Diffusion Bubble Model: A novel MRI approach for detection and subtyping of neonatal punctate white matter lesions

Diffusion magnetic resonance imaging, particularly diffusion tensor imaging (DTI), is an indispensable non-invasive tool for visualizing brain structure and detecting injuries by tracking water molecule motion. However, DTI may overlook subtle microstructural alterations due to its oversimplified model. In this study, we introduced the Diffusion Bubble Model (DBM), a spectrum-based framework that decomposes each voxel's signal into a continuum of isotropic "bubbles" after the anisotropic tensor adjustment, thereby capturing a spectrum with continuous range of restriction levels. From the resulting isotropic-diffusion spectrum we derive metrics representing the spectrum and free-water of the tissue voxel. We applied DBM to diffusion data from 20 infants with punctate white-matter lesions (PWMLs) in the optic radiation and compared lesion regions with contralateral regions as well as matched controls. DBM segregated the lesions into two phenotypes that DTI could not differentiate: wet-type (N=10), showing up to +155.0% elevated free water versus control (+125.1%vs. contralateral), and dry-type (N=10), with -68.4% less free water compared to contralateral and no difference versus controls. Notably, wet-type lesions exhibited stronger slow-diffusion shifts on DBM (-37.6% in the 1/4 area line, -52.7% in left FWHM) than changes in mean diffusivity (-30.3%) from DTI. These findings suggest that DBM can reveal microstructural heterogeneity invisible to conventional DTI, offering a promising tool for refined characterization and monitoring of neonatal brain injury.