Joint Intensity Fusion Image Synthesis Applied to Multiple Sclerosis Lesion Segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Neurology

Background:

• Accurate segmentation of Multiple Sclerosis (MS) lesions is crucial for disease monitoring and treatment.
• Existing segmentation methods can be limited by image quality and inter-subject variability.

Purpose of the Study:

• To develop and evaluate a novel image synthesis technique for enhancing MS lesion segmentation.
• To integrate synthesized images as features into a classification model for improved lesion detection.

Main Methods:

• A new image synthesis method, joint intensity fusion (JIF), was developed using a library of registered and normalized atlases.
• JIF synthesizes images by spatially weighting atlas information, applying the joint label fusion (JLF) framework directly to MRI signal.
• Synthesized JIF images were incorporated as features into the OASIS classifier for lesion segmentation.

Main Results:

• The integration of JIF synthesized images improved lesion segmentation accuracy.
• The Dice-Sorensen coefficient increased by 0.0462 ± 0.0050 compared to standard segmentation methods.
• Improvements were observed across multiple subjects and training/testing folds.

Conclusions:

• Joint intensity fusion (JIF) offers a promising method for generating synthetic MRI data to augment existing datasets.
• This approach enhances the performance of multiple sclerosis lesion segmentation models.
• The direct application of JLF to MRI signal represents a significant methodological advancement.