Related Experiment Videos
Selection of MR images for automated segmentation
K J McClain1, Y Zhu, J D Hazle
1University of Texas M.D. Anderson Cancer Center, Department of Diagnostic Radiology, Houston 77030, USA.
Journal of Magnetic Resonance Imaging : JMRI
|September 1, 1995
Summary
This study found that three magnetic resonance (MR) images are sufficient for accurate brain tumor tissue segmentation. Using statistical analysis, researchers identified the optimal number and type of MR images for segmenting normal brain, edema, necrosis, and active tumor tissues.
Area of Science:
- Medical Imaging
- Radiology
- Biomedical Engineering
Background:
- Magnetic Resonance (MR) imaging offers excellent tissue contrast, making it suitable for multispectral segmentation.
- Current segmentation methods often rely on a limited number of MR images, typically dual-echo series.
- The availability of additional MR images suggests potential for improved segmentation accuracy.
Purpose of the Study:
- To determine the optimal type and number of MR images for segmenting brain tumor-associated tissues.
- To evaluate the effectiveness of pattern recognition methods for image selection in segmentation.
- To establish a systematic approach for optimizing MR image selection for tissue segmentation.
Main Methods:
- Analysis of MR images from 40 patients with brain tumors.
- Application of pattern recognition techniques, including feature selection and feature extraction.
- Validation of segmentation results through visual examination.
Main Results:
- Three MR images from the same slice location were found to be adequate for accurate tissue segmentation.
- Feature selection and extraction measures confirmed the sufficiency of three images.
- Visual inspection corroborated the quantitative findings across all patient datasets.
Conclusions:
- A systematic approach using statistical analysis (feature selection and extraction) can identify the optimal MR image set for segmentation.
- Three MR images provide sufficient information for segmenting critical brain tumor tissues.
- This method enhances the efficiency and accuracy of brain tumor segmentation using existing MR data.