Mechanistic cardiac electrophysiology models allow for personalized simulations of the electrical activity in the heart and the ensuing electrocardiogram (ECG) on the body surface. As such, synthetic signals possess known ground truth labels of the underlying disease and can be employed for validation of machine learning ECG analysis tools in addition to clinical signals. Recently, synthetic ECGs were used to enrich sparse clinical data or even replace them completely during training leading to improved performance on real-world clinical test data. We thus generated a novel synthetic database comprising a total of 16,900 12 lead ECGs based on electrophysiological simulations equally distributed into healthy control and 7 pathology classes. The pathological case of myocardial infraction had 6 sub-classes. A comparison of extracted features between the virtual cohort and a publicly available clinical ECG database demonstrated that the synthetic signals represent clinical ECGs for healthy and pathological subpopulations with high fidelity. The ECG database is split into training, validation, and test folds for development and objective assessment of novel machine learning algorithms.
In this study the performance of a planar array for magnetic induction tomography (MIT) was investigated and the results of measurements to determine the precision and sensitivity of the sensor were undertaken. A planar-array MIT system utilizing flux-linkage minimization for the primary field has been constructed and evaluated. The system comprises 4 printed excitation coils of 4 turns which were shielded, 8 surface-mount inductors of inductance 10 microH as sensor, mounted such that in principle no primary-field flux threads them, and a calibration coil to produce a strong primary field. The excitation current was multiplexed via relays to drive the excitation and reference coils. The noise values were similar in real and imaginary components in the lower frequencies and the factor to which the primary field could be reduced was greatest in the nearest coil. Methods for determining the true real and imaginary components and for flux-linkage minimization for the primary field for variations in channel sensitivities are described and the results of measurements of the system's noise and drift are given. A SNR of 47 dB was observed at 4 MHz when a 0.3 Sm-1 saline filled tank of dimensions 20 cmx20 cmx10 cm was placed centrally over the array. Finally, images were reconstructed from measurements of saline samples in a free space background, with the samples moved past the array in 21 1 cm steps to emulate mechanical scanning of the array. The image reconstruction characteristics of the planar array in conjunction with the reconstruction technique employed are discussed.
In magnetic induction tomography reducing the influence of the primary excitation field on the sensors can provide a significant improvement in SNR and/or allow the operating frequency to be reduced. For the purposes of imaging, it would be valuable if all, or a useful subset, of the detection coils could be rendered insensitive to the primary field for any excitation coil activated. Suitable schemes which have been previously suggested include the use of axial gradiometers and coil-orientation methods (Bx sensors). This paper examines the relative performance of each method through computer simulation of the sensitivity profiles produced by a single sensor, and comparison of reconstructed images produced by sensor arrays. A finite-difference model was used to determine the sensitivity profiles obtained with each type of sensor arrangement. The modelled volume was a cuboid of dimensions 50 cmx50 cmx12 cm with a uniform conductivity of 1 S m-1. The excitation coils were of 5 cm diameter and the detection coils of 5 mm diameter. The Bx sensors provided greater sensitivity than the axial gradiometers at all depths, other than on the surface layer of the volume. Images produced using a single-planar array were found to contain distortion which was reduced by the addition of a second array.
Abstract. Atrial fibrillation (AF) is the most common cardiac arrhyth- mia. Patient-specific computational modeling of the atria can provide a better understanding about mechanisms underlying the arrhythmia and will potentially be used for model-based ablation therapy evaluation and planning. Electrical excitation spreads from the left to the right atrium at discrete locations. The location of the muscular bridges cannot be determined from image data. In the present study, left atrial activation sources were manually identified in local activation time maps of 4 AF patients. This information was used to adjust rule-based placed intera- trial bridges in anatomical atrial models of the patients. Sinus rhythm simulations showed a better qualitative agreement to the measured left atrial activation patterns after the adjustment of the bridges. For one patient, the simulated body surface potential (BSP) pattern after the adjustment correlated better to measured BSP maps. The results show that the fusion of intracardiac electrical measurements of early left atrial activation can be used to refine patient atria models with information of the myocardial structure which cannot be imaged. In future, such personalized atrial models may be used to support EP interventions.