椎动脉支架内再狭窄的人工神经网络分析

探讨椎动脉狭窄支架植入术后再狭窄的危险因素，并利用人工神经网络对椎动脉支架内再狭窄（ISR）进行预测分析。首先，随访97例临床患者，对 12种可能影响椎动脉支架内再狭窄的因素进行单因素分析，总结出具有统计学意义的相关因素。然后，利用BP神经网络建立影响因素样本集与对应的ISR之间的隐性联系模型。最后，利用神经网络预测患者是否会发生支架内再狭窄，并对预测准确率进行评估。结果表明，置入支架后，再狭窄组中支架长度平均值为15 mm，无再狭窄组患者中支架长度平均值为17 mm，两者具有显著差异(P=0.005)；再狭窄组患者平均扩张比为1.15，无再狭窄组患者平均扩张比为1.17，两者具有显著差异（P=0.01）；再狭窄组和无再狭窄组患者椎动脉侧别也具有显著差异（P=0.045）。同时，评估结果显示，BP神经网络模型预测结果令人满意，不会发生ISR的确诊率q175%，会发生ISR确诊率q2=100%。支架长度、椎动脉侧别和支架扩张比对椎动脉ISR具有显著性影响。BP神经网络模型可用于预测椎动脉ISR的发生。

Artificial Neural Network Analysis for Vertebral Artery In Stent Restenosis

The purpose of this paper is to analyze risk factors causing the vertebral artery in stent restenosis (ISR). Applying the artificial neural network (ANN), we further tried to give predictions on the occurrence of ISR. The first step of our strategy was to follow up 97 randomly picked clinical patients, figuring out the statistical relevance for each of the 12 possible factors leading to ISR. We then established a model to connect the factors with ISR using the back propagation (BP) neural network. The last procedure was to assess our results by comparing our predictions with the actual occurrence of ISR. Our observations show that, the average length of stent is (15±1.52) mm for ISR patients, and（17±1.50）mm for non ISR patients. The value of P=0.005 indicates a statistical significance. The average stent expansion is 1.17±0.16 for ISR patients and 115±025 for non ISR patients. P=0.01 exhibits a statistical significance. The side of vertebral artery for both ISR and non ISR patients is of statistical significance P=0.045. Moreover, the assessments show that the BP neural network prediction model is efficient. Diagnosis rate of non ISR is q1>=75%, while that of ISR is q2=100%. Stent length, vertebral sides and stent expansion are significant factors for the vertebral artery ISR. BP neural networks can be used to predict the occurrence of the vertebral artery ISR.