A spatiotemporal-based scheme for efficient registration-based segmentation of thoracic 4-D MRI

Area of Science:

• Medical Imaging
• Radiotherapy
• Pulmonology

Background:

• Dynamic 4-D MR imaging is crucial for studying lung and tumor motion in respiratory diseases and radiotherapy.
• Manual segmentation of 4-D MRI data is time-consuming and prone to user variability.
• Automated segmentation methods are needed for efficient quantitative analysis of 4-D thoracic MRI.

Purpose of the Study:

• To develop an automated 4-D registration-based segmentation scheme for thoracic 4-D MR lung images.
• To improve the efficiency and accuracy of segmenting anatomical structures in dynamic lung MRI.
• To facilitate rapid visualization and tracking of lung and tumor motion.

Main Methods:

• Proposed an automated 4-D registration-based segmentation scheme utilizing spatiotemporal information.
• Applied the scheme to segment thoracic 4-D MR lung images.
• Compared the computational efficiency and segmentation accuracy against direct registration-based segmentation.

Main Results:

• The proposed scheme achieved comparable segmentation accuracy to direct registration-based methods.
• Reduced computational cost by up to 95% compared to direct application on 4-D datasets.
• Enabled efficient 3-D/4-D visualization and potential tumor tracking during radiation delivery.

Conclusions:

• The developed automated 4-D registration-based segmentation scheme is efficient and accurate for thoracic 4-D MR lung images.
• This method significantly reduces computational load, aiding quantitative analysis and visualization of pulmonary and tumor dynamics.
• Facilitates advancements in radiotherapy planning and respiratory disease research through improved motion analysis.