Peter Hirschmann ECE 539 1 Abstract Diagnosis of Parkinson’s Peter Hirschmann 12/14/2010Parkinson’s disease is a degenerate disease which affects the motor and cognitive skills. There is currently no cure, which makes treatment and early diagnosis important. The focus of this project is to test the functionality of speech problems as a diagnosable symptom for Parkinson’s. Current methods rely on a doctor’s judgment which varies with the physician. Hopefully, this would standardize the way Parkinson’s is diagnosed, and make early detection easier. I chose three classification methods to test, Polynomial Model, Maximum Likelihood, and Nearest Neighbor. I found data pertaining to patient’s voice on UCI Machine Learning Repository which has twenty-two feature vectors and one label vector. Using the data, the success of each method can be tested, in order to determine its efficacy. The results of the project have shown that the Polynomial Model is not a good indicator for Parkinson’s. The LOO method performs well, but the testing and training do not. The Maximum Likelihood method has a classification rate of a little over 80% at best, and with LOO, a classification rate of around 93%. Nearest Neighbor performed the best out of all the methods. Using 1% of the data for testing and rest for training, each maximum neighbor has a 100% classification rate. Using the LOO method, a nearest neighbor of K=1, correctly predicts the label of all the data. The conclusion I have drawn from this is the Nearest Neighbor method is almost good enough to be used a standard clinical test. It would certainly be useful as a pre-screen for anyone who wanted to take it. However, there are not enough results to prove that this test could be used as a medical standard. As a medical standard, this test would be used on tens of thousands of people, 195 samples from 31 people is not enough to accurately test its ability. It is strong encouragement for further tests, and if successful, would provide a clear test for Parkinson’s.Bibliography Anders Björklund, S. B. (2003). Neural transplantation for the treatment of Parkinson’s disease. THE LANCET Neurology Vol 2 , 2:437-445. Diagnosis. (2010). Retrieved from Parkinson's Disease Foundation: http://www.pdf.org/en/diagnosis Frank, A. &. (2010). Retrieved from UCI Machine Learning Repository: http://archive.ics.uci.edu/ml Jankovic, J. (n.d.). Parkinson’s disease: clinical features and diagnosis. Neurol Neurosurg Psychiatry . Little MA, M. P. (2007). 'Exploiting Nonlinear Recurrence and Fractal Scaling Properties for Voice Disorder Detection'. BioMedical Engineering OnLine
View Full Document