An improved computational model for classification of 3D spatio temporal FMRI data
3D spatial temporal functional magnetic resonance imaging (fMRI) for classification has gained wide attention in the literature to be applied in the application of data mining techniques. Similarly, Spiking Neural Networks (SNN) has successfully applied in many problems to process and classify fMRI...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/522/1/24p%20SHAZNOOR%20SHAKIRA%20BINTI%20SAHARUDDIN.pdf http://eprints.uthm.edu.my/522/2/SHAZNOOR%20SHAKIRA%20BINTI%20SAHARUDDIN%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/522/3/SHAZNOOR%20SHAKIRA%20BINTI%20SAHARUDDIN%20WATERMARK.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | 3D spatial temporal functional magnetic resonance imaging (fMRI) for classification has gained wide attention in the literature to be applied in the application of data mining techniques. Similarly, Spiking Neural Networks (SNN) has successfully applied in many problems to process and classify fMRI data. However, the network still has a drawback in terms of processing noise, redundant and irrelevant features especially in fMRI data. To an extent, standard machine learning techniques has effectively process and classify fMRI data. Although, these techniques are only best at dealing spatial data, which completely neglect the temporal information inside the data. In order to achieve higher classification accuracy, there is a need to filter out noise from the dataset. Studies have shown that the presence of noise in the data effects the classification process thereby reducing the classification accuracy. In this study, the feature selection technique has been used as a filter at the pre-processing part of the dataset. Thus, this study proposed a feature selection technique called iReliefF to overcome the complexity in selecting the important features in fMRI dataset. This technique has been trained and tested by using StarPlus dataset. Based on the obtained results, the new computational model with proposed method iReliefF has shown better performance by achieving 85% accuracy compared to the existing model which is 80%. Therefore, it can be concluded that the proposed iReliefF has achieved reasonable accuracy and is very effective as well as ideal for fMRI dataset. |
|---|