Deep learning-based magnetic resonance imaging image reconstruction in the assessment of brain microstructural changes in Parkinson's disease patients

This study was aimed to investigate the application value of deep learning-based magnetic resonance imaging (MRI) image reconstruction technology in the assessment of brain microstructural changes in patients with Parkinson's disease (PD). A total of 78 early-stage PD patients and 78 healthy controls were enrolled. Diffusion tensor imaging (DTI) was performed using MRI, and images were reconstructed using a multi-object interactive neural network-based brain MRI segmentation model. Parameters including fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RD) were calculated to assess microstructural changes in brain regions such as the substantia nigra, basal ganglia, and hippocampus. Pearson correlation analysis was employed to examine the association between regional parameters and Montreal cognitive assessment (MoCA) scores, while receiver operating characteristic (ROC) curves were used to evaluate the diagnostic efficacy of these parameters for PD. The constructed segmentation model achieved a Dice similarity coefficient (DSC) of 0.922, with relative volume difference (RVD) and root mean square (RMS) values of 0.042 and 0.46, respectively, outperforming related algorithms. The PD group exhibited significantly reduced FA and increased RD in the substantia nigra, hippocampus, and thalamus. Hippocampal RD demonstrated a strong negative correlation with MoCA scores ( r= -0.67, P< 0.001). ROC analysis indicated that hippocampal RD had the best diagnostic efficacy for PD [area under the curve (AUC) = 0.90, sensitivity 88 %/specificity 87 %], followed by substantia nigra RD (AUC = 0.88) and thalamic RD (AUC = 0.87). Deep learning-based MRI reconstruction technology can accurately quantify early brain microstructural damage in PD patients. The RD of the hippocampus and substantia nigra are sensitive biomarkers for diagnosing PD and screen cognitive impairment, providing a new imaging strategy for the early precise diagnosis and treatment of PD.