Related Experiment Videos
Normal brain volume measurements using multispectral MRI segmentation
M Vaidyanathan1, L P Clarke, C Heidtman
1Department of Radiology, University of South Florida, Tampa, Florida, USA.
Magnetic Resonance Imaging
|January 1, 1997
Summary
The semisupervised fuzzy c-means (SFCM) method offers reproducible brain tissue and cerebrospinal fluid volume measurements. SFCM demonstrated superior stability and accuracy compared to the k-nearest neighbor (kNN) classifier.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Computational Biology
Background:
- Accurate and reproducible measurement of brain tissue and cerebrospinal fluid (CSF) volumes is crucial for neurological research and clinical diagnosis.
- Segmentation methods are essential for quantifying these volumes from medical images, but their reproducibility can be affected by various factors.
Purpose of the Study:
- To evaluate the performance and reproducibility of a supervised k-nearest neighbor (kNN) classifier and a semisupervised fuzzy c-means (SFCM) clustering method for segmenting normal brain tissues and CSF.
- To assess the stability of these segmentation methods across different conditions, including operator variability, imaging session repeats, subject variations, and imaging parameters.
Main Methods:
- The study compared a k-nearest neighbor (kNN) classifier and a semisupervised fuzzy c-means (SFCM) clustering method for brain tissue and CSF segmentation.
- Segmentation performance was evaluated based on reproducibility during repeat imaging sessions, inter-subject variability, and operator influence on training data selection.
- Variability was assessed across different tissue types and imaging parameters.
Main Results:
- The semisupervised fuzzy c-means (SFCM) method exhibited high reproducibility, with less than 3% variability during repeat imaging sessions.
- Intraobserver and interobserver reproducibility for SFCM were below 4% for soft tissues and 6% for CSF.
- The k-nearest neighbor (kNN) segmentation method showed higher variability compared to the SFCM method across all evaluated conditions.
- Absolute brain matter and CSF volumes varied significantly between subjects (9% to 13%).
Conclusions:
- The semisupervised fuzzy c-means (SFCM) clustering method provides a stable and reproducible approach for quantifying normal brain tissue and cerebrospinal fluid volumes.
- SFCM demonstrates superior performance over the k-nearest neighbor (kNN) classifier in terms of reproducibility and stability in neuroimaging segmentation.
- The findings highlight the importance of method selection for reliable volumetric analysis in neuroscience research.