偏最小二乘法+神经网络用于大肠癌组织自体荧光的模式识别

Pattern recognition of laser-induced autofluorescence spectrum from colorectal cancer tissues using Partial-least-square and neural network

摘要:

目的大肠癌组织自体荧光的模式识别算法的优化.方法本文将大肠癌自体荧光光谱的判别分析归结为模式识别问题,并首次采用偏最小二乘法 神经网络判别法,即偏最小二乘法进行模式特征分析,完成特征提取后利用主因子作为人工神经网络输入变量,实现类别预测的同时也完成了数学建模与优化分析工作.结果实践证明,该方法可以以较高的灵敏度、特异性和可靠性对组织荧光光谱进行模式分类.结论该方法优于目前该领域同类判别方法.

Abstract:

Objective: To realize pattern recognition of Laser-induced autofluorescence spectrum from colorectal cancer tissues using Partial-least-square. Methods: The auto-fluorescence spectrum classifying was the pattern recognition problem. By using partial least squares, the spectrum was reduced to some factors, which was taken as the input of artificial neural network. Through the training and prediction, artificial neural network outputs classified spectrum. At the same time, the work of building mathematics model and optimizing the algorithm was completed. Results: The result was very good in sensibility and reliability. Conclusion: This method is better than other classification methods in the same field.