| 3 | 0 | 27 |
| Downloads | Citas | Reads |
Conventional normal tissue complication probability(NTCP) models often fail to adequately account for biological heterogeneity and provide individualized toxicity risk assessment in radiotherapy. To address these limitations,this study proposed a multimodal machine learning framework integrating clinical, computed tomography(CT)-based radiomic, and three-dimensional dosiomic features for three-class risk stratification of acute radiation esophagitis(ARE) after radiotherapy. A multimodal dataset was constructed from a cohort of 161 patients with locally advanced non-small cell lung cancer, comprising 56 patients with grade 0 ARE, 68 with grade 1 ARE, and 37 with grade ≥2 ARE. First, high-throughput features were extracted from planning CT images and three-dimensional dose distributions using PyRadiomics. A multi-stage feature selection strategy was then applied: unstable features were removed based on the intraclass correlation coefficient(ICC), followed by dimensionality reduction using Spearman correlation analysis and least absolute shrinkage and selection operator(LASSO) regression. This process yielded 28 key features,including 2 clinical, 19 radiomic, and 7 dosiomic features. The selected features were subsequently used to construct an XGBoost-based multimodal fusion classifier, with hyperparameters optimized by ten-fold cross-validation and grid search. SHapley Additive exPlanations(SHAP) were further employed to interpret the model at both global and local levels. On the test set, the multimodal fusion model achieved an area under the receiver operating characteristic curve(AUROC) of 0.770 and an F1-score of 0.611. Ablation experiments showed that the fusion model achieved an AUROC of 0.770, compared with 0.758 for the radiomics model, while obtaining better overall performance in terms of F1-score, recall, and specificity. The fusion model also outperformed the clinical model and the dosiomics model(AUROC = 0.580). SHAP analysis identified CT-based radiomic features, including InverseVariance and GrayLevelVariance, as well as several dosiomic texture features, as important contributors to model predictions. These results suggest that the proposed multimodal feature fusion framework can improve the graded prediction of complex radiation toxicities and may provide a useful computational decision-support approach for intelligent and individualized radiotherapy planning.
[1]Baker S , Fairchild A. Radiation-induced esophagitis in lung cancer[J]. Lung Cancer:Targets and Therapy, 2016 ,7:119-127.
[2]McCaughan H , Boyle S , McGoran J J. Update on the management of the gastrointestinal effects of radiation[J].World Journal of Gastrointestinal Oncology, 2021, 13(5):400-408.
[3]Liang B, Yan H, Tian Y, et al. Dosiomics:extracting 3D spatial features from dose distribution to predict incidence of radiation pneumonitis[J]. Frontiers in Oncology, 2019,9:269.
[4]Oberije C, de Ruysscher D, Houben R, et al. A validated prediction model for overall survival from stage III nonsmall cell lung cancer:toward survival prediction for individual patients[J]. International Journal of Radiation Oncology, Biology, Physics, 2015, 92(4):935-944.
[5]Desideri I, Loi M, Francolini G, et al. Application of radiomics for the prediction of radiation-induced toxicity in the IMRT era:current state-of-the-art[J]. Frontiers in Oncology, 2020, 10:1708.
[6]Wang L, Gao Z H, Li C M, et al. Computed tomographybased delta-radiomics analysis for discriminating radiation pneumonitis in patients with esophageal cancer after radiation therapy[J]. International Journal of Radiation Oncology Biology Physics, 2021, 111(2):443-455.
[7]Bourbonne V , Da-ano R , Jaouen V , et al. Radiomics analysis of 3D dose distributions to predict toxicity of radiotherapy for lung cancer[J]. Radiotherapy and Oncology,2021, 155:144-150.
[8]Bourbonne V, Lucia F, Dissaux G, et al. Pulmonary and esophageal toxicity in lung cancer treated by(chemo)-radiotherapy:a radiomics-based prediction model[J]. International Journal of Radiation Oncology Biology Physics, 2020,108(3):S31.
[9]Aerts H J W L, Velazquez E R, Leijenaar R T H, et al.Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach[J]. Nature Communications, 2014, 5:4006.
[10]Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics:extracting more information from medical images using advanced feature analysis[J]. European Journal of Cancer, 2012, 48(4):441-446.
[11]Rossi L, Bijman R, Schillemans W, et al. Texture analysis of 3D dose distributions for predictive modelling of toxicity rates in radiotherapy[J]. Radiotherapy and Oncology ,2018, 129(3):548-553.
[12]El Naqa I, Bradley J, Blanco A I, et al. Multivariable modeling of radiotherapy outcomes, including dose-volume and clinical factors[J]. International Journal of Radiation Oncology Biology Physics, 2006, 64(4):1275-1286.
[13]Huynh E, Hosny A, Guthier C, et al. Artificial intelligence in radiation oncology[J]. Nature Reviews Clinical Oncology, 2020, 17(12):771-781.
[14]Nohara Y, Matsumoto K, Soejima H, et al. Explanation of machine learning models using shapley additive explanation and application for real data in hospital[J]. Computer Methods and Programs in Biomedicine, 2022, 214:106584.
[15]Lundberg S M, Lee S I. A unified approach to interpreting model predictions[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. New York:ACM, 2017:4768-4777.
[16]Lam S K, Zhang Y P, Zhang J, et al. Multi-organ omicsbased prediction for adaptive radiation therapy eligibility in nasopharyngeal carcinoma patients undergoing concurrent chemoradiotherapy[J]. Frontiers in Oncology , 2021 , 11:792024.
[17]van Griethuysen J J M, Fedorov A, Parmar C, et al. Computational radiomics system to decode the radiographic phenotype[J]. Cancer Research, 2017, 77(21):e104-e107.
[18]Zwanenburg A, Vallières M, Abdalah M A, et al. The image biomarker standardization initiative:standardized quantitative radiomics for high-throughput image-based phenotyping[J]. Radiology, 2020, 295(2):428-438.
[19]Tibshirani R. Regression shrinkage and selection via the Lasso[J]. Journal of the Royal Statistical Society Series B:Statistical Methodology, 1996, 58(1):267-288.
[20]Tibshirani R. Regression shrinkage and selection via the Lasso:a retrospective[J]. Journal of the Royal Statistical Society Series B:Statistical Methodology, 2011, 73(3):273-282.
[21]O′Rourke N, RoquéI F M, FarréBernadóN, et al. Concurrent chemoradiotherapy in non-small cell lung cancer[J]. Cochrane Database of Systematic Reviews, 2010, 2010(6):CD002140.
[22]Zheng X L, Guo W, Wang Y H, et al. Multi-omics to predict acute radiation esophagitis in patients with lung cancer treated with intensity-modulated radiation therapy[J]. European Journal of Medical Research, 2023, 28(1):126.
[23]Ma Z L, Liang B, Wei R, et al. Enhanced prediction of postoperative radiotherapy-induced esophagitis in nonsmall cell lung cancer:dosiomic model development in a real-world cohort and validation in the PORT-C randomized controlled trial[J]. Thoracic Cancer, 2023, 14(28):2839-2845.
[24]Gong B Y, Jiang N, Yan G M, et al. Predictors for severe acute esophagitis in lung cancer patients treated with chemoradiotherapy:a systematic review[J]. Current Medical Research and Opinion, 2016, 32(10):1701-1708.
Basic Information:
DOI：10.12194/j.ntu.20251229001
China Classification Code:R734.2;R730.55
Citation Information:
[1]Cheng Xiuyan,Zheng Xiaoli,Mao Ronghu ,et al.A SHAP-integrated multimodal model for three-class risk prediction of acute radiation esophagitis in non-small cell lung cancer[J].Journal of Nantong University (Natural Science Edition),2026,25(03):25-34+57.DOI:10.12194/j.ntu.20251229001.
2026-09-20
2026-09-20