This study aims to improve the resolution of reconstructed images from industrial ultrasonic tomography (UST) by determining the most effective neural network structure for solving the inverse problem based on the measurements taken. The study analyzed three types of neural networks: Artificial Neural Networks, Convolutional Neural Networks (CNNs) and a hybrid of CNNs and Long Short-Term Memory networks (LSTM). After evaluating the reconstructions and quality indicators, the CNN-LSTM combination provided the most accurate image reconstructions of the industrial ultrasound tomography, highlighting the importance of selecting an appropriate neural network to improve the resolution of the reconstructed images.

This research aimed to develop a high accuracy machine learning algorithm that can diagnose cardiovascular diseases from the stream of data from multiple body surface potential mapping devices equipped with 102 textile electrodes. The algorithm is based on the 1D convolutional neural network, trained on the comparable real-life data gathered from the FLUKE ECG simulator connected to the resistance-based human phantom. The developed neural network achieved an accuracy of 99.91% on the test data. Additionally, an additional algorithm was developed that can use the neural network to analyse the data streamed from the medical device and notice the medical staff about dangerous heart rhythms detected by the system.