Parkinson-s disease is regarded as a universal health issue and progressive neurodegenerative disorder in the recent health era. Diagnosis of Parkinson-s disease based on the medical observations and assessment of clinical signs includes characterization of varied motor symptoms. Owing to the non-motor symptoms, early diagnosis of Parkinson-s disease is challenging.
Although the medicaments to cure Parkinson-s disease haven-t been developed, the early detection and treatment directly increase the lifetime of the afflicted patients. Currently, machine learning (ML) techniques are applied to diagnose Parkinson-s disease through the resting and activation states of Electroencephalography (EEG) of speech and gait signals.
Promising Machine Learning Techniques for Parkinson-s Disease Diagnosis: Various machine learning algorithms are utilized to improve the accuracy of Parkinson-s disease prediction and detection, which are highlighted here;
• Recently, Artificial Neural Networks (ANN) and clustering algorithms detect Parkinson-s disease using sound measurements.
• K-Nearest neighbor classifier (K-NN) and other classifiers are utilized to find Parkinson-s disease infection.
• Support Vector Machines (SVM) recently accomplished great precision and accuracy in predicting Parkinson-s disease.
• Multilayer perceptron (MLP) detects Parkinson-s disease using dysphonic measures and clinical scores.
• Deep neural networks like Convolutional Neural Network (CNN) use EEG signals to diagnose Parkinson-s disease.
Future Scopes of Machine Learning for Parkinson-s Disease Diagnosis: Only a limited number of research works on machine learning-based Parkinson-s disease diagnosis are investigated. Thus this research scope needs attention to support the healthcare domain.
• More research on machine learning for Parkinson-s disease diagnosis is requisite to address the issues of open to error class labeling in supervised learning.
• Machine learning-based Parkinson-s disease detection faces many datasets issues such as over-fitting, low generalizability, and data insufficiency, which need to contend with in the future.
• Parkinson-s disease diagnosis with multimodal data should be developed using machine learning to improve the diagnosis and treatment of Parkinson-s disease.
• Evolutionary algorithms such as Genetic algorithms and Extreme Learning Machine for Parkinson-s disease detection and classification need to be utilized in the future for more accurate diagnosis.
• The incorporation of machine learning-based Parkinson-s disease diagnosis applications is classified into a clinical decision support system to aid in the healthcare industry.