CT Radiomic Features of the Crystalline Lens and Association with Age, Hypertension and Cerebral White Matter Lesions

<b>Background:</b> Radiomic analyses have been extensively explored in oncologic imaging and more recently in neuroimaging. However, radiomic characterization of the crystalline lens using computed tomography has not yet been systematically investigated. <b>Methods:</b> In this retrospective study, semiautomatic segmentation of the eye lens on orbital CT was performed on 112 patients (mean age 48 ± 20 years, 38% female). After radiomics feature extraction, a Boruta feature selection approach based on the random forest algorithm was applied to select the most relevant radiomics features. Severity of white matter lesions were graded according to the Fazekas scale for each patient on axial non-contrast head CT. <b>Results:</b> In total, 17 important features were associated with age-related changes in the eye lens and three important radiomic features for the differentiation between patients with a Fazekas score > 1 and a control group. Significantly higher values were found in patients with a Fazekas score > 1 compared to the control group regarding all three features, "ClusterShade", "Skewness" and "DifferenceVariance" (<i>p</i> = 0.0006, 0.0023 and 0.0376, respectively), which are all measures of heterogeneity. No important radiomic features of the eye lens were confirmed between patients with and without hypertension. <b>Conclusions:</b> To the best of our knowledge, this is the first study to use CT-based radiomic analysis of the crystalline lens to detect differences among demographic or clinical groups with small vessel disease. The present results might help to expand the range of applications of radiomics regarding ophthalmic (patho-)physiology and suggest a possible new biomarker for systemic vascular diseases.