Electrolyte imbalance (EI) is a common and potentially fatal complication of chronic cardiovascular and kidney diseases. Timely detection of EI would enable preemptive medical intervention before the onset of life-threatening events. However, EI is often asymptomatic and virtually undetectable without a blood test, which cannot be done in ambulatory monitoring. Thus, there is a need to create noninvasive technologies for blood electrolyte monitoring. Recent studies have demonstrated the potential of using electrocardiograms (ECGs) to derive ventricular repolarization markers (VRMs) sensitive to electrolyte fluctuations during hemodialysis. Despite the encouraging results, the proposed VRMs require 12-lead or precordial-lead ECG systems that are uncomfortable for ambulatory monitoring. Accordingly, methods to derive VRMs from reduced-lead ECGs are needed to facilitate noninvasive ambulatory blood electrolyte monitoring. This thesis presents two methods for this purpose and unveils the current challenges of ECG-based monitoring of blood electrolyte levels. The first method measures T-wave morphology changes from single-lead ECGs via model-based parameterization. It is applied to explore the feasibility of capturing blood potassium fluctuations in ambulatory settings for the first time. The second method uses a deep-learning model with an original composite loss function to estimate one of the most well-established VRMs—the spatial QRS-T angle—from reduced-lead ECGs.