Electromyography Signals as Biomarkers for Parkinson's Disease
Electromyography Signals as Biomarkers for Parkinson's Disease
A simple, painless and reliable method to detect Parkinson's disease at an early stage is very important to patients, doctors and researchers. Doctors want to help patients early, and scientists want to select patients for their research who will help in development of better drugs. We hope that the changes in electrical activity of hand muscles during handwriting will help in early detection of this disease.
This study will use the analysis of electrical activity recorded from hand muscles during handwriting and at rest. There will be two groups of subjects: early Parkinson's disease patients and healthy people. The researcher analyzing the recorded data will not know who is a patient and who is healthy, as subjects will be identified only by numbers. Healthy volunteers will be of similar age as patients. In the course of this study, various properties of hand muscle electrical activity will be examined, and results will be verified by third party.
A neurologist will accrue 10 early PD patients with mild symptoms and 10 healthy controls. Inclusion and exclusion criteria of participants (early PD patients and healthy controls) will be described by the clients after a discussion with the neurologist. It is important that the healthy controls should be similar to the early PD patients in terms of age, gender, and other factors which might also cause differences in EMG signals. The neurologist's diagnosis of the participants' disease status (early PD or health) will be considered the "reference standard test" results, and will be kept confidential until the end of the study. That is, only the neurologist knows the diagnosis for each participant accrued at the end of the study.
An assistant (recorder), who does not know the disease status of these participants and does not know the study design (e.g., how many PD participants and how many health controls), will record EMG signals of these participants following the pre-specified protocol. The order of these participants being examined by the recorder will be randomized. The signals will be analyzed by the software provided by the clients and results needed for diagnosis will be outputted and saved in individual files, one for each participant.
Another assistant (reader), who has no contact with these participants and does not know the study design (e.g., how many PD participants and how many health controls), would then diagnose each individual as early PD or health based on the analysis outputs, according to pre-specified rules as described in the proposal.
At the end of the study, the reader's diagnoses will be compared to the neurologist's diagnosis by a third party. A diagnosis by the reader is defined as correct if this diagnosis is the same as the neurologist's diagnosis. The success rate of our approach of diagnosing early PD disease is defined by the total number of corrected diagnoses by the reader divided by the total number of diagnoses, which equals to the total number of participants.
Statistical analysis The null hypothesis will be rejected, i.e., the client's claim about the capability of their approach in diagnosing early PD should be accepted, if the number of correct diagnosis equals to or exceeds 15, Otherwise, the null hypothesis will not be rejected and the clients' claim about the capability of their approach in diagnosing early PD will not be accepted. We claim that the success rate of their approach should be no less than 0.8.
We denote P0 (= 0.5) as the success rate under the null hypothesis, and P1 as the success rate under the alternative hypothesis. We expect P1 >= 0.80 based on pilot study results. A sample size of 20 participants achieves 80% power to detect a difference (P1-P0) of 0.30 using a one-sided binomial test. The target significance level is 0.05. The actual significance level achieved by this test is 0.0207. These results assume that the population proportion under the null hypothesis is 0.50.
As a secondary objective, the clients could also generate estimates and 95% confidence intervals of sensitivity and specificity of this approach for diagnosing PD.
However, it should be noted that given the small sample size, we couldn't produce accurate estimate of sensitivity and specificity. For example, with 10 PD participants, and assume that the sensitivity is about 0.9, the width of the 95% CI for the estimated sensitivity would be as large as 0.44.
Inclusion Criteria:
Exclusion Criteria: