Brain Tumor Segmentation of T1w MRI Images Based on Clustering Using Dimensionality Reduction Random Projection...

Area of Science:

• Medical imaging analysis
• Computational neuroscience
• Machine learning in healthcare

Background:

• Early brain tumor diagnosis is crucial for improving patient survival rates.
• Magnetic resonance imaging (MRI) with segmentation algorithms is a key diagnostic tool.
• High-dimensional MRI data presents computational and storage challenges, necessitating dimensionality reduction techniques like random projection technique (RPT).

Purpose of the Study:

• To evaluate and compare the effectiveness of RPT and principal component analysis (PCA) for dimensionality reduction in T1-weighted MRI brain tumor segmentation.
• To assess the performance of K-means and fuzzy c-means (FCM) clustering algorithms on MRI images processed by both PCA and RPT.

Main Methods:

• T1-weighted MRI images were subjected to dimensionality reduction using PCA and RPT.
• K-means and FCM clustering algorithms were applied to the processed images for brain tumor detection.
• The study analyzed image sizes of 512x512, 256x256, 128x128, and 64x64 to assess method performance across different resolutions.

Main Results:

• RPT demonstrated superior performance over PCA across all clustering techniques, showing lower average reconstruction errors, Euclidean distances, and segmentation errors.
• Fuzzy c-means (FCM) coupled with RPT achieved the best clustering performance.
• The effectiveness of RPT was consistent across all tested MRI image resolutions.

Conclusions:

• Random projection technique (RPT) is a more effective method than principal component analysis (PCA) for brain tumor segmentation in MRI.
• The combination of RPT and fuzzy c-means (FCM) provides optimal clustering performance for brain tumor detection using MRI data.
• RPT offers significant advantages in managing high-dimensional MRI data for improved diagnostic accuracy and efficiency.