Radiomics feature similarity: A novel approach for characterizing brain network changes in patients with behavioral variant frontotemporal dementia

Network modeling is increasingly used to study brain alterations in neurological disorders. In this study, we apply a novel modeling approach based on the similarity of regional radiomics feature to characterize gray matter network changes in patients with behavioral variant frontotemporal dementia (bvFTD) using MRI data. In this cross-sectional study, we assessed structural 3 T MRI data from twenty patients with bvFTD and 20 cognitively normal controls. Radiomics features were extracted from T1-weighted MRI based on cortical and subcortical brain segmentation. Similarity in radiomics features between brain regions was used to construct intra-individual structural gray matter networks. Regional mean connectivity strength (RMCS) and region-toregion radiomics similarity were compared between bvFTD patients and controls. Finally, associations between network measures, clinical data, and biological features were explored in bvFTD patients. Relative to controls, patients with bvFTD showed higher RMCS values in the superior frontal gyrus, right inferior temporal gyrus and right inferior parietal gyrus (FDR-corrected p < 0.05). Patients with bvFTD also showed several edges of increased radiomics similarity in key components of the frontal, temporal, parietal and thalamic pathways compared to controls (FDR-corrected p < 0.05). Network measures in frontotemporal circuits were associated with Mini-Mental State Examination scores and cerebrospinal fluid total-tau protein levels (Spearman r > |0.7|, p < 0.005). Our study provides new insights into frontotemporal network changes associated with bvFTD, highlighting specific associations between network measures and clinical/biological features. Radiomics feature similarity analysis could represent a useful approach for characterizing brain changes in patients with frontotemporal dementia.