An Enhanced Hybrid Machine Learning Model for Diagnosis of Parkinson's Disease

Parkinson's Disease (PD) is a progressive disease caused due to degeneration of nerve cells in brain which is associated with movement control. These nerve cells die or become damaged, resulting in a loss of ability to produce an important chemical called dopamine. Currently, there is no any cure for this disorder and also no way to reverse the condition. Therefore, it is very essential to identify the presence of disease at its initial stage rather than allowing it to develop into a more lethal threat. In order to predict the Parkinson's in its earlier stage, we have developed an efficient machine learning model using Random Forest and XGBoost classifier algorithms. The model is trained with the help of data collected from various hospitals with the help of neurologists. To predict whether the individual is affected by PD or healthy, we use a videocam to monitor the individual and SensorTile - sensor node, which contains accelerometer, Gyroscope and Magnetometer incorporated on a board with built-in transmitter. All the data are transmitted in real-time and stored in a PC for data analysis. With the data collected from the individual, the trained ML model is evaluated and the disease can be detected on comparison with threshold value. If PD is detected, the patient is provided with a wearable glove which aids in performing their day-to-day tasks.