Local-spatiotemporal Change Monitoring in Extracellular Fluid by Time-variation-constraint Sparse Bayesian Learning Implemented into Frequency-difference Electrical Impedance Tomography (tvcSBL-fdEIT)

Local-spatiotemporal change (LSTC) of frequency-difference conductivity distribution $Delta sigma $ in extracellular fluid (ECF) has been monitored in subcutaneous adipose tissue (SAT) to evaluate leg edema by time-variation-constraint sparse Bayesian learning implemented into frequency-difference electrical impedance tomography (tvc SBL- fd EIT). The tvc SBL- fd EIT has three steps—Step 1: formulation of blocked column vector (BCV), Step 2: SAT separation by time variation constraint, and Step 3: temporal correlation extraction by hyperparameter learning. The tvc SBL- fd EIT was applied to the monitoring of 15 subjects’ calves along with an experimental protocol of prolonged standing and leg elevation. The spatial-mean conductivity $langle Delta sigma rangle ^{mathrm {SAT}}$ in the separated SAT has a strong positive correlation with conventional impedance $z^{mathrm {BIA}}$ by a bioelectrical impedance analysis (BIA) (a correlation coefficient $0.715< R < 0.957$ ; $n$ = 15 and $p < 0.05$ ), which is decreased during the prolonged standing, while $langle Delta sigma rangle ^{mathrm {SAT}}$ is increased during the leg elevation. The frequency dependence of $Delta sigma $ is associated with LSTC of sodium ion concentration in ECF, while the local maximum position of $Delta sigma $ is associated with great saphenous vein (GSV) position. Moreover, the superiority of the proposed algorithm is numerically evaluated under the unstable-background fields.