Patients suffering from renal failure face an increased risk of sudden cardiac death. The connection between those two diseases is not fully understood, yet. One possible reason could be the varying ionic concentrations of Ca2+ and K+ in this patient group. However, the monitoring of the concentrations is always done with a blood sample and a subsequent laboratory analysis.
It is widely known, that ionic concentrations have an impact on the ECG waves. For example, hyperkalaemia can result in spiky narrow T waves. In previous studies, it was tried to estimate the ionic concentrations continuously with the help of the ECG. First results were promising but results were not as precise as needed. Another step could be to classify hypo-/hyperkalaemia and hypo-/hypercalcaemia with the ECG instead of the precise concentration value. This would allow to alarm patients during critical states and to advise them to go to a hospital.
At first, an extensive literature research is needed. The goal is to find techniques appropriate for the reconstruction. Furthermore, existing augmentation techniques should be found. Existing input data have to be preprocessed. In a second step, appropriate algorithms should be selected, implemented and evaluated. Possible improvements or new algorithms should be considered, too. At the end, and possibly with the help of the augmented datasets, the problem of ionic concentration reconstruction will be solved with supervised learning methods in comparison with other methods existing in literature.
• Literature research
• Fundamentals of cardiac electrophysiology
• Good foundations and interest in signal processing, machine learning and statistics
• Experience with deep learning could be beneficial
• Programming skills in MATLAB and Python