BioMed research international · 2025
Vocal Cord Dysfunction in Nonlaryngeal Head and Neck Cancer After Chemoradiation Therapy: Predictive Modeling Using CT Radiomics and Machine Learning.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Bagherzadeh Sakineh, Fadavi Pedram, Abdollahi Hamid, Arefpour Amir Mohamad, Asgari Mahdi, Ahmadabad Foad Goli, Safari Mojtaba, Beigi Manijeh
Аннотация
INTRODUCTION: This study aims to investigate computed tomography (CT) radiomic features and dosimetric-clinical biomarkers to predict vocal cord dysfunction (VCD) in nonlaryngeal head and neck cancer (HNC) patients treated with chemoradiation therapy (CRT), using machine learning (ML) models.
METHODS: Sixty-five HNC patients who underwent CRT were recruited to assess radiation-induced VCD 6 months posttreatment. For each patient, CT radiomic features of the laryngeal region, clinical, and dose-volume histogram (DVH) metrics were collected to develop ML models. Nine classifiers were trained using selected features obtained from three feature selection algorithms: least absolute shrinkage and selection operator (LASSO), extra trees, and elastic net. The models were built using imaging features alone (radiomics model) and in combination with clinical and dosimetric features (combined model). Model performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC).
RESULTS: Of the 65 patients, 31 developed VCD. Among radiomics models, the AdaBoost and random forest (RF) classifiers performed best, with AUCs of 0.74 and 0.84, respectively. For the combined models, nine classifiers achieved an AUC greater than 0.95 using LASSO and elastic net algorithms. In contrast, only one classifier surpassed an AUC of 0.95 when using the extra trees algorithm.
CONCLUSION: Our findings demonstrate that pretreatment CT radiomic features are predictive biomarkers for radiation-induced toxicities, including VCD. Furthermore, combining radiomic features with clinical and dosimetric data can improve the predictive modeling of radiotherapy outcomes.
METHODS: Sixty-five HNC patients who underwent CRT were recruited to assess radiation-induced VCD 6 months posttreatment. For each patient, CT radiomic features of the laryngeal region, clinical, and dose-volume histogram (DVH) metrics were collected to develop ML models. Nine classifiers were trained using selected features obtained from three feature selection algorithms: least absolute shrinkage and selection operator (LASSO), extra trees, and elastic net. The models were built using imaging features alone (radiomics model) and in combination with clinical and dosimetric features (combined model). Model performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC).
RESULTS: Of the 65 patients, 31 developed VCD. Among radiomics models, the AdaBoost and random forest (RF) classifiers performed best, with AUCs of 0.74 and 0.84, respectively. For the combined models, nine classifiers achieved an AUC greater than 0.95 using LASSO and elastic net algorithms. In contrast, only one classifier surpassed an AUC of 0.95 when using the extra trees algorithm.
CONCLUSION: Our findings demonstrate that pretreatment CT radiomic features are predictive biomarkers for radiation-induced toxicities, including VCD. Furthermore, combining radiomic features with clinical and dosimetric data can improve the predictive modeling of radiotherapy outcomes.