**9. Classifiers**

For classification over the regions of interest, detected with VBM, texture information is very useful since the measurement of texture requires statistical analysis to determine how voxel intensity values are distributed. Texture measurement involves the computation of the first-order and second-order statistics of the regions of interest.

The number of features extracted from one atlas is very large because of the number of regions of interest, the number of directions for second-order statistics features (co-occurrence matrix), number of views (sagittal, axial, coronal), number of different first-order statistics features, and number of second-order statistics features. Thus, *Principal Component Analysis* and *Wrappers* were applied to detect the most relevant features for the classification of regions of interest.

## **10. Regions of interest**

From the results of applying VBM to brain MRI, it has been found that regions of interest for PD detection in men are the *basal ganglia*, *brainstem*, *fourth ventricle*, *lateral ventricle*, *cerebellum*, *frontal lobe*, *temporal lobe*, *putamen,* and *thalamus*. The generation of signals for involuntary movement and instincts is generated within the putamen and thalamus. Other regions of interest lie in the upper cortex, which is related to brain functions such as reasoning, decision making. On the other hand, the application of VBM to female brain MRI shows that regions of interest, for PD detection in women, are *occipital lobe*, *basal ganglia*, a small part of the *cerebellum*, *frontal lobe*, *thalamus*, *brainstem*, and *temporal gyrus*. The last three regions are

associated with visual stimuli processing and spatial awareness. Regions of interest, within the cortex area, are not as large as those in men. These results are significant since most works, for automated PD detection, have been focused on the *striatum region* of the brain to detect damage. Another finding is that regions of interest in men are bigger than those in women, which agrees with medical findings that state that men are more prone to PD than women by almost twice. It is also found that the number of regions of interest are more in women than in men and that regions of interest, in men and women, are generally scattered over the same brain zones. Regions of interest, in men, are found within areas where more and smaller regions of interest for women occur. The number of selected features for PD detection in women is more reduced than in men.

Another finding from Solana et al. [6] is that the regions of interest from which most features are selected for PD detection, vary if the image is acquired with a different magnetic field. When the scanner uses 1.5-T for obtaining the MRI images, the features from the striatum region of the brain were chosen for the classification algorithm. On the other hand, when 3-T MRI are analyzed, features from regions like the primary *somatosensory cortex*, the *cerebellum*, and *temporal lobe* are selected as it is shown in **Figures 3** and **4**. Detection of PD with MRI achieves good performance with both genders and magnetic fields. When classifying female patients' MRI, accuracies of 96.77% and 93.28% for 1.5 and 3 T respectively. For male patients' MRI, excellent results were obtained, with 99.01% and 95.56% accuracy for 1.5 and 3 T respectively. **Table 2**, shows the results obtained by different methods in recent years, and how they compare to the proposed work.

### **Figure 3.**

*The most relevant regions of interest for PD detection in 1.5 and 3 T MRI of female patients, are highlighted with colors red, yellow, and green.*

### **Figure 4.**

*The most relevant regions of interest for PD detection in 1.5 and 3 T MRI of male patients, are highlighted with colors red, yellow, and green.*

*Analysis of Voice and Magnetic Resonance Images to Assist Diagnosis of Parkinson's Disease… DOI: http://dx.doi.org/10.5772/intechopen.99973*


### **Table 2.**

*A comparison between different works on PD detection based on MRI.*
