Application of artifact removal algorithms in walking signals for diagnosis of parkinson's disease

Parkinson's Disease (PD) is a neurological disease that progresses over time. People affected by this condition experience a shortage of dopamine, a chemical found in the nerve cells of the brain that plays a vital role in coordinating people's movement. As PH progresses, the sick person cannot control their movements normally. Individuals affected by Parkinson's disease show distinctive symptoms such as gait disturbances and episodes of tremors at different stages of the disease. Gait disorders are manifested in the form of decreased walking speed and small steps. Gait disturbance is progressive in all stages of the disease. Gait disturbance is due to muscle stiffness, decreased strength, abnormal rhythmicity, asymmetry of the left and right parts of the body, and abnormal scaling of stride lengths. The power distribution on the foot varies for PH and healthy people. The characteristic features of these power distributions can be used to distinguish between healthy and sick individuals. Gait analysis is often used as part of routine clinical testing to evaluate walking performance. In this study, a new approach is proposed to differentiate individuals with Parkinson's disease from healthy individuals by gait signals. Gait and trembling signs are widely used in diagnosing PH. Gait signals are usually measured multi-channel. When measuring signals on one channel, other channels may have a noise effect. Therefore, CAR, MCAR and WCAR methods have been applied to the walking signs in order to remove possible noise in the signs. Statistical features were extracted from the generated clean signals after applying the LBP transform. These feature groups are classified by classification algorithms such as KNN, LR and RF. To test the proposed approach in this study, a total of 93 Parkinson's patients (PD) (mean age: 66.3 years; 63% male) and 73 healthy subjects (CO) for control purposes (mean age: 66.3 years; 55% male) were recruited from three different studies. includes gait measurements. The data set labeled "Ga" includes 113 (75 PH+38 Control) records obtained from 29 Parkinson's patients and 18 control subjects. The data set labeled "Ju" includes 129 (104 PH+25 Control) records obtained from 29 Parkinson's patients and 26 control subjects. Finally, the data set labeled as "Si" includes 64 (35PH+29 Control) records obtained from 35 Parkinson's patients and 29 control subjects. Looking at the results, the WCAR method, which is an adaptive method that cleans the noises best, was found. As for the classification method, the best classifier was observed as the KNN method. The best success rate was found as 92.96%. The study also examined which foot marks are effective in diagnosing PH. According to the data set used, it was seen that the left foot signs provided better distinguishing features in diagnosing PD. Vertical ground reaction force (VGRF) recordings of individuals were obtained by running the data set at normal, self-selected speeds for approximately 2 minutes on a flat surface. Under each foot of individuals, there are 8 sensors that measure power (in Newtons) as a function of time. The success rate observed by using the extracted features after applying filter methods to the recorded signals from each sensor was obtained as 84.02%. Sensor 5 is located in the middle of the soles of the feet. Therefore, it was seen that the signs obtained from the centers of the soles of the feet provided more effective signs to the signs in other regions. When the results were examined, it was seen that the noise removal methods in the signs increased the success in the diagnosis of PH.