Nursing Department, Kaohsiung Armed Forces General Hospital, Kaohsiung 80201, Taiwan.
Evid Based Complement Alternat Med. 2013;2013:478202. doi: 10.1155/2013/478202. Epub 2013 Jun 18.
Background. This study purposed to validate the use of artificial neural network (ANN) models for predicting myofascial pain control after dry needling and to compare the predictive capability of ANNs with that of support vector machine (SVM) and multiple linear regression (MLR). Methods. Totally 400 patients who have received dry needling treatments completed the Brief Pain Inventory (BPI) at baseline and at 1 year postoperatively. Results. Compared to the MLR and SVM models, the ANN model generally had smaller mean square error (MSE) and mean absolute percentage error (MAPE) values in the training dataset and testing dataset. Most ANN models had MAPE values ranging from 3.4% to 4.6% and most had high prediction accuracy. The global sensitivity analysis also showed that pretreatment BPI score was the best parameter for predicting pain after dry needling. Conclusion. Compared with the MLR and SVM models, the ANN model in this study was more accurate in predicting patient-reported BPI scores and had higher overall performance indices. Further studies of this model may consider the effect of a more detailed database that includes complications and clinical examination findings as well as more detailed outcome data.
本研究旨在验证人工神经网络（ANN）模型在预测肌筋膜疼痛经皮针刺治疗后的控制效果，并比较 ANN 与支持向量机（SVM）和多元线性回归（MLR）的预测能力。方法：共有 400 名接受经皮针刺治疗的患者在基线和术后 1 年时完成了简明疼痛量表（BPI）的评估。结果：与 MLR 和 SVM 模型相比，ANN 模型在训练数据集和测试数据集的平均平方误差（MSE）和平均绝对百分比误差（MAPE）值通常更小。大多数 ANN 模型的 MAPE 值范围在 3.4%至 4.6%之间，且具有较高的预测准确性。全局敏感性分析还表明，治疗前 BPI 评分是预测经皮针刺治疗后疼痛的最佳参数。结论：与 MLR 和 SVM 模型相比，本研究中的 ANN 模型在预测患者报告的 BPI 评分方面更为准确，具有更高的整体性能指标。对该模型的进一步研究可能需要考虑更详细的数据库，包括并发症和临床检查结果以及更详细的结局数据的影响。
