The region-of-interest based measurement selection process for electrical impedance tomography in radiofrequency cardiac ablation with known anatomical information

Radiofrequency (RF) cardiac ablation is a common technique for efficiently treating Cardiac Arrhythmias. However, the information of the dynamic lesion heating is still currently unavailable in a conventional catheterization laboratory. Electrical Impedance Tomography (EIT) is a temporal impedance imaging modality which might be able to track the heat-caused changes within the electrical property distribution of myocardium. Within this paper, a measurement selection process focusing on the region-of-interest (ROI), i.e. the ablated site, is proposed to deal with the EIT's illposedness, the biggest obstruction to EIT's feasibility this application. By gathering an adequate number of only high-sensitive measurements, the aim is to balance between the information need out of ROI and a well-posed Err inverse solver. The process takes advantage of (1) known anatomical information due to preprocedural CT scans and (2) known intracardiac catheter location from electroanatomical mapping systems. The L-2-norm of ROI conductivity changes was used to predict two lesion parameters including the depth and the width. A model was made including a 2.5D realistic thorax boundary and lungs, and an elliptical heart of 16 mm in thickness with 16 boundary electrodes and 3 internal catheter-based electrodes. 8 ablated positions were tested, in each one reference ablation for fitting the EIT-lesion size prediction curves and 100 testing ablations were simulated using the Pennes' Bioheat transfer equation. Three disturbances including (1) varied myocardium conductivity, (2) varied blood flow convection effect and (3) catheter location mapping error, and the Gaussian noises up to 100 mu V in standard deviation were introduced in the test. The results showed that, with no more than 12 measurements, EIT with the optimized measurement sets performed well when dealing with all three disturbances while allowing acquiring a high image rate in short time to flatten the noise up 100 mu V. (C) 2019 Elsevier Ltd. All rights reserved.