ITK-SNAP: An interactive tool for semi-automatic segmentation of multi-modality biomedical images

Area of Science:

• Medical image analysis
• Computational biology
• Radiology

Background:

• Quantitative analysis of biomedical images often necessitates segmentation.
• Multi-modality imaging, especially MRI, is increasingly prevalent.
• Semi-automatic segmentation offers a balance between automation and expert guidance for complex cases.

Purpose of the Study:

• To extend ITK-SNAP for semi-automatic segmentation of multi-modality imaging data.
• To develop a method that simultaneously utilizes information from all available imaging modalities.
• To evaluate the approach for segmenting high-grade glioblastoma.

Main Methods:

• Implementation of new extensions within the ITK-SNAP tool.
• Integration of Random Forest classifiers trained by user input (brushstrokes).
• Combination with the active contour segmentation algorithm for a hybrid approach.

Main Results:

• The developed extensions enable semi-automatic segmentation of multi-modality datasets.
• The approach effectively utilizes information across different imaging modalities.
• Evaluation in high-grade glioblastoma segmentation demonstrated the method's utility.

Conclusions:

• The enhanced ITK-SNAP tool provides a valuable solution for semi-automatic multi-modality image segmentation.
• This approach can improve efficiency and accuracy in medical image analysis tasks.
• The method shows promise for clinical applications, particularly in neuro-oncology.