Idiopathic rapid eye movement sleep behavior disorder (iRBD), a sleep abnormality characterized by loss of normal rapid eye movement sleep muscle tone and dreaming behavior, is considered to be a prodromal stage of Parkinson's disease (PD) and other α-synucleinopathies [1]. The possible pathophysiology of iRBD is thought to stem from dysfunction of the pontine subcoeruleus complex [2]. Importantly, neuroimaging studies on iRBD have indicated minor levels of nigrostriatal dysfunction and have demonstrated that neuroinflammation in brain regions is commonly affected in Parkinson's disease, such as the substantia nigra, dorsal striatum, and posterior cortex [3,4]. Therefore, iRBD and PD may share lesion characteristics in some brain regions.

PD is a chronic neurodegenerative disease primarily caused by loss of dopamine-producing nerve cells and the aggregation of Lewy bodies. Dopaminergic neurons directly influence cerebral microvessels, thereby modifying local cerebral perfusion [5,6]. In addition, Lewy body pathology in PD contributes to a compromised blood-brain barrier, impacting capillary function and blood flow control [7,8]. It has been shown that PD patients have significantly reduced cerebral blood flow (CBF) in several brain regions such as the caudate nucleus, supplementary motor area, precuneus, and insula compared with healthy control (HC) [9,10,11,12]. Matthews et al. revealed that hypometabolism in parieto-occipital and premotor cortices could be employed for discriminating PD from HC [13]. Similar to the reduction in dopamine content observed in PD, the decreased dopamine transporter (DAT) [14,15,16] and amino-acid decarboxylase (AADC) [3,4,17] in iRBD also lead to decrease dopamine levels. This alteration in dopamine content affects the microvascular environment, potentially influencing local cerebral perfusion. Recent research on iRBD has demonstrated reduced perfusion in brain regions such as the frontal and tempo-parietal cortices, parieto-occipital lobe, and insula, when compared to HC [18,19,20]. Notably, these identified abnormal brain regions partly align with the metabolic profile observed in Parkinson's disease.

Arterial spin labeling (ASL), as a non-invasive MRI technique, enables quantitative measurement of CBF, with lower inter-individual variability compared to PET and SPECT [21]. Furthermore, ASL has been used to study resting-state perfusion changes in patients with PD, and results have shown that local CBF changes are consistent with PET metabolic imaging [11] and they correlate with the severity and progression of PD [22]. ASL holds promise for providing an objective basis for the adjunctive diagnosis of iRBD and PD. In addition, with the deep development of medical-industrial crossover, the combination of multiple imaging methods with machine learning (ML) has been developing rapidly, which takes into account inter-regional correlations and thus improves the sensitivity to subtle variations and differences in spatial distributions, as compared with the traditional group level analysis. In a previous study, Matthews et al. combined PET scanning with machine learning, which showed good efficacy in identifying PD from HC [13]. Furthermore, Pang et al. found that a combination of resting-state fMRI with machine learning was capable of differentiating PD subtypes [23]. We therefore hypothesize that the combination of ASL with machine learning techniques might provide imaging biomarkers for the diagnosis of PD and iRBD at the individual level. Specifically, we have two primary objectives: (1) To extract CBF changes as potential diagnostic markers for PD or iRBD; and (2) To investigate whether any CBF derived features can predict the transition from iRBD to PD.