Fig. 3

Hyperplanes separating MCI and mild AD according to XMITTN according to age, Aβ42, and C3 in men with APOE ε4 allele. Using ADNI data, XMITTN constructed a high-dimensional space to model the interaction between biomarkers and diagnosis. To visualize the space, we assigned fixed values to CSF t-Tau and FH, and plotted the planes which distinguishes between MCI and mild AD according to age (depth), CSF Aβ42 levels (X-axis), and zlog(CSF C3) (Y-axis; panel a). Panel A represents a model of C3 in MCI-AD and mild AD with matching gender (male), age (Y-axis), t-Tau, and FH. Subjects whose biomarker combinations fall into the filled regions are considered to have mild AD, and subjects whose biomarker combinations into the empty regions are considered to have MCI. Examining zlog(C3) levels between MCI and mild AD subjects with CSF < 125 pg/mL showed trend detected by XMITTN (panel b, p = 0.07)