Post-gadolinium T1 alterations in Parkinson's disease and essential tremor Authors: Junsung Kim, Eunseon Jeong, Yangsean Choi, Yoonho Nam, Ho Sung Kim, Sungyang Jo, Sun Ju Chung Published: 2026 ABSTRACT Post-contrast T1 changes (ΔT1) on brain MRI may reflect alterations in cerebrospinal fluid dynamics and microvascular exchange in movement disorders. We investigated regional post-contrast ΔT1 differences in patients with idiopathic Parkinson’s disease (IPD) and essential tremor (ET). In this retrospective single-center study, 277 patients who underwent brain MRI for suspected movement disorders between November 2023 and February 2025 were classified as ET (n = 109), early IPD (n = 81), or advanced IPD (n = 87). Post-contrast ΔT1 values were calculated by subtracting pre-contrast from post-contrast T1 measurements in the choroid plexus (CP), basal ganglia perivascular spaces, cerebral cortex, cerebellar cortex, and thalamus. Advanced IPD demonstrated significantly greater T1 shortening in the CP compared with ET, while showing reduced T1 shortening in the cerebellar cortex and thalamus. No significant differences were observed in other regions. These findings suggest region-specific alterations in contrast kinetics in advanced IPD, potentially reflecting disease-related changes in contrast distribution and microvascular permeability. Quantitative post-contrast T1 mapping may provide complementary imaging markers for distinguishing Parkinson’s disease from essential tremor. Introduction The glymphatic system is an emerging theory that describes a brain-wide clearance mechanism protecting against the accumulation of neurotoxic substances. Its key components include the perivascular space (PVS) and the choroid plexus (CP), which reflect the blood-brain barrier (BBB) and the blood-cerebrospinal fluid (CSF) barrier, respectively1. In clinical neuroradiology, gadolinium-based contrast agents (GBCAs) are routinely administered intravenously to detect brain abnormalities. These agents may cross a disrupted BBB and enter the CSF via diffusion or convection, resulting in contrast-related signal changes. Given growing evidence supporting the role of glymphatic-related processes in neurodegenerative diseases, exploring barrier permeability and fluid exchange kinetics in patients with idiopathic Parkinson’s disease (IPD) may yield valuable insights into its pathophysiology. IPD, the second most common neurodegenerative disease after Alzheimer’s disease, is characterized by resting tremor, rigidity, bradykinesia, and gait disturbance2. Disease severity is commonly assessed using the Hoehn and Yahr (H&Y) staging system, which ranges from stage 1 to 5. IPD lasting over 5 years is generally considered advanced3. Essential tremor (ET), another common movement disorder, is primarily characterized by involuntary, rhythmic shaking, often affecting the hands. Unlike IPD, ET presents with action tremor, and has a less clearly defined pathological basis4. The clinical overlap between IPD and ET often complicates differential diagnosis5. The pathology of IPD involves dopaminergic neuronal loss in the substantia nigra and the accumulation of Lewy bodies. In contrast, ET may involve abnormal cerebellar circuitry, including Purkinje cells and basket cells, as well as climbing fibers6. Additionally, thalamic volume loss, particularly in the intralaminar nuclei, has been reported among patients with IPD7,8. Alterations in glymphatic-related imaging markers in IPD have been demonstrated using Diffusion Tensor Imaging Analysis along the Perivascular Space (DTI-ALPS), which showed reduced DTI-ALPS indices in patients with IPD compared to those with ET or healthy controls9–11. However, existing MRI-based methods, including the DTI-ALPS index12,13 and intrathecal GBCA administration14,15 present notable limitations. The DTI-ALPS index can be affected by imaging plane orientation and user-dependent region-of-interest selection, while intrathecal GBCA injection is invasive and impractical in routine clinical settings. To address these limitations, several recent studies have explored the possibility of conducting putative glymphatic assessment using intravenous GBCA administration16–19. Prior studies have demonstrated meaningful relationships between PVS burden, CP volume, and disease severity in IPD10,20; however, these analyses were confined to volumetric measurements. The PVS plays a key role in interstitial fluid drainage and glymphatic circulation—processes increasingly implicated in IPD pathophysiology10. The CP, as the principal interface of the blood–CSF barrier, regulates CSF production and solute exchange21; thus, evaluating both the PVS and CP would offer complementary insight into disease-related alterations in fluid transport and barrier function reflected by post-contrast T1 mapping. To our knowledge, no prior study has yet examined the post-contrast T1 alterations of the CP, PVS, and cortex following intravenous GBCA administration in patients with IPD. Quantitative post-contrast T1 mapping provides a sensitive, region-specific measure of GBCA distribution that reflects vascular and interstitial microenvironmental changes. Despite increasing interest in fluid and barrier dysfunction in Parkinsonian disorders, the spatial characteristics of post-contrast T1 alterations across disease stages remain largely unexplored. Accordingly, the present study aimed to investigate the regional differences in post-contrast T1 changes between IPD and ET, and to explore potential disease-related alterations in microvascular and cerebrospinal fluid dynamics. Results Patients The mean ages of patients with ET, early IPD, and advanced IPD were 67 ± 10, 69 ± 9, and 70 ± 10 years, respectively (p = 0.09); the respective proportions of female patients were 64.2%, 51.9%, and 49.4% (p = 0.08). The mean disease duration was 6.3 ± 7.6 years for ET, 2.0 ± 1.4 years for early IPD, and 7.7 ± 5.5 years for advanced IPD. The median H&Y stage was stage 2 for early IPD and stage 3 for advanced IPD (Table 1).Table 1Baseline characteristics of the included patientsVariablesET (n = 109)Early IPD (n = 81)Advanced IPD (n = 87)p-valueAge (years), mean ± SD66.6 ± 9.968.5 ± 8.769.6 ± 9.70.09Sex n (%)0.08Male39 (35.8%)39 (48.1%)44 (50.6%)Female70 (64.2%)42 (51.9%)43 (49.4%)Duration (years), mean ± SD6.3 ± 7.62.0 ± 1.47.7 ± 5.5<0.001H&Y stage, median [IQR]NA2 [2, 2]3 [3, 3.5]ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, SD Standard Deviation, IQR Interquartile Range, H&Y stage Hoehn and Yahr stage, NA Not Assessable. Table 1Baseline characteristics of the included patientsVariablesET (n = 109)Early IPD (n = 81)Advanced IPD (n = 87)p-valueAge (years), mean ± SD66.6 ± 9.968.5 ± 8.769.6 ± 9.70.09Sex n (%)0.08Male39 (35.8%)39 (48.1%)44 (50.6%)Female70 (64.2%)42 (51.9%)43 (49.4%)Duration (years), mean ± SD6.3 ± 7.62.0 ± 1.47.7 ± 5.5<0.001H&Y stage, median [IQR]NA2 [2, 2]3 [3, 3.5]ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, SD Standard Deviation, IQR Interquartile Range, H&Y stage Hoehn and Yahr stage, NA Not Assessable. Baseline characteristics of the included patients ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, SD Standard Deviation, IQR Interquartile Range, H&Y stage Hoehn and Yahr stage, NA Not Assessable. Group-wise comparisons of post-contrast ΔT1 Choroid Plexus and Basal Ganglia PVS Table 2 summarizes the group-wise comparisons of post-contrast ΔT1 values. The mean post-contrast ΔT1 in the CP did not differ significantly between ET (–859.6 ± 116.9 ms) and early IPD (–871.7 ± 88.9 ms; p = 0.72), nor between early and advanced IPD (–902.2 ± 106.7 ms; p = 0.15). However, a significant post-contrast ΔT1 shortening was observed between advanced IPD and ET (p = 0.02; Fig. 1A). In basal ganglia PVS, no significant differences in mean post-contrast ΔT1 were observed among the three groups (ET vs. early IPD, p = 0.99; ET vs. advanced IPD, p = 0.93; early IPD vs. advanced IPD, p = 0.97; Fig. 1B).Fig. 1Post-contrast ΔT1 of choroid plexus and basal ganglia perivascular spaces across disease groups.A Choroid plexus and B basal ganglia PVS among patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. The mean post-contrast ΔT1 is indicated by asterisks.Table 2MRI characteristics of the included patientsPost-contrast ΔT1 (ms)*ET (n = 109)Early IPD (n = 81)Advanced IPD (n = 87)p-valueChoroid plexus–859.6 ± 116.9–871.7 ± 88.9–902.2 ± 106.70.02†Basal ganglia PVS–93.0 ± 23.4–93.3 ± 23.6–94.2 ± 21.40.94Thalamus–91.5 ± 13.8–89.1 ± 11.4–86.5 ± 12.50.02†Cerebral cortex–170.5 ± 25.0–165.7 ± 20.8–168.3 ± 26.90.7Cerebellar cortex–175.7 ± 24.7–169.7 ± 19.5–167.0 ± 24.60.03†All numbers are shown as mean ± standard deviation.ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, PVS perivascular space.*Post-contrast T1 value change = [Post-contrast T1 value] – [Pre-contrast T1 value].†Significant difference between advanced IPD and ET. Post-contrast ΔT1 of choroid plexus and basal ganglia perivascular spaces across disease groups. A Choroid plexus and B basal ganglia PVS among patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. The mean post-contrast ΔT1 is indicated by asterisks. Table 2MRI characteristics of the included patientsPost-contrast ΔT1 (ms)*ET (n = 109)Early IPD (n = 81)Advanced IPD (n = 87)p-valueChoroid plexus–859.6 ± 116.9–871.7 ± 88.9–902.2 ± 106.70.02†Basal ganglia PVS–93.0 ± 23.4–93.3 ± 23.6–94.2 ± 21.40.94Thalamus–91.5 ± 13.8–89.1 ± 11.4–86.5 ± 12.50.02†Cerebral cortex–170.5 ± 25.0–165.7 ± 20.8–168.3 ± 26.90.7Cerebellar cortex–175.7 ± 24.7–169.7 ± 19.5–167.0 ± 24.60.03†All numbers are shown as mean ± standard deviation.ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, PVS perivascular space.*Post-contrast T1 value change = [Post-contrast T1 value] – [Pre-contrast T1 value].†Significant difference between advanced IPD and ET. MRI characteristics of the included patients All numbers are shown as mean ± standard deviation. ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease, PVS perivascular space. *Post-contrast T1 value change = [Post-contrast T1 value] – [Pre-contrast T1 value]. †Significant difference between advanced IPD and ET. Cerebral and cerebellar cortex No significant differences were found in mean post-contrast ΔT1 values of the cerebral cortex among the three groups (Fig. 2A). In contrast, the mean post-contrast ΔT1 of the cerebellar cortex was least shortened in the advanced IPD group (–167.0 ± 24.6 ms), followed by early IPD (–169.7 ± 19.5 ms) and ET (–175.7 ± 24.7 ms). Pairwise comparison revealed a significantly less T1 shortening in the advanced IPD group compared to ET (p =.03; Fig. 2B).Fig. 2Post-contrast ΔT1 of cerebral cortex, cerebellar cortex, and thalamus across disease groups.A Cerebral cortex, B cerebellar cortex, and C thalamus among patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. The mean post-contrast ΔT1 is indicated by asterisks. Post-contrast ΔT1 of cerebral cortex, cerebellar cortex, and thalamus across disease groups. A Cerebral cortex, B cerebellar cortex, and C thalamus among patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. The mean post-contrast ΔT1 is indicated by asterisks. Thalamus Patients with advanced IPD showed the least T1 shortening in the thalamus (–86.5 ± 12.5 ms), followed by early IPD (–89.1 ± 11.4 ms) and ET (–91.5 ± 13.8 ms). No significant differences were observed between the ET and early IPD groups (p = 0.40) or between early and advanced IPD (p = 0.38). However, advanced IPD showed significantly less T1 shortening in the thalamic mean post-contrast ΔT1 compared to ET (p = 0.02; Fig. 2C). Ventricular post-contrast ΔT1 Among the three groups, the mean post-contrast ΔT1 of the CP, inferior lateral ventricle, lateral ventricle, third ventricle, and fourth ventricle were –877.0 ± 107.0 ms, –239.0 ± 67.0 ms, –102.0 ± 64.9 ms, –55.5 ± 68.2 ms, and –249.0 ± 70.9 ms, respectively. Statistically significant differences were observed among these regions (p < 0.001; Supplementary Fig. S1), except between the inferior lateral ventricle and the fourth ventricle (p = 0.11). An interactive 3D visualization of the post-contrast ΔT1 values across all participants, with the choroid plexus represented as point coordinates, is available at https://olnsnlyy.github.io/Post-GD-T1-alterations-in-PD-and-ET/subt1map_ventricle_gradient_3d.html. Voxel-wise comparison of post-contrast ΔT1 Exploratory voxel-wise analysis identified significant regional differences in post-contrast ΔT1 among groups (Fig. 3). After applying a cluster size threshold (\ge≥50 mm3) and excluding CSF/white matter regions, the most prominent and consistent differences across all three pairwise comparisons were observed in the cerebellar cortex and thalamus.Fig. 3Voxel-wise comparison of post-contrast ΔT1 among essential tremor and idiopathic Parkinson's disease groups.Montreal Neurological Institute (MNI)-registered brain MRI showing regions with statistically significant differences in post-contrast ΔT1. A Essential tremor (ET) versus early idiopathic Parkinson's disease (IPD). B ET versus advanced IPD. C Early IPD versus advanced IPD. The color bar represents p-values from 0.00 (red) to 0.05 (yellow). Voxel-wise comparison of post-contrast ΔT1 among essential tremor and idiopathic Parkinson's disease groups. Montreal Neurological Institute (MNI)-registered brain MRI showing regions with statistically significant differences in post-contrast ΔT1. A Essential tremor (ET) versus early idiopathic Parkinson's disease (IPD). B ET versus advanced IPD. C Early IPD versus advanced IPD. The color bar represents p-values from 0.00 (red) to 0.05 (yellow). In the ET vs. early IPD comparison (Fig. 3A), clusters were primarily localized to the cerebellar and thalamic regions. The pattern became more extensive and intense, with the largest and most intense clusters spanning the cerebellar cortex and thalamus in the ET vs. advanced IPD comparison (Fig. 3B), followed by the intermediate spatial patterns seen in the early IPD vs. advanced IPD comparison (Fig. 3C). Less shortening of post-contrast ΔT1 values was observed in these regions from ET to early and advanced IPD, suggesting a pattern of reduced contrast enhancement in advanced IPD. Representative color-mapped images are presented in Fig. 4.Fig. 4Representative images of post-contrast ΔT1 in cerebellar cortex and thalamus.A Cerebellar cortex and B thalamus in patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. Arrows indicate visible changes of post-contrast ΔT1. Representative images of post-contrast ΔT1 in cerebellar cortex and thalamus. A Cerebellar cortex and B thalamus in patients with essential tremor (ET), early idiopathic Parkinson’s disease (IPD), and advanced IPD. Arrows indicate visible changes of post-contrast ΔT1. PVS scales and correlation analyses Interobserver agreement for PVS counts was excellent, with an intraclass correlation coefficient of 0.957 (95% confidence interval [CI]: 0.945–0.966; p <.01). The mean PVS scale ratings (averaged between the two raters) of early (10.5 ± 6.4) and advanced IPD (10.2 ± 6.8) were significantly higher compared to that of ET (8.0 ± 6.0) (ET vs. early IPD, p = 0.03; ET vs. advanced IPD, p = 0.04). However, no significant difference was observed between early and advanced IPD (p = 0.67; Supplementary Fig. S2). Supplementary Table S1 summarizes the correlation analyses among post-contrast ΔT1, age, and PVS scale. Post-contrast ΔT1 exhibited significantly negative correlations with age in the CP (r = –0.28), basal ganglia PVS (r = –0.29), cerebral cortex (r = –0.41), and thalamus (r = –0.27) (all p < 0.01). PVS scale showed a significantly positive correlation with age (r = 0.47, p < 0.01). Additionally, PVS scale was negatively correlated with post-contrast ΔT1 in the CP (r = –0.25), basal ganglia PVS (r = –0.19), and cerebral cortex (r = –0.26) (all p < 0.01). Discussion Overall, this study demonstrated that the post-contrast ΔT1 of the CP was significantly shortened in patients with advanced IPD compared to those with ET, indicating greater GBCA enhancement. In contrast, the post-contrast ΔT1 of the cerebellar cortex and thalamus was less shortened in advanced IPD compared with ET, indicating lower contrast-related signal changes. These findings suggest that the post-contrast ΔT1 values of IPD patients exhibit region-specific alterations, even on a single-phase post-contrast MRI scan. Age may partially influence the degree of CP enhancement. Our study identified a significant negative correlation between age and post-contrast ΔT1 of the CP. Previous studies have similarly reported age-related increases in the CP volume21,22, presumably due to inflammation and weakening of the blood-CSF barrier, which may partly account for the association between age and CP enhancement in our study. However, no significant age differences were observed among the study groups, thereby minimizing the potential effect of age. No significant differences in post-contrast ΔT1 were observed in the cerebral cortex across groups. However, significant differences were found in the cerebellar cortex and thalamus. These regional variations may be caused by the timing of post-contrast image acquisition. Based on known CSF flow dynamics, CSF is generally described as moving downward from the lateral ventricles through the third and fourth ventricles, and finally into the subarachnoid space23. However, early GBCA distribution may not strictly follow this flow pattern. Several studies have previously demonstrated that the GBCA concentrations within a few minutes post-intravenous administration in the CSF are higher around the third and fourth ventricles than in the lateral ventricles or cerebral cortex18,19,24. At 5 minutes post-injection, intravascular GBCA and a small extracellular fraction of GBCA may contribute to the observed signal changes, potentially reflecting regional differences in cerebral blood volume rather than conventional CSF flow dynamics. Consequently, the significant post-contrast ΔT1 differences between ET and advanced IPD at 5 minutes post-injection may be partly attributable to disease-related alterations in intravascular GBCA dynamics and extracellular contrast distribution. Although this early regional T1 shortening may partly reflect differences in CSF flow pathways or delayed contrast arrival, recent dynamic MRI studies have reported that intravenous GBCA can be detected in the dura mater and subarachnoid space within 40 seconds in healthy volunteers, suggesting that GBCA may enter the CSF through pathways other than the CP25. Our study demonstrated significant regional differences in ventricular post-contrast ΔT1. The most pronounced enhancement in the inferior lateral ventricle may be attributable to its close proximity to the CP and the confined geometry of the ventricular space26. Unfortunately, a dedicated CP mask for the fourth ventricle was not available in the current study. Therefore, partial volume effects from the CP could not be excluded in this region and may have contributed to the relatively high contrast-related signal change. Nevertheless, the observed differences in contrast-related signal change between the inferior lateral ventricle and the third and lateral ventricles suggest regional variation in post-contrast ΔT1, which may reflect differences in proximity to the CP and the geometric characteristics of the region of interest (ROI), rather than bulk CSF flow. Notably, this pattern was not identical to previously reported CSF gadolinium concentration gradients18,24. Interestingly, in contrast to the findings in the CP, the cerebellar cortex and thalamus exhibited opposite changes in the post-contrast ΔT1. However, our study has inherent limitations in accurately interpreting these findings, as it was based on a cross-sectional comparison using a single-phase contrast-enhancement protocol. A further dynamic contrast-enhanced study is required to accurately interpret the true BBB and blood-CSF barrier permeabilities among the CP, cerebellar cortex, and thalamus27,28. Contrary to our expectations, the post-contrast ΔT1 of the basal ganglia PVS did not differ significantly across groups. This may be attributable to the short MRI acquisition time after GBCA injection. According to Sun et al., dynamic-susceptibility-contrast MRI signal changes were observed in the CSF around the CP within 50–250 seconds, whereas signal changes in the lateral ventricular CSF appeared approximately 4 hours post-injection24. To address this limitation, we additionally compared the PVS scale between the ET and IPD groups. The PVS scale ratings were significantly higher in both the early and advanced IPD groups compared to the ET group (Supplementary Results). This finding aligns with previous studies suggesting that enlarged PVS may reflect microvascular structural impairment and serve as a potential biomarker for neurodegenerative diseases, including IPD and Alzheimer’s disease29–31. Post-contrast ΔT1 revealed a significant negative correlation with age in the basal ganglia PVS, cerebral cortex, and cerebellar cortex (Supplementary Results), consistent with previous reports indicating an age-related decline in glymphatic function32–34. A previous longitudinal study investigating T1 relaxation time changes in IPD similarly found decreased T1 relaxation times in the temporoparietal and occipital cortices35. Considering that the glymphatic system and BBB are closely interrelated systems, our findings may suggest a potential role for GBCA in assessing age-related changes in BBB permeability36. This study has several limitations beyond those inherent to its single-phase post-contrast design. First, it was a retrospective study. A single-phase post-contrast protocol was inevitable due to the retrospectively acquired dataset in a clinical setting in which MRI acquisition time is constrained. Second, healthy controls without neurological symptoms were not included, as such individuals are unlikely to undergo the dedicated brain MRI protocol in our institution. Instead, patients diagnosed with or suspected of having ET were used as the comparison group. Third, the post-contrast ΔT1 difference identified between the advanced IPD and ET groups may hold limited clinical relevance compared with a difference observed at earlier disease stages. Future prospective studies with external validations are needed to acquire MRI data at multiple time points after GBCA administration. Finally, there was a substantial overlap of post-contrast ΔT1 among the disease groups, which may limit the discriminative ability for individual patients. The modest effect sizes and substantial overlap in post-contrast ΔT1 between groups indicate that, although group differences are present, the immediate diagnostic utility for individual patients remains limited and should be interpreted cautiously in clinical translation. In conclusion, this study showed that patients with advanced IPD demonstrated significant post-contrast ΔT1 shortening in the CP and lower contrast concentration in the cerebellar cortex and thalamus. These regional differences in contrast distribution may reflect disease-related alterations in microvascular exchange and contrast dynamics. Methods This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of Asan Medical Center, Seoul, Republic of Korea, and the requirement for informed consent was waived due to the retrospective nature of the study (2024-0465). Patients Between November 2023 and February 2025, all patients who underwent brain MRI for suspected movement disorders were retrospectively reviewed. During this period, 1778 brain MRI scans were performed in 1778 individual patients. Inclusion criteria were: (1) age >40 years; and (2) availability of pre-contrast and post-contrast T1 maps. Exclusion criteria included: (1) absence of T1 mapping data (n = 577); (2) indeterminate clinical diagnosis (n = 415; difficult to distinguish between ET and IPD or between IPD and atypical PD); (3) cerebrovascular disease diagnosis (n = 204); (4) atypical Parkinson’s disease diagnosis (n = 125); (5) normal pressure hydrocephalus diagnosis (n = 36); (6) unspecified diagnosis (n = 113); and (7) prior deep brain stimulation (n = 31). Clinical diagnoses Two experienced neurologists diagnosed patients with IPD according to the criteria established by the International Parkinson and Movement Disorder Society37, and ET based on the consensus statement on the classification of tremors38. Among patients diagnosed with IPD, those with a disease duration ≤5 years and a H&Y stage ≤2.5 were classified as having early IPD, whereas those with a disease duration >5 years or a H&Y stage >2.5 were classified as having advanced IPD39. Consequently, a total of 277 brain MRI scans were included in the study, comprising 109 patients with ET, 81 with early IPD, and 87 with advanced IPD. Figure 5 presents the flowchart of the selection process.Fig. 5Flowchart of the selection process for eligible patients.CVD Cerebrovascular Disease, Atypical PD Atypical Parkinson’s Disease, NPH Normal Pressure Hydrocephalus, DBS Deep Brain Stimulation, ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease. Flowchart of the selection process for eligible patients. CVD Cerebrovascular Disease, Atypical PD Atypical Parkinson’s Disease, NPH Normal Pressure Hydrocephalus, DBS Deep Brain Stimulation, ET Essential Tremor, Early IPD Early Idiopathic Parkinson’s Disease, Advanced IPD Advanced Idiopathic Parkinson’s Disease. MRI acquisition All MRI scans were conducted using a 3T scanner with a 64-channel head coil (Magnetom Vida, Siemens Healthineers, Erlangen, Germany). The MRI protocol included pre-contrast Compressed Sensing-Magnetization Prepared 2 Rapid Acquisition Gradient Echoes (CS-MP2RAGE), with multiplanar reconstructions of T1-weighted images and quantitative T1 mapping. Post-contrast CS-MP2RAGE sequences were acquired approximately 5 minutes post-intravenous administration of GBCA (0.1 mmol/kg of gadoterate meglumine, Dotarem, Guerbet). Supplementary Table S2 summarizes the detailed acquisition parameters. Pharmacokinetics of GBCA Following intravenous administration, the majority of GBCAs remain in the intravascular space. However, a small fraction may cross the BBB or the blood-CSF barrier and enter the brain parenchyma and CSF compartments. GBCAs reach the cerebral arterial circulation through the carotid and vertebral arteries, which branch into the penetrating arteries or arterioles. Concurrently, GBCAs are distributed through the CP capillaries. Within the brain interstitium, GBCAs may be exchanged between the interstitial fluid and CSF through mechanisms such as arterial pulsation and CSF pressure gradients, which may be influenced by respiratory cycling. Ultimately, GBCAs may exit the brain through the perivenular PVS into the subarachnoid space32,40,41. The PVSs are CSF-filled spaces surrounding the penetrating arteries and arterioles. Previous studies have predominantly focused on the basal ganglia, in which enlarged PVSs are most frequently observed. However, recent investigations have expanded to include the PVSs of the cerebral cortex and juxtacortical white matter42–44. Image preprocessing Figure 6A shows schematic illustrations of the PVS of the basal ganglia and cerebral cortex. Pre-contrast T1-weighted images were registered to post-contrast T1-weighted images using a rigid-body transformation approach with 6 degrees of freedom, implemented using FLIRT (FMRIB’s Linear Image Registration Tool) with a least-squares cost function45–47. The resulting transformation matrix was applied to both the pre-contrast T1-weighted images and T1 maps to ensure spatial alignment with the post-contrast images. Following registration, brain segmentation was performed on the registered pre-contrast T1-weighted images using FreeSurfer’s SynthSeg48, a deep learning-based segmentation tool for parcellating brain structures used in subsequent quantitative analysis of contrast enhancement. Figure 6B summarizes the image preprocessing steps. Supplementary Fig. S3 illustrates the overall automated segmentation workflow, while Supplementary Table S3 provides the detailed deep learning model hyperparameters.Fig. 6Glymphatic system overview and image preprocessing pipeline.A Schematic representation of the glymphatic system in perivascular spaces of the basal ganglia and cerebral cortex; B Image preprocessing steps for generating post-contrast ΔT1. Glymphatic system overview and image preprocessing pipeline. A Schematic representation of the glymphatic system in perivascular spaces of the basal ganglia and cerebral cortex; B Image preprocessing steps for generating post-contrast ΔT1. Deep learning-based segmentation method We developed a deep learning-based model for the automated segmentation of choroid plexus (CP) and basal ganglia perivascular spaces (BGPVS). For model training, ground truth labels were generated for 101 subjects. The trained model was then applied to perform CP and BGPVS segmentation across the entire dataset. The details of each step are described below. Choroid Plexus For the CP, initial lateral ventricle and CP masks were estimated from FreeSurfer49. These preliminary masks were refined using a Gaussian Mixture Model (GMM) approach50 to improve the delineation of structure boundaries. Subsequently, an expert radiologist (blinded) reviewed and manually corrected these labels to ensure anatomical accuracy. Basal ganglia perivascular space The BGPVS were segmented via a multi-step algorithmic process. The basal ganglia regions (i.e., putamen and caudate nucleus) were identified using SynthSeg-derived segmentation labels48. Restricting the analysis to slices containing these structures, the T1-weighted images were standardized using histogram equalization. The Frangi filter51 was then applied to detect tubular structures characteristic of perivascular spaces, using scale parameters optimized for BGPVS (scale-range: 1–2). The basal ganglia masks were further refined through morphological operations (binary closing followed by dilation with a spherical structuring element) to ensure comprehensive coverage. A threshold was established based on the mean vesselness response within these masks plus one standard deviation. Voxels exceeding this threshold were classified as BGPVS, with separate labels maintained for left and right hemispheres. The final BGPVS masks were reviewed and corrected by an expert radiologist (blinded) to ensure anatomical accuracy and proper delineation of these structures. Deep learning model training The manually corrected CP and BGPVS masks were used to train an nnU-Net model52 for automated segmentation. The model was trained using the registered pre-contrast T1-weighted images and post-contrast T1-weighted images as dual-channel input. The training dataset consisted of 101 paired images, and model parameter tuning was performed using 5-fold cross-validation. Segmentation performance was evaluated using the Dice similarity coefficient and Intersection over Union (IoU). The model achieved a Dice coefficient of 0.806 ± 0.013 (IoU: 0.682 ± 0.017) for BGPVS and 0.754 ± 0.007 (IoU: 0.608 ± 0.008) for CP across the 5 folds. These metrics demonstrate segmentation performance comparable to previous studies for these small and challenging anatomical structures; prior studies on CP segmentation typically report Dice scores of approximately 0.58–0.82 for deep learning-based methods and 0.24-0.37 for FreeSurfer53. The trained model was then applied to the entire dataset for inference. All automated segmentations underwent slice-by-slice visual quality control using 3D Slicer (v.5.2.2, https://www.slicer.org/), and were reviewed by an expert neuroradiologist. For both CP and BGPVS, segmentations were evaluated based on (1) spatial confinement within expected anatomical boundaries, (2) morphological consistency with known structural characteristics, and (3) careful inspection for false positives and false negatives. Manual corrections were performed when necessary to ensure segmentation accuracy. The complete workflow pipeline is illustrated in Supplementary Fig. S1. Generation of post-contrast T1 value changes GBCAs shorten T1 relaxation times; thus, a more intense contrast enhancement corresponds to a greater reduction in T1 values54,55. To generate post-contrast T1 value changes, both the registered pre-contrast and post-contrast T1 maps were subjected to Gaussian smoothing (sigma = 1 voxel, equivalent to 1 mm at 1 mm3 isotropic resolution) to reduce noise and improve signal-to-noise ratio. This smoothing step was essential to minimize the impact of small registration errors and to enhance the reliability of the subsequent quantitative analysis. Following smoothing, post-contrast T1 value changes (hereafter referred to as post-contrast ΔT1) were calculated as the post-contrast T1 value minus the pre-contrast T1 value. Post-contrast ΔT1 values of the CP, basal ganglia PVS, cerebral cortex, cerebellar cortex, and thalamus were calculated for all patients. The thalamic region was included due to its neuronal activity in the motor circuits responsible for motor and nonmotor symptoms of IPD7,8. In addition, post-contrast ΔT1 values were compared across ventricular regions of the entire study cohort to assess regional gradients across the ventricular system. The analyzed regions included the CP, lateral ventricle, inferior lateral ventricle, third ventricle, and fourth ventricle. To quantify within the CSF space and avoid contamination from CP tissues, the CP mask was excluded from the inferior lateral ventricle and the lateral ventricle. Because a dedicated CP mask was not available for the fourth ventricle, measurements in this region were performed using the ventricle mask including the CP. Voxel-wise analysis of post-contrast ΔT1 differences between groups All post-contrast ΔT1 maps were nonlinearly registered to the Montreal Neurological Institute (MNI) template using FNIRT (part of the FSL software package). Voxel-wise independent samples t-tests were conducted to detect areas with group-wise differences (i.e., p < 0.05). False discovery rate (FDR) correction (q < 0.05) was applied to control for multiple comparisons. To further reduce the risk of spurious findings, a spatial extent threshold of 50 mm3 was applied, which was empirically determined by evaluating a range of cluster size thresholds to identify the minimum value that effectively suppressed isolated noise-driven voxels while preserving spatially coherent clusters. Additionally, CSF and white matter regions were excluded using masks derived from the MNI template. The resulting p-value maps were visually inspected to identify any anatomically meaningful regions showing consistent patterns across comparisons. Qualitative PVS scale ratings and correlation analyses The PVS scales were independently reviewed by two neuroradiologists (Reader A and Reader B; both blinded), with 7 and 11 years of experience in neuroradiology, respectively. PVS counts were qualitatively evaluated on axial 3D T1-weighted and axial 3D FLAIR images at two anatomical levels: 2 mm (basal ganglia) and 37 mm (centrum semiovale) above the anterior commissure. If there were 20 or more PVS in a patient, the number was recorded as ≥2056. Pearson’s correlation analyses were conducted to examine relationships among post-contrast ΔT1 values, PVS scale ratings, and age. Statistical analysis Demographic and clinical data for the ET, early IPD, and advanced IPD groups were compared using one-way analysis of variance or the Kruskal–Wallis test, as appropriate. Group comparisons of post-contrast ΔT1 values in the CP, basal ganglia PVS, cerebral cortex, cerebellar cortex, and thalamus were performed using one-way analysis of variance, followed by Tukey’s Honestly Significant Difference test, with adjustments for age and sex. Ventricular post-contrast ΔT1 differences were evaluated using repeated-measures ANOVA with Bonferroni-corrected post-hoc pairwise comparisons. PVS scale ratings for the three groups were compared using the Kruskal–Wallis test with post-hoc Dunn’s test. Interobserver agreement on PVS scale ratings was evaluated using the intraclass correlation coefficient. All statistical analyses were conducted using R (version 4.4.3), with statistical significance set at p <.05. Supplementary information Supplementary Information Supplementary Information