Robust Inter-Modality Multi-Atlas Segmentation for PACS-based DTI Quality Control.

Area of Science:

• Medical Imaging Analysis
• Neuroimaging
• Computational Anatomy

Background:

• Anatomical context is crucial for medical image interpretation.
• Current segmentation methods often require standardized imaging sequences, limiting their application across diverse data types.
• Expanding Picture Archive and Communication System (PACS) archives necessitate generalizable segmentation approaches.

Purpose of the Study:

• To develop a generalizable, automated cross-modality segmentation method.
• To address the challenge of segmenting unlabeled imaging modalities, specifically B0 images in diffusion tensor imaging (DTI).
• To enable informed structure-wise noise estimation for fractional anisotropy (FA) measurements in DTI.

Main Methods:

• A multi-tier multi-atlas segmentation framework was proposed.
• Label information was transferred from labeled T1-weighted MRI data to unlabeled B0 DTI data.
• The approach facilitates automated, cross-modality label transfer.

Main Results:

• The framework successfully enabled automated, generalizable cross-modality segmentation.
• The method allowed for the construction of informed structure-wise noise estimates for DTI's fractional anisotropy (FA) measurements.
• Demonstrated utility in quality control for DTI images.

Conclusions:

• The proposed multi-tier multi-atlas segmentation framework provides a generalizable solution for segmenting unlabeled imaging modalities.
• This methodology is applicable beyond DTI quality control, supporting applications with limited atlas availability.
• The automated label transfer enhances the interpretation and utility of diverse medical imaging data.