AI-enabled electrical impedance tomography (EIT) system for detecting and quantifying metabolic associated steatotic...

Worldwide, 376 million people have chronic fatty liver disease which if not treated, progresses towards fibrosis, cirrhosis and liver cancer, irreversible conditions causing 2 million deaths, 4% of all deaths. Therefore, there is a need for screening and monitoring fatty liver disease at its reversible stages on a community level. However, there is a lack of devices fit for large-scale, high throughput community screening. Here, we demonstrate a system for screening metabolic associated steatotic liver disease (MASLD) using multi-spectral electrical impedance tomography (EIT) frequency difference reconstructions and anthropometric measures. This is done through the prediction of the Controlled Attenuation Parameter (CAP). The study also leverages state-of-the-art deep learning models such as Resnet-18 to significantly boost performance of such predictive systems. Overall, our predicted CAP is strongly correlated with clinical standard Vibration-controlled Transient Elastography (VCTE) recorded CAP with a coefficient of determination R² = 0.740 together with 91.7% sensitivity, 88.1% specificity, and 0.95 receiver operating characteristic area under the curve in classifying patients with MASLD from those without. By enabling accurate CAP prediction with a portable, non-invasive, high throughput, and cost-effective system, this approach facilitates easy screening and monitoring of patients' conditions in community settings, promoting early intervention and management of MASLD.Clinical RelevanceOur portable AI-enabled multi-spectral EIT system has predictive accuracy to facilitate early detection and monitoring as a cost effective and convenient alternative to current solutions for MASLD at a community level.