Validation and Optimisation of a White Matter Lesions Pipeline to be used in Large-Data Analysis in Parkinson's Disease

Abstract

White Matter Lesions (WMLs) are an accepted marker of small vessel disease. There is evidence of small vessel pathology in Parkinson's disease (PD), yet large-data analysis is required to understand this contribution. The ENIGMA-PD works toward understanding the neuroimaging markers of PD through standardized analysis protocols with the aim of developing a truly reliable quantitative assessment of WMH on MRI. In this study, the focus was to validate and optimize a containerized pipeline that incorporates FSL functions and UNet-pgs implementation, to process T1-weighted and FLAIR MR images to be applied to ENIGMA-PD consortia data. The pipeline was tested across a dataset of 97 subjects, encountering and addressing several challenges, particularly related to memory requirements and segmentation accuracy. To enhance the pipeline's performance, nonlinear transformations and ventricular distance mapping were integrated, allowing for more precise spatial alignment and classification of WMLs into periventricular and deep white matter lesions. The results demonstrated significant improvements in segmentation accuracy when utilizing nonlinear transformations compared to the initial linear approach. The optimized pipeline effectively distinguished between varying levels of WMLs. Additionally, this work included detailed instructions to ensure the pipeline's usability by other research groups, addressing the previously identified limitations in user-friendliness. These findings underscore the importance of advanced preprocessing and transformation techniques in neuroimaging analysis and highlight the potential of this pipeline to be adopted as a standard tool for large-scale studies within the ENIGMA consortium and beyond.