support vector machine (svm)

Parkinson’s disease (PD) is a complex, progressive neurodegenerative disorder known to negatively impair patient gait. Therefore, with gait and vertical ground reaction force (VGRF) data, an association can be made between the data and Parkinson’s disease.

Data from 146 participants; 93 with Parkinson’s disease and 73 without Parkinson’s disease was obtained from a PhysioNet database for use in this article. A Fourier Analysis and several support vector machine learning models were computed in MATLAB to classify whether an individual had Parkinson’s disease.

From the Fourier analysis, it was determined that a statistically significant difference was present between the VGRF data of individuals with and without Parkinson’s disease. Additionally, it was found that a Minimum Classification Error Optimized SVM machine learning model using Bayesian statistics was able to classify individuals with Parkinson’s disease using VGRF data at an accuracy of 67.1%, and sensitivity of 80.43%.

Therefore, it can be determined that vertical ground reaction force can predict Parkinson’s Disease with considerable accuracy which could be improved with an increased number of participants.

Schizophrenia (SZ) is a mental illness that impairs a person's mental capacity, emotionaldispositions, and personal and social quality of life. Manual SZ patient screening is timeconsuming,expensive, and prone to human mistakes. As a result, a autonomous, relativelyaccurate, and reasonably economical system for diagnosing schizophrenia patients isrequired. Machine learning methods are capable of learning subtle hidden patterns fromhigh dimensional imaging data and achieve significant correlations for the classificationof Schizophrenia. In this study, the diverse types of symptoms of the affected person areselected which have the weights assigned by cross-correlations and the model classifiesthe probability of schizophrenia in the person based on the highest weighted symptomspresent in the report of the patient using machine learning classifiers. The classificationis made by various classifiers in which the Support Vector Machine (SVM) gives thebest result. In the neuroscience domain, it has been one of the most popular machinelearningtools. SVM with Radial Basis Function kernel helps to distinguish betweenpatients and healthy controls with significant accuracy of 76% without normalization andPrincipal Component Analysis (PCA). The K nearest neighbor’s algorithm also with nonormalization and PCA showed an accuracy of 73% in predicting SZ which is remarkablyclose to the SVM given the small size dataset.

Electrocardiogram (ECG) is defined as an electrical signal, which represents cardiac activity. Heart rate variability (HRV) as the variation of interval between two consecutive heartbeats represents the balance between the sympathetic and parasympathetic branches of the autonomic nervous system.
In this study, we aimed to evaluate the efficiency of discrete wavelet transform (DWT) based features extracted from HRV which were further selected by genetic algorithm (GA), and were deployed by support vector machine to HRV classification.
In this paper, 53 ECGs including 3 different beat types (ventricular fibrillation (VF), atrial fibrillation (AF) and also normal sinus rhythm (NSR)), were selected from the MIT/BIH arrhythmia database. The approach contains 4 stages including HRV signal extraction from each ECG signal, feature extraction using DWT (entropy, mean, variance, kurtosis and spectral component β), best features selection by GA and classification of normal and abnormal ECGs using the selected features by support vector machine (SVM).
The performance of the classification procedure employing the combination of selected features were evaluated using several measures including accuracy, sensitivity, specificity and precision which resulted in 97.14%, 97.54%, 96.9% and 97.64%, respectively.
A comparative analysis with the related existing methods illustrates the proposed method has a higher potential in the classification of AF and VF. The attempt to classify the ECG signal has been successfully achieved. The proposed method has shown a promising sensitivity of 97.54% which indicates that this technique is an excellent model for computer-aided diagnosis of cardiac arrhythmias.

Functional magnetic resonance imaging (fMRI) is a popular method to probe the functional organizationof the brain using hemodynamic responses. In this method, volume images of the entire brain areobtained with a very good spatial resolution and low temporal resolution. However, they always sufferfrom high dimensionality in the face of classification algorithms. In this work, we combine a supportvector machine (SVM) with a self-organizing map (SOM) for having a feature-based classification byusing SVM. Then, a linear kernel SVM is used for detecting the active areas. Here, we use SOM forfeature extracting and labeling the datasets. SOM has two major advances: (i) it reduces dimension ofdata sets for having less computational complexity and (ii) it is useful for identifying brain regions withsmall onset differences in hemodynamic responses. Our non-parametric model is compared withparametric and non-parametric methods. We use simulated fMRI data sets and block design inputsin this paper and consider the contrast to noise ratio (CNR) value equal to 0.6 for simulated datasets.fMRI simulated dataset has contrast 1–4% in active areas. The accuracy of our proposed method is93.63% and the error rate is 6.37%.

In cataract surgery, the defected lens is replaced with an artificial intraocularlens (IOL). The refraction power of this lens is specified by ophthalmologistsbefore the surgery. There are different formulas that propose the IOL powerbased on corneal power and axial length. Six common formulas is used in thisstudy and the one which minimizes the postoperative error for a specific patienthave to be selected. Refraction is measured three times at most, during six month after surgeryand the best result is considered as postoperative refraction for each patient. Asupport vector machine (SVM) is used to classify the data to two groups based onaxial length and corneal power. Each class needs a formula with a specific tendencytoward stronger or weaker IOL lenses according to the feature vector. Experimental tests lead to a nearly diagonal confusion matrix whichsupports the performance of the proposed method strongly. Mean Absolute Error(MAE) is 0.47 which shows 6% decrease in postoperative refraction error comparedto the best reported result. In calculating IOL power, we expect stronger IOL powers for eyeshaving shorter axial length or weaker corneal power. In the contrary, a weaker IOLpower is expected for longer axial length and stronger corneal power. But experimentalresults show that for the first group, formulas proposing weaker powers winthe race for decreased postoperative refraction error while for the second group,formulas leading to stronger powers perform better. This shows that these formulasoverestimate and underestimate for marginal cases.