Neural network algorithm for the early detection of Parkinson's disease from blood plasma by FTIR micro-spectroscopy

Abstract

Parkinson's disease (PD) is a progressive bradykinetic disorder. Diagnosis of PD is entirely clinical, no biochemical parameters exist to diagnose PD. The diagnostic problem associated with the early detection has led to an interest in the development of spectroscopic based diagnostic techniques. Fourier-transform infrared micro-spectroscopy analysis (FTIR) was applied to analyze human plasma samples of 69 healthy and 61 drug-naive PD patients in order to detect spectral parameters, which serve as biomarkers for monitoring and identification of PD. The analysis showed the bands at 1078, 1169 and 1244 cm −1 corresponding to carbohydrates, significantly increased in all tested patient samples ( p ≤ 0.01). Several other spectral regions that attribute to amino acids, lipids and proteins indicate the unique detection of disease stages. Cluster analysis of variable regions provided excellent grouping of the healthy and the patient samples, which correlate completely with the clinical diagnosis. For automatic detection, artificial neural network (ANN) was implemented on the variable regions showed 96.29% accuracy in the detection of disease progression. These parameters could be used, as a basis for developing a spectral method for detecting PD. Execution of neural network will be useful in clinical screening and rapid detection of PD.