Population Based Image Imputation.

Area of Science:

• Medical Imaging
• Computational Anatomy
• Artificial Intelligence in Medicine

Background:

• Clinical brain MRI scans often have large inter-slice spacing due to acquisition constraints, resulting in sparse data.
• This sparsity limits the effectiveness of standard computational analysis algorithms, hindering medical research and diagnostics.
• Existing specialized algorithms for sparse data lack generalizability across different medical imaging applications.

Purpose of the Study:

• To develop a method for generating high-resolution, anatomically plausible brain MRI images from sparse clinical scans.
• To enable the application of existing high-resolution image analysis algorithms to undersampled clinical MRI data.
• To overcome limitations imposed by large inter-slice spacing in clinical brain MRI.

Main Methods:

• A novel algorithm was developed to capture fine-scale anatomical similarity across subjects in clinical image collections.
• The algorithm effectively fills in missing data in scans with large slice spacing.
• The model leverages anatomical priors from large clinical datasets to reconstruct missing information.

Main Results:

• The proposed method significantly outperforms current upsampling techniques for sparse MRI data.
• Generated images are anatomically plausible and consistent with acquired clinical brain MRI scans.
• The algorithm successfully enhances the quality of undersampled MRI scans.

Conclusions:

• The developed algorithm facilitates subsequent computational analysis of clinical brain MRI scans previously not possible.
• This approach promises to unlock the wealth of information in large clinical image databases.
• The method offers a generalizable solution for enhancing sparse medical imaging data.