IEEE Transactions on Instrumentation and Measurement · 2021 · 16 citations · 36 references
EngineeringBioengineeringBiomedical EngineeringBiomedical ImagingDiagnostic ImagingSensors3D ImagingDiagnostic Imaging SystemsGastrointestinal ImagingImage ReconstructionMedical ImagingMedical Imaging InstrumentationConductivityElectrical Impedance TomographyNumerical SimulationSignal-to-noise RatioSignal-to-noise Ratio EstimationThree-dimensional Medical ImagingTomographic Imaging SystemsTomography Systems
In this study, a functional gastric-shape imaging method based on a pocket electrical impedance tomography (p-EIT) system is proposed which consist of the development of (1) a buffer-mirrored current source (BMCS) as a wide impedance range circuit and (2) a filter-trained quasi-3D method as an image reconstruction algorithm. The BMCS provides a wide impedance range of constant current Ic with a high signal-to-noise ratio SNR. The filter-trained quasi-3D method iteratively trains the threshold α to obtain the most sensitive impedance Z pair from a multilayer sensor. Using simulations and experiments, the performance of BMCS is validated by comparison with the conventional impedance circuit which is based on an enhanced Howland current source (EHCS). With the p-EIT system, gastric-shaped objects are imaged under the various content conductivity Δσc, different levels of gas-liquid ratio Δσl, and different gastric volume ΔσV. Based on simulations and experiments, the BMCS successfully provides Ic 1 mA on 1-27.54 kΩ of Z range, meanwhile, it provides 1-10 kΩ in a practical test. By SNR over than 80 dB, the BMCS is proper to be examined for several cases. In Δσc, using the BMCS and filter-trained quasi-3D method on the p-EIT system improves the accuracy of filtered images σ* which has a 24.18% higher cross-correlation CC, 23.11% lower image error IAE, and 41.03% lower root-mean-square error RMSE. In Δσl and ΔσV, the quasi-3D conductivity distribution of numerical simulation Θsim and experiment Θexp studies are evaluated by volumetric content estimation on liquid storage φV emptying φl process. From the results, in ΔσV, φV,sim increases from 5.87 cm3 to 9.62 cm3 and φV,exp increases from 3.29 cm3 to 10.23 cm3. Also, in Δσl, show that φl,sim decreases from 13.26 cm3 to 2.28 cm3 and φl,exp decreases from 12.11 cm3 to 2.93 cm3. The developed p-EIT system reaches average volumetric improvement 〈Δeφ〉 = 57.71% in the case of Δφl and 12.30% in the case of ΔφV. Both of Δφsim and Δφexp have a similarity pattern close to Δφtrue.
36