Abstract:

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 16 862 iterations，and the sensitivity and specificity of breast cancer prediction were 0.941 2 and 0.795 5 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. 

Key words:

breast cancer, blood cell parameter, OPLS?DA, artificial neural network, diagnosis