to the prediction with baseline xerostomia and mean cPG dose alone. MATERIALS AND METHODS: The prediction endpoint was patient-rated moderate-to-severe xerostomia at 12 months after radiotherapy. The PET-IBMs were extracted from pre-treatment 18 F-FDG PET images. The performance of the model (base model) with baseline xerostomia and mean cPG dose alone and models with additionally P90-SUV or Mean-SUV were tested in the current independent validation cohort. Specifically, model discrimination (area under the curve: AUC) and calibration (calibration plot) were evaluated. RESULTS: The current validation cohort consisted of 137 patients of which 40% developed moderate-to-severe xerostomia at 12 months. Both the PET-P90 model (AUC:PET-P90 = 0.71) and the PET-Mean model (AUC: PET-Mean = 0.70) performed well in the current validation cohort. Moreover, their performance were improved compared to the base model (AUC:base model= 0.68). The calibration plots showed a good fit of the prediction to the actual rates for all tested models. CONCLUSION: PET-IBMs showed an improved prediction of late-xerostomia when added to the base model in this validation cohort. This contributed to the published hypothesis that PET-IBMs include individualized information and can serve as a pre-treatment risk factor for late-xerostomia.

Deep learning-based 4D-synthetic CTs from sparse-view CBCTs for dose calculations in adaptive proton therapy.

BACKGROUND: Time-resolved 4D cone beam-computed tomography (4D-CBCT) allows a daily assessment of patient anatomy and respiratory motion. However, 4D-CBCTs suffer from imaging artifacts that affect the CT number accuracy and prevent accurate proton dose calculations. Deep learning can be used to correct CT numbers and generate synthetic CTs (sCTs) that can enable CBCT-based proton dose calculations. PURPOSE: In this work, sparse view 4D-CBCTs were converted into 4D-sCT utilizing a deep convolutional neural network (DCNN). 4D-sCTs were evaluated in terms of image quality and dosimetric accuracy to determine if accurate proton dose calculations for adaptive proton therapy workflows of lung cancer patients are feasible. METHODS: A dataset of 45 thoracic cancer patients was utilized to train and evaluate a DCNN to generate 4D-sCTs, based on sparse view 4D-CBCTs reconstructed from projections acquired with a 3D acquisition protocol. Mean absolute error (MAE) and mean error were used as metrics to evaluate the image quality of single phases and average 4D-sCTs against 4D-CTs acquired on the same day. The dosimetric accuracy was checked globally (gamma analysis) and locally for target volumes and organs-at-risk (OARs) (lung, heart, and esophagus). Furthermore, 4D-sCTs were also compared to 3D-sCTs. To evaluate CT number accuracy, proton radiography simulations in 4D-sCT and 4D-CTs were compared in terms of range errors. The clinical suitability of 4D-sCTs was demonstrated by performing a 4D dose reconstruction using patient specific treatment delivery log files and breathing signals. RESULTS: 4D-sCTs resulted in average MAEs of 48.1 ± 6.5 HU (single phase) and 37.7 ± 6.2 HU (average). The global dosimetric evaluation showed gamma pass ratios of 92.3% ± 3.2% (single phase) and 94.4% ± 2.1% (average). The clinical target volume showed high agreement in D98 between 4D-CT and 4D-sCT, with differences below 2.4% for all patients. Larger dose differences were observed in mean doses of OARs (up to 8.4%). The comparison with 3D-sCTs showed no substantial image quality and dosimetric differences for the 4D-sCT average. Individual 4D-sCT phases showed slightly lower dosimetric accuracy. The range error evaluation revealed that lung tissues cause range errors about three times higher than the other tissues. CONCLUSION: In this study, we have investigated the accuracy of deep learning-based 4D-sCTs for daily dose calculations in adaptive proton therapy. Despite image quality differences between 4D-sCTs and 3D-sCTs, comparable dosimetric accuracy was observed globally and locally. Further improvement of 3D and 4D lung sCTs could be achieved by increasing CT number accuracy in lung tissues.

A Decision Support Tool to Optimize Selection of Head and Neck Cancer Patients for Proton Therapy.

Selection of head and neck cancer (HNC) patients for proton therapy (PT) using plan comparison (VMAT vs. IMPT) for each patient is labor-intensive. Our aim was to develop a decision support tool to identify patients with high probability to qualify for PT, at a very early stage (immediately after delineation) to avoid delay in treatment initiation. A total of 151 HNC patients were included, of which 106 (70%) patients qualified for PT. Linear regression models for individual OARs were created to predict the Dmean to the OARs for VMAT and IMPT plans. The predictors were OAR volume percentages overlapping with target volumes. Then, actual and predicted plan comparison decisions were compared. Actual and predicted OAR Dmean (VMAT R2 = 0.953, IMPT R2 = 0.975) and NTCP values (VMAT R2 = 0.986, IMPT R2 = 0.992) were highly correlated. The sensitivity, specificity, PPV and NPV of the decision support tool were 64%, 87%, 92% and 51%, respectively. The expected toxicity reduction with IMPT can be predicted using only the delineation data. The probability of qualifying for PT is >90% when the tool indicates a positive outcome for PT. This tool will contribute significantly to a more effective selection of HNC patients for PT at a much earlier stage, reducing treatment delay.

A two-stage genome-wide association study of radiation-induced acute toxicity in head and neck cancer.

BACKGROUND: Most head and neck cancer (HNC) patients receive radiotherapy (RT) and develop toxicities. This genome-wide association study (GWAS) was designed to identify single nucleotide polymorphisms (SNPs) associated with common acute radiation-induced toxicities (RITs) in an HNC cohort. METHODS: A two-stage GWAS was performed in 1279 HNC patients treated with RT and prospectively scored for mucositis, xerostomia, sticky saliva, and dysphagia. The area under the curve (AUC) was used to estimate the average load of toxicity during RT. At the discovery study, multivariate linear regression was used in 957 patients, and the top-ranking SNPs were tested in 322 independent replication cohort. Next, the discovery and the replication studies were meta-analyzed. RESULTS: A region on 5q21.3 containing 16 SNPs showed genome-wide (GW) significance association at P-value < 5.0 × 10-8 with patient-rated acute xerostomia in the discovery study. The top signal was rs35542 with an adjusted effect size of 0.17*A (95% CI 0.12 to 0.23; P-value < = 3.78 × 10-9). The genome wide significant SNPs were located within three genes (EFNA5, FBXL17, and FER). In-silico functional analysis showed these genes may be involved in DNA damage response and co-expressed in minor salivary glands. We found 428 suggestive SNPs (P-value < 1.0 × 10-5) for other toxicities, taken to the replication study. Eleven of them showed a nominal association (P-value < 0.05). CONCLUSIONS: This GWAS suggested novel SNPs for patient-rated acute xerostomia in HNC patients. If validated, these SNPs and their related functional pathways could lead to a predictive assay to identify sensitive patients to radiation, which may eventually allow a more individualized RT treatment.

Clinical suitability of deep learning based synthetic CTs for adaptive proton therapy of lung cancer.

PURPOSE: Adaptive proton therapy (APT) of lung cancer patients requires frequent volumetric imaging of diagnostic quality. Cone-beam CT (CBCT) can provide these daily images, but x-ray scattering limits CBCT-image quality and hampers dose calculation accuracy. The purpose of this study was to generate CBCT-based synthetic CTs using a deep convolutional neural network (DCNN) and investigate image quality and clinical suitability for proton dose calculations in lung cancer patients. METHODS: A dataset of 33 thoracic cancer patients, containing CBCTs, same-day repeat CTs (rCT), planning-CTs (pCTs), and clinical proton treatment plans, was used to train and evaluate a DCNN with and without a pCT-based correction method. Mean absolute error (MAE), mean error (ME), peak signal-to-noise ratio, and structural similarity were used to quantify image quality. The evaluation of clinical suitability was based on recalculation of clinical proton treatment plans. Gamma pass ratios, mean dose to target volumes and organs at risk, and normal tissue complication probabilities (NTCP) were calculated. Furthermore, proton radiography simulations were performed to assess the HU-accuracy of sCTs in terms of range errors. RESULTS: On average, sCTs without correction resulted in a MAE of 34 ± 6 HU and ME of 4 ± 8 HU. The correction reduced the MAE to 31 ± 4HU (ME to 2 ± 4HU). Average 3%/3 mm gamma pass ratios increased from 93.7% to 96.8%, when the correction was applied. The patient specific correction reduced mean proton range errors from 1.5 to 1.1 mm. Relative mean target dose differences between sCTs and rCT were below ± 0.5% for all patients and both synthetic CTs (with/without correction). NTCP values showed high agreement between sCTs and rCT (<2%). CONCLUSION: CBCT-based sCTs can enable accurate proton dose calculations for APT of lung cancer patients. The patient specific correction method increased the image quality and dosimetric accuracy but had only a limited influence on clinically relevant parameters.