NTCP Models Research Articles

Comparison of deep-learning multimodality data fusion strategies in mandibular osteoradionecrosis NTCP modelling using clinical variables and radiation dose distribution volumes

Objective.Normal tissue complication probability (NTCP) modelling is rapidly embracing deep learning (DL) methods, acknowledging the importance of spatial dose information. Finding effective ways to combine information from radiation dose distribution maps (dosiomics) and clinical data involves technical challenges and requires domain knowledge. We propose different multi-modality data fusion strategies to facilitate future DL-based NTCP studies.Approach.Early, joint and late DL multi-modality fusion strategies were compared using clinical and mandibular radiation dose distribution volumes. These were contrasted with single-modality models: a random forest trained on non-image data (clinical, demographic and dose-volume metrics) and a 3D DenseNet-40 trained on image data (mandibular dose distribution maps). The study involved a matched cohort of 92 osteoradionecrosis cases and 92 controls from a single institution.Main results.The late fusion model exhibited superior discrimination and calibration performance, while the join fusion achieved a more balanced distribution of the predicted probabilities. Discrimination performance did not significantly differ between strategies. Late fusion, though less technically complex, lacks crucial inter-modality interactions for NTCP modelling. In contrast, joint fusion, despite its complexity, resulted in a single network training process which included intra- and inter-modality interactions in its model parameter optimisation.Significance.This study is a pioneering effort in comparing different strategies for including image data into DL-based NTCP models in combination with lower dimensional data such as clinical variables. The discrimination performance of such multi-modality NTCP models and the choice of fusion strategy will depend on the distribution and quality of both types of data. Multiple data fusion strategies should be compared and reported in multi-modality NTCP modelling using DL.

2251: NTCP modeling for severe oral mucositis from HNC radiotherapy using radiomics and machine learning

2251: NTCP modeling for severe oral mucositis from HNC radiotherapy using radiomics and machine learning

1677: First late rectal bleeding biological NTCP model for multiple fractionation schedules

1677: First late rectal bleeding biological NTCP model for multiple fractionation schedules

2190: NTCP modeling for rectal complication based on variable RBE-weighted proton dose for prostate cancer

2190: NTCP modeling for rectal complication based on variable RBE-weighted proton dose for prostate cancer

Blood biomarkers for cardiac damage during and after radiotherapy for esophageal cancer: A prospective cohort study

PurposeThe aim of this study was to test the hypothesis that the levels of High Sensitive Troponin T (HS-TNT) and N-terminal Brain Natriuretic Peptide (NT-ProBNP) increase after radiation therapy in a dose dependent way and are predictive for clinical cardiac events. Materials and MethodsBlood samples during and after radiotherapy of 87 esophageal cancer patients were analysed regarding the course of HS-TNT and NT-ProBNP levels and their relationship with clinical toxicity endpoints and radiation dose volume parameters. ResultsHS-TNT values at the end of treatment correlated with the mean heart dose (p = 0.02), whereas the rise of NT-ProBNP correlated with the mean lung dose (p = 0.01). Furthermore, the course of both HS-TNT (p < 0.001) and NT-ProBNP (p < 0.01) levels were significantly different for patients who developed new cardiac events as opposed to those without new cardiac events. ConclusionSignificant correlations were found for both biomarkers with radiation dose and clinical toxicity endpoints after treatment. Therefore, these markers might be of additional value in NTCP models for cardiac events and might help us unravelling the mechanisms behind these toxicity endpoints.

Acute toxicities in proton therapy for head and neck cancer – A matched analysis of the DAHANCA 35 feasibility study

Background and purposeAs preparation for a national randomized study comparing proton radiotherapy to photon radiotherapy, DAHANCA 35, we performed a non-randomized pilot study to investigate patient selection, logistics, planning, and treatment delivery. With the present study, as a comprehensive safety analysis, we want to compare toxicity during and up to two months after therapy to a historically matched group of patients treated with photon radiotherapy. Materials and methods62 patients treated with protons were matched to 124 patients who received photon treatment outside a protocol. Available data were retrieved from the DAHANCA database. Patients were matched on treatment centre, concurrent chemotherapy, tumour site, stage, p16 status for oropharynx cancers. Selection of patients for proton therapy was based on comparative treatment plans with a NTCP reduction for dysphagia and xerostomia at six months. ResultsBaseline characteristics between groups were well balanced, except for the type of drug used concurrently; more photon patients received Carboplatin (21.2 % vs 5.8 %, p = 0.01). Proton therapy was associated with significantly less weight loss at the end of treatment, mean weight loss of 3 % for protons and 5 % for photons (p < 0.001). There were more grade 3 skin reactions and grade 3 mucositis after proton treatment compared with photons at the end of treatment, Risk Ratio (RR) 1.9 (95 % CI: 1.01–3.5, p = 0.04) and RR 1.5 (95 % CI: 1.3–1.7, p < 0.001), respectively. All differences resolved at follow up two months after treatment. There were no significant differences between groups on opioid use, use of feeding tubes, or hospitalization during the observation period. ConclusionProton treatment resulted in excess objective mucositis and dermatitis, which was transient and did not seem to negatively influence weight or treatment compliance and intensity. Selection bias was likely especially since NTCP models were used for selection of proton treatment and photon treated patients were matched manually. We are currently including patients in a randomized controlled trial.

Incidence of alopecia in brain tumour patients treated with pencil scanning proton therapy and validation of existing NTCP models

Background and purposeRadiation-induced alopecia (RIA) is one of the most frequent and upsetting cosmetic side effects after radiotherapy (RT) for brain cancer. We report the incidence of RIA in a cohort of brain tumours patients treated with Proton Therapy (PT) and externally validate published NTCP models of grade 2 (G2) RIA for their implementation in clinical practice. MethodsData for patients treated for brain tumours with scanning beam PT between 2018 and 2022 were extracted. Acute, late and permanent RIA events were evaluated according to CTCAE 5.0. Lyman-Kutcher-Burman (LKB) and multivariable logistic regression (MLR) published models were computed from the relative dose-surface histogram of the scalp. External validity of models was assessed in terms of discrimination and calibration. ResultsIn the 264 patients analysed, rates of any grade acute (≤90 days after PT completion), late (>90 days) and permanent RIA (persisting for> 12 months) were 61.8 %, 24.7 % and 14.4 %, respectively. In our independent cohort, LKB- and MLR-NTCP showed a good discrimination for G2 RIA (0.71≤ROC-AUC≤0.83) while model calibration was unsatisfactory possibly due to a different outcome evaluation between training and validation cohorts, as well as differences in clinical and treatment related variables between the two groups. ConclusionsDespite the reasonable sensitivity and specificity of the NTCP models for RIA in the validation cohort, our study emphasizes the significance of differences between the cohorts utilized for model development and validation. Specifically, variations in the reporting of clinical outcomes inevitably jeopardize the validation of NTCP models. A standardize and objective RIA scoring system is essential.

3359: In silico simulations to support statistics, planning and analysis of clinical trials: The role of TCP and NTCP models

3359: In silico simulations to support statistics, planning and analysis of clinical trials: The role of TCP and NTCP models

Assessment of the Dosimetric Index from IMRT and Rapid arc Plan for Oropharyngeal Cancer with Simultaneous Integrated Boost (SIB) Technique in Combination with EUD-based NTCP and TCP Radiobiological Models.

The current research compared radiobiological and dosimetric results for simultaneous integrated boost (SIB) plans employing RapidArc and IMRT planning procedures in oropharyngeal cancer from head-and-neck cancer (HNC) patients. The indigenously developed Python-based software was used in this study for generation and analysis. Twelve patients with forty-eight total plans with SIB were planned using Rapid arc (2 and 3 arcs) and IMRT (7 and 9 fields) and compared with radiobiological models Lyman, Kutcher, Burman (LKB) and EUD (Equivalent Uniform Dose) along with physical index such as homogeneity index(HI), conformity index(CI) of target volumes. These models' inputs are the dose-volume histograms (DVHs) calculated by the treatment planning system (TPS). The values obtained vary from one model to the other for the same technique and patient. The maximum dose to the brainstem and spinal cord and the mean dose to the parotids were analysed both dosimetrically and radiobiologically, such as the LKB model effective volume, equivalent uniform dose, EUD-based normal tissue complication probability, and normal tissue integral dose. The mean and max dose to target volume with conformity, homogeneity index, tumor control probability compared with treatment times, and monitor units. Rapid arc (3 arcs) resulted in significantly better OAR sparing, dose homogeneity, and conformity. The findings indicate that the rapid arc plan has improved dose distribution in the target volume compared with IMRT, but the tumor control probability obtained for the two planning methods, Rapid arc (3 arcs) and IMRT (7 fields), are similar. The treatment time and monitor units for the Rapid arc (3 arcs) were superior to other planning methods and considered to be standard in head & neck radiotherapy.

3126: Strategies for continuous NTCP model adaptation in an evolving esophageal cancer treatment landscape

3126: Strategies for continuous NTCP model adaptation in an evolving esophageal cancer treatment landscape

1262: NTCP model predicting temporal lobe necrosis in skull-base tumors treated with proton therapy

1262: NTCP model predicting temporal lobe necrosis in skull-base tumors treated with proton therapy

1333: Predicting late taste loss with deep learning NTCP model using 3D dose, CT and OAR segmentations

1333: Predicting late taste loss with deep learning NTCP model using 3D dose, CT and OAR segmentations

2516: NTCP model of acute toxicity after whole breast irradiation: training and extended validation

2516: NTCP model of acute toxicity after whole breast irradiation: training and extended validation

Prediction of auditory brain stem responses damage in patients with head-and-neck cancers receiving radiotherapy using the functional assays of normal tissue complication probability models.

The purpose of this study was to set four NTCP models on clinical data and develop a model that calculates the possibility of hearing damage due to irradiation of healthy and at-risk brainstem tissue. ABR tests were performed on 50 head-and-neck cancer patients three years after radiotherapy for evaluation of lesions in a part of the auditory nerve or the auditory pathway in the brainstem. It indicated a significant difference in the latency of the waves assessed by the ABR test between the two groups. The paired sample t-test indicated the latency time of waves I, III, V, I-III, and I-V (P < 0.001) in the right ear, and in the left ear latency time of waves III, V, I-III, I-V, and III-V (P < 0.001) were significantly higher in the case group's ear than those in the control group. The confidence interval of the fitted parameters was 95% for NTCP models. ABR test's binary outcome with differential dose-volume histograms (dDVHs) was calculated and imported as input to the NTCP modeling. The values of the parameters n = 2.3-2.9 and the value s = 1 were obtained, which indicated that the brainstem organ is seriality. The best model ranked for the prediction of brainstem hearing damage was the logit model, which had the lowest Akaike value. The nervousness of the auditory organ of the brainstem (VIII nerve) can be declared as one of the reasons for being independent of the received dose.

Exploring Helium Ions' Potential for Post-Mastectomy Left-Sided Breast Cancer Radiotherapy.

Proton therapy presents a promising modality for treating left-sided breast cancer due to its unique dose distribution. Helium ions provide increased conformality thanks to a reduced lateral scattering. Consequently, the potential clinical benefit of both techniques was explored. An explorative treatment planning study involving ten patients, previously treated with VMAT (Volumetric Modulated Arc Therapy) for 50 Gy in 25 fractions for locally advanced, node-positive breast cancer, was carried out using proton pencil beam therapy with a fixed relative biological effectiveness (RBE) of 1.1 and helium therapy with a variable RBE described by the mMKM (modified microdosimetric kinetic model). Results indicated that target coverage was improved with particle therapy for both the clinical target volume and especially the internal mammary lymph nodes compared to VMAT. Median dose value analysis revealed that proton and helium plans provided lower dose on the left anterior descending artery (LAD), heart, lungs and right breast than VMAT. Notably, helium therapy exhibited improved ipsilateral lung sparing over protons. Employing NTCP models as available in the literature, helium therapy showed a lower probability of grade ≤ 2 radiation pneumonitis (22% for photons, 5% for protons and 2% for helium ions), while both proton and helium ions reduce the probability of major coronary events with respect to VMAT.

Estimating how contouring differences affect normal tissue complication probability modelling

Background and purposeNormal tissue complication probability (NTCP) models are developed from large retrospective datasets where automatic contouring is often used to contour the organs at risk. This study proposes a methodology to estimate how discrepancies between two sets of contours are reflected on NTCP model performance. We apply this methodology to heart contours within a dataset of non-small cell lung cancer (NSCLC) patients. Materials and methodsOne of the contour sets is designated the ground truth and a dosimetric parameter derived from it is used to simulate outcomes via a predefined NTCP relationship. For each simulated outcome, the selected dosimetric parameters associated with each contour set are individually used to fit a toxicity model and their performance is compared. Our dataset comprised 605 stage IIA-IIIB NSCLC patients. Manual, deep learning, and atlas-based heart contours were available. ResultsHow contour differences were reflected in NTCP model performance depended on the slope of the predefined model, the dosimetric parameter utilized, and the size of the cohort. The impact of contour differences on NTCP model performance increased with steeper NTCP curves. In our dataset, parameters on the low range of the dose-volume histogram were more robust to contour differences. ConclusionsOur methodology can be used to estimate whether a given contouring model is fit for NTCP model development. For the heart in comparable datasets, average Dice should be at least as high as between our manual and deep learning contours for shallow NTCP relationships (88.5 ± 4.5 %) and higher for steep relationships.

Dosimetric Analysis of Proximal Bronchial Tree Subsegments to Assess The Risk of Severe Toxicity After Stereotactic Body Radiation Therapy of Ultra-central Lung Tumors

Background and purposeStereotactic body radiotherapy (SBRT) of ultra-central lung tumors (UCLT) is associated with an increased risk of severe toxicity. The aim of this study was to perform a detailed dosimetric analysis of the proximal bronchial tree (PBT) anatomical sub-segments to evaluate the safety of risk-adapted SBRT and to evaluate potential differences in radiation tolerance between PBT sub-segments. Material and methodsFifty-seven patients treated with SBRT for UCLT between 2014 and 2021 were included. UCLT were defined as tumor abutting or overlapping with the trachea, PBT, or esophagus. This study analyzed overall survival, local control, progression-free survival, and grade ≥3 toxicity events. Bayesian inference was used to build a dose-response model with upper limits for toxicity. ResultsTwenty-seven (47.4%) of the irradiated lesions were primary or locoregionally recurrent NSCLC and 30 (52.6%) oligometastases. All patients were treated with risk-adapted SBRT of median 45.0 Gy (range: 30.0-60.0 Gy) in 8 or 10 fractions. Grade ≥3 radiation pneumonitis was observed in two patients (3.5%), while no bronchial stenosis, hemorrhage or fistula were observed. The dose-response model predicted a grade ≥3 toxicity (stenosis, hemorrhage or fistula) limited to 4.9% (0 - 11.4%) when delivering EQD2_3 = 100 Gy to any location of the PBT (D0.2cc). Detailed dosimetric analysis of PBT substructures showed no variation in the dose-response model between the anatomical PBT sub-segments. ConclusionRisk-adapted SBRT regimens delivered in 8 or 10 fractions for ultra-central lung tumors resulted in high rates of local tumor control with low toxicity rates, without differences in radiation tolerance between the anatomical PBT sub-segments.

Radiation-induced long-term dysphagia in survivors of head and neck cancer and association with dose-volume parameters

BackgroundAlthough dysphagia is a common side effect after radiotherapy (RT) of head and neck cancer (HNC), data on long-term dysphagia is scarce. We aimed to 1) compare radiation dose parameters in HNC survivors with and without dysphagia, 2) investigate factors associated with long-term dysphagia and its possible impact on health-related quality of life (HRQoL), and 3) investigate how our data agree with existing NTCP models. MethodsThis cross-sectional study conducted in 2018–2020, included HNC survivors treated in 2007–2013. Participants attended a one-day examination in hospital and filled in patient questionnaires. Dysphagia was measured with the EORTC QLQ-H&N35 swallowing scale. Toxicity was scored with CTCAE v.4. We contoured swallowing organs at risk (SWOAR) on RT plans, calculated dose-volume histograms (DVHs), performed logistic regression analyses and tested our data in established NTCP models. ResultsOf the 239 participants, 75 (31%) reported dysphagia. Compared to survivors without dysphagia, this group had reduced HRQoL and the DVHs for infrahyoid SWOAR were significantly shifted to the right. Long-term dysphagia was associated with age (OR 1.07, 95% CI 1.03–1.10), female sex (OR 2.75, 95% CI 1.45–5.21), and mean dose to middle pharyngeal constrictor muscle (MD-MPCM) (OR 1.06, 95% CI 1.03–1.09). NTCP models overall underestimated the risk of long-term dysphagia. ConclusionsLong-term dysphagia was associated with higher age, being female, and high MD-MPCM. Doses to distally located SWOAR seemed to be risk factors. Existing NTCP models do not sufficiently predict long-term dysphagia. Further efforts are needed to reduce the prevalence and consequences of this late effect.

RADT-50. EVIDENCE-BASED MULTIVARIATE NORMAL TISSUE COMPLICATION PROBABILITY (NTCP) STUDY OF DOMAIN-SPECIFIC COGNITIVE DECLINE AFTER PARTIAL BRAIN RADIATION (RT)

Abstract Beyond the hippocampus, there are no evidence-based dose constraints for eloquent brain structures which subserve cognitive domains. We performed a multivariate NTCP analysis of post-RT neurocognitive decline, examining dosimetric predictors of eloquent brain regions in a prospective longitudinal clinical trial. 83 patients with primary brain tumors receiving fractionated RT completed comprehensive neurocognitive evaluation and high-resolution volumetric and diffusion MRI at baseline and 6 months post-RT. Image processing using robust, validated automated segmentation parcellated individual white matter (WM) tracts, cortical regions, and hippocampi. Well-validated neurocognitive tests evaluating multiple cognitive domains were assessed with negative reliable change indices adjusted for practice effects between timepoints scored as decline. Univariate logistic regression and multivariate model building were performed examining dose to eloquent structures and clinical variables. Univariate analysis showed mean dose to the bilateral caudal anterior cingulate cortex and WM were correlated with decline in executive function. Attention/processing speed decline was correlated with mean/max dose to the posterior corpus callosum (CC), volume getting &amp;gt;20 Gy in the combined CC, as well as anti-epileptic use and chemotherapy. On multivariate analysis for attention automated bootstrapped logistic regression performance at nested cross-validation by AUC was 0.59 (0.54-0.65); LASSO model performance by AUC was 0.60 (0.54-0.66) with mean dose to combined CC being the most frequent variable in both. The top three most important variables in Random Forest by mean decrease in Gini coefficient were max dose to posterior and combined CC and mean dose to anterior CC. Model performance by AUC was 0.70 (0.64-0.75). Here we present the first, to our knowledge, NTCP model for decline in attention/processing speed, along with dosimetric predictors of executive function decline. We found that after partial brain RT, dose to several ROIs significantly correlated with post-RT impairment. These data can guide future cognitive-sparing strategies for brain RT.

NTCP Modeling and Dose-Volume Correlations of Significant Hematocrit Drop 3 Months After Prostate Radiation Therapy

Our purpose was to determine and model the dose-response relations of different parts of the pelvis regarding the endpoint of hematocrit level drop after pelvic radiation therapy (RT). Two hundred and twenty-one patients treated with RT for prostate adenocarcinoma between 2014 and 2016 were included. All patients had complete blood counts collected at baseline and 3 months post-RT. The net difference of hematocrit level post-RT versus baseline was calculated, and the level of the 15th percentiles defined the thresholds of response in each case. The doses to 8 different pelvic structures were derived and fitted to the hematocrit levels using the relative seriality normal tissue complication probability model and the biologically equivalent uniform dose (). Pelvic structures that correlated with significant decreases in hematocrit were the os coxae bilaterally superior to the acetabulum (OCUB), the total os coxae bilaterally, and the bone volume of the whole pelvis. The structure showing the highest correlation was OCUB with a maximum area under the curve (AUC) of 0.74. For V20 Gy < 30% the odds ratio was 9.8 with 95% CI of 2.9 to 32.9. For mean dose (Dmean) to OCUB, an AUC of 0.73 was observed where the dose threshold was 23 Gy and the odds ratio was 2.7 and 95% CI 1.3 to 5.6. The values for the D50, γ, and s parameters of the relative seriality model were 26.9 Gy (25.9-27.9), 1.3 (1.2-2.2), and 0.12 (0.10-0.83), respectively. The AUC of was 0.73 and patients with to OCUB ≥ 27 Gy had 8.2 times higher rate of significant hematocrit drop versus <27 Gy. These findings confirm the association of radiation-induced damage to pelvic bone marrow with a drop in hematocrit. A threshold of V20 Gy < 30%, Dmean < 23 Gy, or < 27 Gy to OCUB may significantly reduce the risk for this endpoint.