Department of ECE, Vardhaman College of Engineering, Hyderabad, India.
Department of CSE, Prasad V Potluri Siddhartha Institute of Technology, Vijayawada, India.
Eur Spine J. 2023 Jun;32(6):2140-2148. doi: 10.1007/s00586-023-07712-6. Epub 2023 Apr 15.
Due to the diversity of patient characteristics, therapeutic approaches, and radiological findings, it can be challenging to predict outcomes based on neurological consequences accurately within cervical spinal cord injury (SCI) entities and based on machine learning (ML) technique. Accurate neurological outcomes prediction in the patients suffering with cervical spinal cord injury is challenging due to heterogeneity existing in patient characteristics and treatment strategies. Machine learning algorithms are proven technology for achieving greater prediction outcomes. Thus, the research employed machine learning model through extreme gradient boosting (XGBoost) for attaining superior accuracy and reliability followed with other MI algorithms for predicting the neurological outcomes. Besides, it generated a model of a data-driven approach with extreme gradient boosting to enhance fault detection techniques (XGBoost) efficiency rate. To forecast improvements within functionalities of neurological systems, the status has been monitored through motor position (ASIA [American Spinal Injury Association] Impairment Scale [AIS] D and E) followed by the method of prediction employing XGBoost, combined with decision tree for regression logistics. Thus, with the proposed XGBoost approach, the enhanced accuracy in reaching the outcome is 81.1%, and from other models such as decision tree (80%) and logistic regression (82%), in predicting outcomes of neurological improvements within cervical SCI patients. Considering the AUC, the XGBoost and decision tree valued with 0.867 and 0.787, whereas logistic regression showed 0.877. Therefore, the application of XGBoost for accurate prediction and decision-making in the categorization of pre-treatment in patients with cervical SCI has reached better development with this study.
由于患者特征、治疗方法和影像学表现的多样性，基于机器学习（ML）技术，准确预测颈椎脊髓损伤（SCI）实体和基于神经后果的结果具有挑战性。由于患者特征和治疗策略存在异质性，准确预测颈椎脊髓损伤患者的神经预后具有挑战性。机器学习算法是实现更高预测结果的成熟技术。因此，该研究通过极端梯度提升（XGBoost）采用机器学习模型，以获得更高的准确性和可靠性，然后再采用其他 MI 算法来预测神经预后。此外，它生成了一个具有极端梯度提升的数据驱动方法模型，以提高故障检测技术（XGBoost）的效率。为了预测神经功能系统的改善，通过运动位置（美国脊髓损伤协会 [ASIA] 损伤量表 [AIS] D 和 E）监测状态，然后使用 XGBoost 结合决策树进行回归逻辑预测方法。因此，通过提出的 XGBoost 方法，达到结果的增强准确性为 81.1%，而其他模型，如决策树（80%）和逻辑回归（82%），在预测颈椎 SCI 患者神经改善的结果方面。考虑到 AUC，XGBoost 和决策树的值分别为 0.867 和 0.787，而逻辑回归显示为 0.877。因此，XGBoost 在准确预测和决策分类中的应用在患者的颈椎 SCI 患者的治疗前分类中取得了更好的发展。
