User-Guided Segmentation of Multi-modality Medical Imaging Datasets with ITK-SNAP

Area of Science:

• Medical image analysis
• Computational imaging
• Machine learning in medicine

Background:

• ITK-SNAP is a key tool for 3D medical image segmentation.
• Previous versions primarily supported manual segmentation.
• Need for improved semi-automatic segmentation, especially for multi-modal data.

Purpose of the Study:

• To introduce and evaluate new semi-automatic segmentation features in ITK-SNAP.
• To enhance segmentation of multi-modality imaging datasets.
• To assess the performance against established benchmarks.

Main Methods:

• Development of decision forest classifiers for interactive training.
• Integration of probability maps into active contour evolution algorithms.
• Evaluation using the MICCAI 2013 Multi-Modal Brain Tumor Segmentation Challenge (BRATS) dataset.

Main Results:

• Semi-automatic segmentation accuracy is competitive with top BRATS challenge methods.
• ITK-SNAP results surpassed the second-best and fourth-best BRATS performers.
• Segmentation time was reduced by 2.5 to 5 times compared to manual methods.

Conclusions:

• New ITK-SNAP features significantly enhance semi-automatic segmentation capabilities.
• The tool provides accurate and efficient segmentation for brain tumors and other applications.
• Functionality demonstrates generalizability across different medical imaging domains.