Real-time Dynamic Imaging Method for Flexible Boundary Sensor in Wearable Electrical Impedance Tomography
Panji Nursetia Darma, Marlin Ramadhan Baidillah, Martin Wekesa Sifuna, Masahiro Takei
Abstract
A novel real-time dynamic imaging method has been proposed for flexible boundary sensors in wearable electrical impedance tomography (wearable EIT). The novel method has three stages; 1) estimation of flexible boundary shape Ω(t) at time t, 2) computations of clustered Jacobian matrix J(Ω(t)) using parallel cloud computing (PCC) and 3) reconstruction of conductivity distribution images σ(Ω(t)) based on Ω(t). Initially, Ω(t) is estimated based on a shape factor n <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> of Lame curve which is calculated from axial circumference c <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> and axial length m <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> . The c <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> and m <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> are calculated by stretch detectors d <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</sub> and angle detectors θ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</sub> integrated in the wearable sensor. J(Ω(t)) is then clustered into several processors p in PCC based on potential fields produced by the current injected to the e-th electrode and the measured voltage from m-th adjacent electrode patterns. Finally, σ(Ω(t)) is reconstructed by the Gauss-Newton method based on J(Ω(t)) and measured voltage V. The performance of the proposed imaging method is qualitatively and quantitatively evaluated using three different shaped phantoms. From the results, the boundary shape is well estimated with boundary error be = 5.78% compared with the true boundary shape. The clustered Jacobian matrix computation J(Ω(t)) improves the performance speed sp by 20.17 times compared with the standard Jacobian matrix computation. It also provides a more accurate Jacobian matrix with a Jacobian error je = 2.92% compared with the true Jacobian matrix. The σ(Ω(t)) provides more accurate reconstructed images with relatively low rmse = 0.0023[-].