基于OPLS⁃DA和人工神经网络算法研究血细胞参数在乳腺癌诊断中的价值

Study on the value of blood cell parameters in the diagnosis of breast cancer based on OPLS⁃DA and artificial neural network algorithm

目的运用OPLS⁃DA和人工神经网络算法发掘和研究血细胞参数对乳腺癌的诊断价值。方法收集乳腺癌患者和健康女性的血细胞参数,通过OPLS⁃DA发掘两者之间的主要差异血细胞参数;通过ROC法计算主要差异参数对乳腺癌诊断的灵敏度和特异性;基于主要差异参数,建立神经网络诊断模型,用于乳腺癌的诊断预测。结果OPLS⁃DA提示基于血细胞参数,乳腺癌患者和健康女性之间存在显著差异,平均血小板体积(MPV)、嗜碱性粒细胞绝对值(BA#)、血小板计数(PLT)、平均红细胞体积(MCV)、红细胞计数(RBC)和淋巴细胞绝对值(LY#)是两组之间的主要差异参数。上述参数对乳腺癌诊断特异性分别为0.564、0.983、0.622、0.674、0.878和0.762,灵敏度分别为0.819、0.614、0.618、0.561、0.393和0.514,ROC曲线下面积(AUC)分别为0.773、0.793、0.657、0.649、0.643和0.635。基于6种主要参数,通过逆向传播算法经16862次迭代建立人工神经网络模型,对乳腺癌预测的灵敏度为0.9412,特异性为0.7955。结论本文通过OPLS⁃DA算法发掘了乳腺癌患者和健康女性之间的主要差异血细胞参数,成功建立了人工神经网络的乳腺癌预测模型,对乳腺癌的诊断筛查具有一定价值。

Objective Investigate the value of blood cell parameters in breast cancer diagnosis by OPLS⁃DA and artificial neural network algorithm.Methods Blood cell parameters from breast cancer patients and healthy females were collected.The OPLS⁃DA algorithm was used to calculate the significant differences between breast cancer patients and normal females.ROC was used to assess the diagnostic efficacy of the prominent parameters,including sensitivity and specificity.A prediction model for breast cancer diagnosis is established by neural network algorithm based on the prominent different parameters.Results According to the OPLS⁃DA analysis,there are significant differences in blood cell parameters between breast cancer patients and healthy women.MPV,BA#,PLT,MCV,RBC and LY#were the main differential parameters between the two groups.The specificity of these parameters for breast cancer diagnosis is 0.564,0.983,0.622,0.674,0.878 and 0.762 respectively;the sensitivity is 0.819,0.614,0.618,0.561,0.393 and 0.393 respectively;and the AUC is 0.773,0.793,0.657,0.649,0.643 and 0.635 respectively.Based on MPV,BA#,PLT,MCV,RBC and LY#,the artificial neural network model was estab⁃lished by 16862 iterations,and the sensitivity and specificity of breast cancer prediction were 0.9412 and 0.7955 respectively.Conclusion The OPLS⁃DA algorithm is used in this study to investigate the main differential parame⁃ters of blood cell parameters between breast cancer patients and healthy women.Breast cancer prediction model based on artificial neural network has been successfully established,which is of certain value for breast cancer screening.