Oral and maxillofacial surgery · 2026
Comparison of multivariate linear regression and random forest model in developing mandibular chin prediction reconstruction equation.
СтатьяСравнительное исследованиеНазвание и аннотация: Перевод в очереди
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.
Soh Hui Yuh, Yu Yao, Feng Zhiheng, Du Wen, Liu Shuo, Peng Xin, Zhang Wen-Bo
Аннотация
PURPOSE: Reconstruction of extensive mandibular defects that cross midline is technically challenging, as the surgeons lack normal anatomical references. This study aims to compare multivariate linear regression and machine learning (random forest) algorithms in predicting the premorbid shape of the mandibular chin to guide surgical reconstruction.
METHODS: Anatomical landmarks and measurements were extracted from the computed tomography (CT) data of the head. The multivariate regression model was developed based on Ordinary Least Squares (OLS), while the dataset was randomly split into training set (80%) and testing set (20%) to develop a machine learning-based algorithm.
RESULTS: Correlation was observed among the maxillary intercanine distance, distance of bilateral zygoma, width of piriform aperture and the width of chin in multivariate linear regression model, and satisfactory performance was achieved, (MAE (2.35), R² (0.25)). No significant difference was noted between the predicted width of chin and actual measurement in paired t-test. (P = 0.092) Bootstrap validation with 500 repetitions demonstrated stable performance (MAE 2.45 mm; R² = 0.17). Machine learning models using the same predictors did not improve predictive accuracy.
CONCLUSION: Both models demonstrated comparable accuracy in predicting chin width; however, the multivariate regression model achieved lower prediction error compared with machine learning approaches using the same predictors. The interpretability and simplicity of multivariate linear regression may make it more clinically practical for reconstructing midline mandibular defects. A straightforward regression model can provide surgeons with a reliable, accessible tool for preoperative planning in complex mandibular reconstructions, especially where advanced computational resources are limited.
METHODS: Anatomical landmarks and measurements were extracted from the computed tomography (CT) data of the head. The multivariate regression model was developed based on Ordinary Least Squares (OLS), while the dataset was randomly split into training set (80%) and testing set (20%) to develop a machine learning-based algorithm.
RESULTS: Correlation was observed among the maxillary intercanine distance, distance of bilateral zygoma, width of piriform aperture and the width of chin in multivariate linear regression model, and satisfactory performance was achieved, (MAE (2.35), R² (0.25)). No significant difference was noted between the predicted width of chin and actual measurement in paired t-test. (P = 0.092) Bootstrap validation with 500 repetitions demonstrated stable performance (MAE 2.45 mm; R² = 0.17). Machine learning models using the same predictors did not improve predictive accuracy.
CONCLUSION: Both models demonstrated comparable accuracy in predicting chin width; however, the multivariate regression model achieved lower prediction error compared with machine learning approaches using the same predictors. The interpretability and simplicity of multivariate linear regression may make it more clinically practical for reconstructing midline mandibular defects. A straightforward regression model can provide surgeons with a reliable, accessible tool for preoperative planning in complex mandibular reconstructions, especially where advanced computational resources are limited.