Integrated Causal Learning Workflow of Brain Iron Deposition in Patients with PD based on QSM.

INTRODUCTION: To investigate the model-inferred dependency pattern of iron deposition in the brain of patients with Parkinson's Disease (PD). MATERIALS AND METHODS: Clinical and imaging data of 25 patients with PD and 36 Healthy Controls (HCs) were collected between January and December 2021. Iron deposition in the brain was quantified using Quantitative Susceptibility Mapping (QSM). An integrated causal learning workflow, combining GESbased causal discovery, causal effect estimation, and refutation analysis, was used to explore the causal relationships between clinical data and image characteristics and estimate their causal effects. RESULTS: Beck Anxiety Inventory (BAI) and Beck Depression Inventory (BDI) scores were significantly greater in the PD group than the HC group (P = 0.003; P < 0.001). Timed Up and Go Test (TUG) and Berg Balance Scale (BBS) scores were lower in the PD group than in the HC group (P < 0.001; P < 0.001). The integrated causal learning workflow suggested a cross-sectional, model-inferred directional dependency structure with TH occupying an upstream position and downstream dependencies involving PUT, CN, RN, SN, and GP. DISCUSSION: The iron deposition tissues in different regions of the brains of patients with PD exhibit a structured dependency pattern, which is of great significance for proposing hypotheses regarding the mechanism of iron deposition related to PD. CONCLUSION: The integrated causal learning workflow combined with QSM may help explore cross-sectional dependency patterns of brain iron deposition in patients with PD. These findings provide a biologically interpretable and hypothesis-generating workflow for understanding PD-related iron deposition patterns and may assist in early clinical detection of PD, rapid determination of its course, and timely administration of interventions and treatments.