Article Preview
TopIntroduction
Neuropsychiatric disorders such as Parkinson’s disease, epilepsy, multiple sclerosis, and Alzheimer’s disease are brain disorder which affects the brain functioning of humans in one way or another. Parkinson's disease also called shaking palsy is the 2nd most common neurodegenerative disorder after Alzheimer’s disease (Kotsavasiloglou et al., 2017). Low production of a brain chemical known as dopamine, which functions as a neurotransmitter in the substantia nigra is the root cause of Parkinson’s. Parkinson's symptoms appear when brain cells that produce dopamine begin to die but the reason behind the deterioration of these cells is still unknown (Lamba et al., 2020). As Parkinson’s is non-curable, an early diagnosis is the only way to ameliorate the well being of the patient with proper medication and exercise as prevention is not possible. Commonly used drugs by which Parkinson's symptoms can be controlled are levodopa, carbidopa, and syndopa. The prime function of these drugs is to stimulate the cells to generate extra dopamine. The dosage depends on the severity of the disease (Ma et al., 2020).
In the starting, the symptoms of Parkinson’s disease are mild and sometimes go unnoticeable but as the disease progresses the symptoms go severe. The symptoms are majorly categorized into motor and non-motor (Ascherioi and Schwarzschild, 2016). The common motor symptoms are muscle stiffness, tremor or shaking in hand/ arms of one side of the body or both sides, slowness during walking called Bradykinesia, steps become shorter during walking, changes in handwriting, and changes in speech. The non-motor symptoms are sweating, weight, sleep, sexual dysfunction, losses of taste/smell, bowel movements, appetite, drooling, fatigue, hallucinations. Dysphonia (difficulty or changes in speech) is one of the predominant symptoms that appear in more than ninety percent of Parkinson’s patients (Bhat et al., 2018).
It is an arduous task to diagnose the onset of Parkinson’s at an early stage because of the indistinct nature of its symptoms. The patients have to visit health care centres repeatedly for the diagnosis of the disease by a neurologist or movement order specialist, where their health history and certain scans are reviewed. As this disease mostly affects the population of around 60 years and above, visiting health care centres can be a strenuous task. With technological advancement, Computer-assisted technologies are gaining momentum in the healthcare field. Moreover, they are remotely accessible and highly accurate. Now doctors and healthcare professionals may seek the assistance of such computer-assisted technologies in diagnosing Parkinson’s disease (Lamba et al., 2021a).
There are numerous ways that the researchers have already applied to diagnose Parkinson’s by implementing deep learning and machine learning methods. They have diagnosed Parkinson’s using handwritten images by extracting features from raw time series data or by applying deep learning techniques directly on the handwritten images (Lamba et al., 2021a), speech signals (Behroozi and Sami, 2016), Freezing of gait (Ertuğrul et al., 2016), EEG signals (Oh et al., 2019), EMG signals (Loconsole et al., 2019), MRI images (Sivaranjini and Sujatha, 2019) and SPECT images. Dysphonia, changes in speech is the earliest symptom that appears in Parkinson’s patients can effectively be used for ascertaining whether a patient is suffering from Parkinson’s or not.
Researchers have proposed various decision support systems (DSS) for Parkinson’s diagnosis which include feature extraction, pre-processing, feature reduction, and classification stages by using machine learning techniques (Lamba et al., 2020). In this paper, the authors have proposed the MIRFE-XGBoost system to diagnose Parkinson’s using machine learning methods. The MIRFE feature selection approach has contributed to increasing the accuracy of the system. The major motivation behind proposing MIRFE-XGBoost system is to improve Parkinson's patient's life through early diagnosis. However there are other reasons, such as the need for an automated Parkinson's diagnosis system, and there is scope to increase accuracy in existing systems.
The prime contributions are