Ying Wang

E-health solutions, such as patient monitoring using wearable technologies, have been widely promoted. The broad usage of e-health solutions can decrease the burden of national health care system and increase quality of care. However, the extremely variable physiological signals and patient demography are barriers to accurate e-health monitoring systems. In this thesis, we stated our research about e-health monitoring systems for individuals with epilepsy or Parkinson’s disease. We developed the systems from the perspective of clinical (chapter 2 & 4), basic science (chapter 5 & 6), and engineering (chapter 3 & 5) research.

With respect to the applications for individual with epilepsy, we first did a clinical study to investigate the causes of unreliable electroencephalography (EEG) analysis for the diagnosis of non-convulsive status epilepticus (chapter 2). The typical pitfalls in EEG visual (by human raters) and automated (by computer technology) analysis were identified in this chapter. We also provided suggestions as to how those pitfalls might be avoided. We found that short ictal discharges with a gradual onset (developing over 3 seconds in length) were liable to be misinterpreted. Other pitfalls were the misinterpretation of abnormal background activity (slow wave activities, other abnormal brain activity, and ictal-like movement artifacts), continuous interictal discharges, and continuous short ictal discharges. We argued that a longer duration criterion for NCSE-EEGs than the one suggested by the Salzburg criteria is needed. Using knowledge of historical EEGs, individualized algorithms, and context-dependent alarm thresholds may also avoid the pitfalls.

To reduce false alarms in the non-convulsive seizure detection, we proposed a detection system to prevent three common errors: over-interpretation of abnormal background activity, dense short ictal discharges and continuous interictal discharges as ictal discharges in chapter 3. To reduce the false alarms caused by abnormal background activity, we used morphological features extracted by visibility graph methods in addition to time-frequency features. To reduce the false alarms caused by over-interpreting short ictal discharges and interictal discharges, we created two synthetic classes—“Suspected Non-ictal” and “Suspected Ictal”—based on the misclassified categories and constructed a synthetic 4-class dataset combining the standard two classes—“Non-ictal” and “Ictal”—to train a 4-class classifier. The 4-class classification model improved the performance of the standard 2-class model, especially increased the precision by 15% at an 80% sensitivity level when only time-frequency features were used. Using the morphological features, the 4-class classification model achieved the best performances: a sensitivity of 93% ± 12% and a precision of 55% ± 30% in the group level. 100% accuracy was reached in a participant’s 4.3-hour recording with 5 ictal discharges.

With respect to the applications for individual with Parkinson’s disease, we firstly executed a clinical practice study to determine the most effective in-place task to provoke freezing of gait (chapter 4). We included 16 patients with Parkinson’s disease and subjective experience of daily freezing episodes. All patients were examined in a practically defined OFF state, i.e., >12 hours after intake of the last dose of dopaminergic medication. The following tasks were performed once by each patient, for 30 seconds each: (a) stepping in place at self-selected speed; (b) making a rapid half turn (180 degree) in place; and (c) making a rapid full turn (360 degree) in place. Rapid half and full turn in place were found to be more effective to provoke freezing episodes than stepping-in-place. We recommend to ask the patient to make rapid alternating 180 or 360 degrees turns on the spot (360 degrees being preferred for patients with milder freezing) when aiming to provoke freezing in daily clinical practice (when the available time for physical examination is limited), and to repeat this when the first result is negative.

To validly and reliably evaluate freezing of gait in daily life, we characterized and detected freezing episodes using multi-modal features from brain, eye, heart, motion, and gait activity (chapter 5). The eye-stabilization speed during turning and lower-body trembling measure were found to be significantly associated with freezing episodes and therefore used for freezing detection. Using a leave-one-subject-out cross-validation, we obtained a sensitivity of 97% ± 3%, a specificity of 96% ± 7%, a precision of 73% ± 21%, a Matthews correlation coefficient of 0.82 ± 0.15, and an area under the Precision-Recall curve of 0.94 ± 0.05. According to the Precision-Recall curves, the proposed freezing detection method using the multi-modal features performed better than using single-modal features.

In chapter 6, we presented a basic science study to further explore the relationship between freezing episodes and saccadic eye movements for gaze-direction and gaze-direction stabilization during turning. The eye movements acquired from electrooculography (EOG) signals were characterized by the average position of gaze, the amplitude of gaze shifts, and the speed of gaze-direction stabilization. We compared these variables before and during freezing episodes with the variables during successful turning. Significant changes of gaze direction were observed almost one turn cycle before freezing episodes. In addition, the speed of gaze-direction stabilization significantly decreased during freezing episodes. We speculate that different gaze direction than the current turning orbit might be predictive for freezing due to continued failure in movement-error correction or an insufficient preparation for eye-to-foot coordination during turning. In addition, we argue that the decreases in the speed of gaze-direction stabilization is an evidence of a healthy vestibular reflex system in individuals with freezing episodes.

The application of e-health solution has been being widely used in daily life with the rapid progress of techniques. The universal application of e-health monitoring systems can help decrease the burden of health care systems and assist timely treatments for individuals. However, accurate e-health monitoring systems are still challenging. Hereby, we would strongly suggest a close collaboration among researchers in the clinical, basic science, and engineering field in future e-health monitoring studies. Inspired by problems during clinical practices, supported by the evidence in basic science, and customizing engineering techniques, we could obtain accurate e-health monitoring systems for people around the world in the future.