Radiomic Features Extracted from Magnetic Resonance Imaging to Identify Arteriopathy in a Humanized Mouse Model of Sickle Cell Anemia and Sickle Cell Trait.

PURPOSE: Sickle cell disease (SCD) is a hereditary blood disorder that increases stroke risk in children. Previously, we showed that SCD caused accelerated arteriopathy in a sickle cell transgenic mouse model compared to sickle cell trait controls. We sought to determine whether radiomics analysis of label-free carotid artery magnetic resonance angiography (MRA) can differentiate between SCD mice (SS) from heterozygous sickle cell trait control mice (AS). Radiomics analysis of MRI data was used to extract quantitative imaging features, then tested for discrimination between SS and AS mice.

METHODS

MRA scans of Townes sickle cell transgenic mice at one and three months of age were completed. 112 radiomic features extracted from segmented carotid artery images using PyRadiomics software were used to develop models predictive of SCD genotypes.

RESULTS

At one month of age, four radiomics features yielded accuracy of 74%. At three months, a single feature achieved 76% accuracy. Analysis of MRA-derived arterial morphology confirmed that incorrectly identified mice carotid arteries resembled the incorrectly predicted genotype: larger luminal areas for AS and smaller luminal areas for SS, reflecting how biological variability of SCD impacted radiomic feature predictions.

CONCLUSION

This study demonstrates feasibility of radiomics in discriminating arterial features between SCD and control mice, and supports radiomics as a non-invasive imaging analytical approach to characterize arterial remodeling in preclinical SCD models, motivating future translational studies linking imaging features to vascular pathology.