基于支持向量机的脊柱离体组织电阻抗识别

目的 测量不同脊柱组织的电阻抗,基于支持向量机建立电阻抗数据的组织分类算法并验证算法的准确性,寻找不同组织电阻抗分类阈值。 方法 取离体脊柱组织,应用电化学分析仪采集10~100 kHz频率范围内皮质骨、松质骨、脊髓、肌肉、髓核的电阻抗。将两只猪采集的数据集分别作为训练集和测试集,应用主成分分析降维至二维数据,训练和验证基于支持向量机(SVM)建立的分类算法,应用集成学习的方法计算不同组织分类的电阻抗阈值。 结果 5种组织在10~100 kHz的测量频率内,电阻抗值差异有统计学意义(P<0.001)。应用主成分分析降维的数据集建立的支持向量机分类算法识别不同组织的准确率为100%。应用集成学习建立的多个分类器计算出了不同组织的电阻抗分类阈值。 结论 基于支持向量机可以实现脊柱术区组织电阻抗的准确识别,有望应用于临床协助医生提升组织识别准确率。

Electrical impedance recognition of spine tissue in vitro based on support vector machine

Objective To measure impedances of different spinal tissues, to develop a tissue classification algorithm based on the impedance data using support vector machine(SVM), in order to verify the accuracy of the algorithm and to search for thresholds for impedance classification of different tissues. Methods Isolated spinal tissue was collected, and the electrical impedances of cortical bone, cancellous bone, spinal cord, muscle, and nucleus pulposus were collected using an electrochemical analyzer in the frequency range of 10-100 kHz. The data sets of two pigs were used as training set and test set, respectively. Principal component analysis(PCA)was used to reduce the dimensionality to two-dimensional data, and SVM based classification algorithm was trained and validated. Ensemble learning method was used to calculate the electrical impedance threshold for different tissue classifications. Results The electrical impedances of the five tissues were significantly different(P<0.001)within the measurement frequency of 10-100 kHz. The accuracy of PCA dimensionality reduction based SVM classification algorithm was 100%. Ensemble learning based multi-classifiers were used to compute impedance classification thresholds for different tissues. Conclusion Accurate identification of the electrical impedance of tissues in spinal surgery can be achieved based on support vector machines, which are expected to help surgeons improve the accuracy of tissue identification.