A deep learning approach for prediction of Parkinson’s disease progression

Abstract

This paper proposes a deep neural network (DNN) model using the reduced input feature space of Parkinson's telemonitoring dataset to predict Parkinson's disease (PD) progression. PD is a chronic and progressive nervous system disorder that affects body movement. PD is assessed by using the unified Parkinson's disease rating scale (UPDRS). In this paper, firstly, principal component analysis (PCA) is employed to the featured dataset to address the multicollinearity problems in the dataset and to reduce the dimension of input feature space. Then, the reduced input feature space is fed into the proposed DNN model with a tuned parameter norm penalty (L2) and analyses the prediction performance of it in PD progression by predicting Motor and Total-UPDRS score. The model's performance is evaluated by conducting several experiments and the result is compared with the result of previously developed methods on the same dataset. The model's prediction accuracy is measured by fitness parameters, mean absolute error (MAE), root mean squared error (RMSE), and coefficient of determination (R 2 ). The MAE, RMSE, and R 2 values are 0.926, 1.422, and 0.970 respectively for motor-UPDRS. These values are 1.334, 2.221, and 0.956 respectively for Total-UPDRS. Both the Motor and Total-UPDRS score is better predicted by the proposed method. This paper shows the usefulness and efficacy of the proposed method for predicting the UPDRS score in PD progression.

A Review of Deep-Learning-Based Approaches for Attenuation Correction in Positron Emission Tomography

“…Artificial intelligence and machine learning are two of the most widely investigated mathematical and engineering techniques in the biomedical engineering field [181][182][183][184][185][186][187]. In recent decades, various techniques based on artificial intelligence and machine learning have been applied to nuclear medicine images.…”

The Dawn of the Human-Machine Era: A forecast of new and emerging language technologies

“…Features that affect articulation, respiration and prosody are used to track the progress of multiple sclerosis (Noffs et al 2018). The automatic detection of Parkinson's disease from speech and voice has already entered a mature stage after more than a decade of active research, and could eventually supplement the neurological and neuropsychological manual examination to diagnose (Shahid & Singh 2020).…”

Regression Networks for Neurophysiological Indicator Evaluation in Practicing Motor Imagery Tasks

“…Thus, related to motor evoked potential time series, nonlinear models (like random forests) can achieve significantly better prediction performance than a linear one (or logistic regression) [ 35 ]. In particular, machine learning analysis in nonlinear regression is extensively employed under two deep learning solutions [ 36 , 37 ]: ( i ) utilizing an ensemble of deep networks that suffer from larger computational complexity and ( ii ) transforming a single nonlinear regression hypothesis to a robust loss function that is jointly optimizable with the deep network usually in terms of the mean square error. However, the generalization ability is a major concern in developing deep regression models and computational complexity and hardware consumption [ 38 ].…”