Ultrasound-guided structural fusion enhances the interpretability of electrical impedance tomography for layered tissue imaging
Source: PubMed Central Open Access, NCBI / U.S. National Library of Medicine
Abstract Electrical Impedance Tomography (EIT) provides low-cost functional impedance imaging, whereas ultrasound offers high-resolution morphological visualization. In this study, a hybrid Ultrasound–EIT (EIT–US) fusion framework is developed for soft-tissue characterization in layered ex vivo biological phantoms using controlled chicken tissue models. A custom EIT acquisition system recorded 208 boundary voltage measurements per acquisition across three tissue layers and multiple frequency–voltage conditions, with experiments repeated across three anatomical tissue configurations. Five FEM-based reconstruction methods—Back Projection (BP), Gauss–Newton (GN), Tikhonov regularization (TIK), GREIT, and a linearized EIDORS-like approach—were implemented within a stand-alone MATLAB framework. Ultrasound images acquired using a GE VScan Air CL probe (musculoskeletal mode) were integrated with EIT reconstructions through a multimodal fusion pipeline incorporating impedance imaging, ultrasound texture–echogenicity mapping, and statistical correlation analysis. Quantitative evaluation demonstrated consistent layer differentiation with moderate structural–electrical agreement (≈ 0.37–0.45). ESI increased by more than 60% at the evaluated layer interfaces during the fusion approach compared to EIT-only reconstructions, along with an increase in boundary detection performance (AUC improvement of approximately 0.13–0.15). Structural similarity (SSIM) of the fused images relative to ultr
Abstract
Abstract Electrical Impedance Tomography (EIT) provides low-cost functional impedance imaging, whereas ultrasound offers high-resolution morphological visualization. In this study, a hybrid Ultrasound–EIT (EIT–US) fusion framework is developed for soft-tissue characterization in layered ex vivo biological phantoms using controlled chicken tissue models. A custom EIT acquisition system recorded 208 boundary voltage measurements per acquisition across three tissue layers and multiple frequency–voltage conditions, with experiments repeated across three anatomical tissue configurations. Five FEM-based reconstruction methods—Back Projection (BP), Gauss–Newton (GN), Tikhonov regularization (TIK), GREIT, and a linearized EIDORS-like approach—were implemented within a stand-alone MATLAB framework. Ultrasound images acquired using a GE VScan Air CL probe (musculoskeletal mode) were integrated with EIT reconstructions through a multimodal fusion pipeline incorporating impedance imaging, ultrasound texture–echogenicity mapping, and statistical correlation analysis. Quantitative evaluation demonstrated consistent layer differentiation with moderate structural–electrical agreement (≈ 0.37–0.45). ESI increased by more than 60% at the evaluated layer interfaces during the fusion approach compared to EIT-only reconstructions, along with an increase in boundary detection performance (AUC improvement of approximately 0.13–0.15). Structural similarity (SSIM) of the fused images relative to ultrasound boundaries ranged from 0.71 to 0.78, indicating preserved spatial coherence. Frequency and voltage variations influenced reconstruction stability and contrast recovery. Overall, multimodal fusion improved the structural interpretability of EIT by anchoring conductivity distributions to ultrasound-derived morphological boundaries. These results indicate that incorporating ultrasound-derived structural priors improves the interpretability of EIT reconstructions by providing spatial context to conductivity distributions. This framework provides a practical experimental approach for integrating complementary electrical and acoustic sensing modalities, with potential relevance to future low-cost and portable soft-tissue imaging systems.