Radiomics Reproducibility in MRI: Influence of Noise Reduction Filters

Radiomic analysis of Magnetic resonance imaging images offers significant potential for the characterization of neurodegenerative diseases. However, the reproducibility of radiomic features can be affected by the image processing step. In this study, the impact of field inhomogeneity correction and two different noise filters on the reproducibility of radiomic features extracted from longitudinal MRI was evaluated. The analysis was conducted on healthy regions in subjects without dementia and on pathological areas in dementia patients, comparing the same references also in healthy individuals. The results provide insights into the robustness of radiomic features in different contexts and the importance of image pre-processing to ensure a reliable analysis.