Diagnostic Potential of Macromolecular Proton Fraction Mapping Combined with Quantitative Susceptibility Mapping as a Subcortical Biomarker for Parkinson’s Disease

Abstract

Purpose

This study aimed to evaluate the diagnostic potential of macromolecular proton fraction (MPF) mapping combined with magnetic susceptibility measurements as subcortical biomarkers in Parkinson’s disease (PD).

Methods

Twenty patients with PD and 9 age-matched healthy controls underwent 3.0T magnetic resonance imaging, including MPF mapping and quantitative susceptibility mapping (QSM). MPF and magnetic susceptibility values of the 16 subcortical nuclei were measured in both the PD group and healthy control group, and these values were compared between the 2 groups. Diagnostic performance of MPF and magnetic susceptibility in the subcortical nuclei was individually evaluated using receiver operating characteristic (ROC) analysis. A logistic regression model was developed for the diagnosis of PD using a combination of MPF and magnetic susceptibility values.

Results

Quantitative analyses revealed significantly reduced MPF values (false discovery rate–corrected P < 0.05) in 7 subcortical nuclei in the PD group, namely the hypothalamus, parabrachial pigmented nucleus, red nucleus, substantia nigra pars reticulata, substantia nigra pars compacta, ventral pallidum, and ventral tegmental area. Additionally, magnetic susceptibility was significantly elevated (false discovery rate–corrected P < 0.05) in the parabrachial pigmented nucleus and substantia nigra pars compacta. ROC curve analysis demonstrated strong diagnostic performance, with the ventral pallidum showing the highest MPF-based diagnostic accuracy (area under the curve [AUC] = 0.82) and the substantia nigra pars compacta showing the highest QSM-based accuracy (AUC = 0.88). The logistic regression model combining MPF and magnetic susceptibility showed the best performance (AUC = 0.93).

Conclusion

MPF mapping, particularly when combined with magnetic susceptibility measurements, may serve as a quantitative diagnostic biomarker for PD, and the observed widespread alterations across multiple subcortical nuclei provide new insights into the pathology of PD beyond the classic nigrostriatal pathway.

Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by the loss of dopaminergic neurons in the substantia nigra, with 30%–50% of neurons already lost at the onset of motor symptoms.^1–3 This substantial neuronal loss prior to clinical manifestation underscores the critical need for sensitive imaging biomarkers capable of detecting neurodegeneration during the prodromal stage. While the primary pathology of PD centers on dopaminergic neuronal loss and iron accumulation in the substantia nigra, mounting evidence suggests widespread involvement of other subcortical nuclei within the basal ganglia.^4–6 Therefore, comprehensive assessment of changes across subcortical nuclei may provide crucial diagnostic information and insights into disease progression.

Multiple MRI-based studies have investigated subcortical nuclei in PD, with quantitative susceptibility mapping (QSM) emerging as the prominent technique.^7–9 QSM has consistently demonstrated elevated susceptibility values in the substantia nigra pars compacta, red nucleus, globus pallidus, and putamen of patients with PD compared with healthy controls.^7,10–12 However, the relationship between iron deposition and neurodegeneration remains unclear in terms of whether iron accumulation initiates or accelerates neurodegeneration, or whether it represents a secondary metabolic consequence of ongoing neuronal loss.^13,14 This mechanistic uncertainty highlights the need for complementary imaging parameters that can assess other aspects of PD pathology, particularly subcortical myelin and neuronal integrity changes.

Macromolecular proton fraction (MPF) mapping represents a promising quantitative MRI technique that measures the proportion of protons bound to macromolecules, providing direct insights into tissue integrity and macromolecular content.^15–17 This technique exhibits particular sensitivity to changes in myelin content, making it a potentially valuable tool for assessing neurodegenerative disorders.¹⁶ Given these capabilities, MPF mapping may serve as a complementary imaging biomarker for detecting early neurodegenerative changes in PD and monitoring disease progression. While the ultimate goal is to develop MPF mapping for early PD diagnosis, this preliminary study focused on establishing the diagnostic capability of MPF in a well-characterized PD cohort that included patients across advanced disease stages. Additionally, this study aimed to validate whether the diagnostic accuracy of PD can be improved by combining MPF and QSM. This approach first allows us to demonstrate the sensitivity of MPF to PD-related pathological changes before proceeding to early-stage disease detection in future studies.

Materials and Methods

Ethics statements

This study was approved by the institutional review board of National Hospital Organization Omuta National Hospital (Approval no. 4-9), and all procedures adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment.

Study design and participants

This cross-sectional study was conducted at the National Hospital Organization Omuta National Hospital using a 3.0T MRI system (Ingenia; Philips Healthcare, Best, the Netherlands) equipped with a 15-channel phased-array coil. This study included 20 patients with PD (mean age: 74.0 ± 8.7 years; 10 males, 10 females) and 9 age-matched healthy controls (mean age: 76.2 ± 7.4 years; 4 males, 5 females). Detailed participant characteristics are summarized in Table 1. The mean disease duration was 7.9 ± 5.9 years.

Table 1.

Demographic and clinical characteristics of the study participants

	Patients with PD (n = 20)	Healthy controls (n = 9)	P-value
Age (years, mean ± SD)	74.0 ± 8.7	76.2 ± 7.4	0.81
Sex (male/female), n (%)	10 (50%)/10 (50%)	4 (44%)/5 (56%)	> 0.99
MMSE (mean ± SD)	24.7 ± 3.4	29.2 ± 0.7	< 0.001*
Disease duration (years)	7.9 ± 5.9	—
Hoehn and Yahr stage, n (%)	Stage 2: 7 (35%)	—
	Stage 2.5: 1 (5%)
	Stage 3: 3 (15%)
	Stage 3.5: 2 (10%)
	Stage 4: 5 (25%)
	Stage 4.5: 1 (5%)
	Stage 5: 1 (5%)

*Significant difference between patients with PD and healthy controls (P < 0.05).MMSE, Mini-Mental State Examination; PD, Parkinson’s disease; SD, standard deviation.

PD diagnosis was based on the Movement Disorder Society Clinical Diagnostic Criteria. All patients were evaluated in their “on medication state.” Exclusion criteria included other neurological disorders, psychiatric illness requiring medication, head trauma, MRI contraindications, and substantial movement artifacts during scanning. All patients with PD were diagnosed by a board-certified neurologist with over 25 years of experience in movement disorders. Participants underwent cognitive assessment using the Mini-Mental State Examination (MMSE). Patients with PD were further assessed for motor symptoms using the Hoehn and Yahr stage. Patients with comorbid neurodegenerative or psychiatric conditions were excluded. Healthy controls were individuals who had no history of major medical, neurological, or psychiatric disorders. They showed no structural brain abnormalities on conventional MRI and no abnormal neurological findings on examination. All healthy controls also demonstrated normal cognitive function, defined as an MMSE score > 27.

MRI acquisition

The imaging protocol comprised several 3D sequences acquired with a 1.0 mm isotropic resolution and 170 slices. T₁-weighted magnetization-prepared rapid gradient echo (MPRAGE) images were acquired using the MPRAGE sequence with the following parameters: FOV, 240 × 240 mm²; TR, 6.1 ms; TE, 2.7 ms; flip angle (FA), 8°; inversion time, 768 ms; bandwidth, 289.4 Hz/pixel; number of excitations (NEX), 1; matrix size, 240 × 240; SENSE factor for slice-encoding direction, 2.0; and acquisition time, 4 mins and 17s.

The MPF mapping protocol was implemented using the single-point reference method with 3 sequences.^15,18 Proton density–weighted images were acquired using a fast spoiled gradient echo sequence with the following parameters: FOV, 256 × 256 mm²; TR, 18.0 ms; multi-echo acquisition (6 echoes; TE, 2.3–13.8 ms; delta TE, 2.3 ms); FA, 4°; bandwidth, 479.4 Hz/pixel; NEX, 1; SENSE factors for phase-encoding direction and slice-encoding direction, 2.0; and acquisition time, 4 mins and 14s. T₁-weighted images were acquired with similar parameters except for a FA of 20°, bandwidth of 433.5 Hz/pixel, and SENSE factors for phase-encoding direction and slice-encoding direction of 2.0 and 1.25, respectively. Magnetization transfer (MT)—weighted images were acquired with FOV of 256 × 256 mm², TR of 68.0 ms, TE of 2.3 ms, FA of 10°, bandwidth of 479.4 Hz/pixel, NEX of 1, SENSE factors for phase and slice-encoding direction of 2.0 and 1.75, respectively, and acquisition time of 7 mins and 58s. Off-resonance saturation was achieved using the manufacturer’s standard 3-lobe sinc pulse with Gaussian apodization (duration, 16.5 ms; amplitude, 9 μT; effective FA, 520°; and offset-frequency, 1.1 kHz).

B₁ mapping was performed using the double-angle method with a 2D single-shot turbo spin echo sequence with the following parameters¹⁹: FOV, 256 × 240 mm²; TR, 1000 ms; effective TE, 100 ms; FAs, 60°/120°; acquisition matrix, 64 × 60; matrix size, 256 × 256; bandwidth, 726.1 Hz/pixel; slice thickness, 4 mm; slice gap, 0 mm; and acquisition time, 43s.

Image analysis

MPF maps were generated via voxel-wise iterative solutions of the pulsed MT equation using customized C++ software.^15,18 The 2 pool model parameters included a cross-relaxation rate (R) of 19 s⁻¹, T₂ of bound macromolecular protons (T₂^B) of 10 µs, and product of the relaxation rate R₁ and T₂ of free-water protons (R₁T₂^F) of 0.22. B₁ inhomogeneity was corrected using the B₁ map.

QSM analysis was conducted using an in-house program with the following sequential processing pipeline^9,20–22: Laplacian-based unwrapping was applied to multi-echo phase images. Brain binary masks were generated using the brain extraction tool in FMRIB Software Library (FSL), based on magnitude images. Background field removal was performed using the variable-kernel sophisticated harmonic artifact reduction for phase data algorithm, with spherical kernel radii from 2 to 30 mm and 0-mm⁻¹ cut-off frequency. Local field maps at each TE were combined using weighted averaging, with weights derived from T₂* maps estimated from magnitude images.^9,20–23 Susceptibility maps were reconstructed using the improved sparse linear equations and least squares method. Mean cerebrospinal fluid susceptibility in the lateral ventricles served as the zero reference for susceptibility values.²⁴

T₁-weighted MPRAGE images were non-linearly registered to the MNI152 space using the Advanced Normalization Tools (ANTs) standard symmetric normalization algorithm.²⁵ A subcortical nuclei atlas was inversely transformed into each participant’s native space to extract MPF and magnetic susceptibility values from 16 predefined subcortical nuclei.²⁶ Figure 1 shows the subcortical anatomical atlas in the MNI152 space used in this study.

Fig. 1.

Display of an anatomical atlas of 16 subcortical nuclei on the MNI152 standard space template in the axial and coronal views. Different colors indicate the segmentation of various subcortical nuclei.

Statistical analysis

Statistical analyses were performed using Prism 10.0 (GraphPad Software, La Jolla, CA, USA) and MATLAB 2024b (The MathWorks, Natick, MA, USA).

The Mann–Whitney U test was performed to compare age and MMSE scores between the groups, and the Fisher exact test was used to compare sex distribution between the groups. Additionally, the Shapiro–Wilk test was used to evaluate normality of MPF and magnetic susceptibility distributions. Given non-normal distributions, the Mann–Whitney U test was employed to compare values between groups across 16 subcortical nuclei, with Benjamini–Hochberg false discovery rate (FDR) correction (5% threshold) applied for multiple comparisons. In all analyses, the differences were considered statistically significant at P < 0.05. Receiver operating characteristic (ROC) curve analysis was used to assess diagnostic accuracy, with areas under the curve (AUCs) > 0.7 considered clinically meaningful. The optimal cut-off values were determined using the Youden index (sensitivity + specificity – 1).

Finally, a logistic regression model was developed for the diagnosis of PD using a combination of MPF and magnetic susceptibility values. Initially, 9 MPF and QSM measurements from subcortical nuclei that showed significant differences were extracted from each participant. Owing to the presence of multicollinearity among these parameters, we first assessed their pairwise correlations and calculated the variance inflation factors. To reduce the number of features and address multicollinearity, we applied the Lasso method (L1-regularized logistic regression) with 10-fold cross-validation to select the optimal regularization parameter (λ). Subsequently, the Lasso procedure selected 5 parameters (MPFs for the substantia nigra pars compacta, substantia nigra pars reticulata, and ventral tegmental area; and QSMs for the parabrachial pigmented nucleus and substantia nigra pars compacta). Then, these selected parameters were used to build the final logistic regression model. Model performance was evaluated using the AUC, sensitivity, specificity, and the Youden index.

Results

Patient population

Statistical analysis revealed no significant difference in age (P = 0.81) or sex distribution (P > 0.99) between the groups. However, MMSE scores were significantly lower in the PD group than in healthy controls (P < 0.001).

MPF analysis

Figure 2 shows representative MPF maps for a patient with PD and an age-matched healthy control. Quantitative assessment revealed reduced MPF values in the subcortical structures of the patient with PD, particularly in the substantia nigra pars compacta and basal ganglia. There were significantly lower MPF values in patients with PD than in healthy controls across 7 subcortical nuclei (FDR-corrected P < 0.05): the hypothalamus, parabrachial pigmented nucleus, red nucleus, substantia nigra pars reticulata, substantia nigra pars compacta, ventral pallidum, and ventral tegmental area. Detailed MPF values for all analyzed regions are presented in Table 2. Figure 3 shows the box plot of MPF values for patients with PD and healthy controls in each subcortical nuclei.

Fig. 2.

A representative example showing an atlas of subcortical nuclei overlaid on a T₁-weighted image, displaying participants’ MPF values. Only subcortical nuclei demonstrating statistically significant differences are shown. The upper rows correspond to a patient with PD, whereas lower rows correspond to a HC. The labeled regions (a)–(g) indicate the hypothalamus, parabrachial pigmented nucleus, red nucleus, substantia nigra pars reticulata, substantia nigra pars compacta, ventral pallidum, and ventral tegmental area, respectively.F, female; HC, healthy control; MPF, macromolecular proton fraction; PD, Parkinson’s disease.

Table 2.

MPF values in subcortical nuclei of patients with PD and healthy controls

Subcortical nuclei	MPF (%)		P-value	FDR adjusted P-value
	Patients with PD (mean ± SD)	Healthy controls (mean ± SD)
Caudate nucleus	5.58 ± 0.63	5.97 ± 0.47	0.1988	0.177
Extended amygdala	4.38 ± 0.54	4.62 ± 0.34	0.4439	0.320
Globus pallidus external	7.59 ± 0.76	8.30 ± 0.82	0.0386	0.056
Globus pallidus internal	7.78 ± 0.75	8.31 ± 0.88	0.1397	0.134
Habenular nuclei	5.15 ± 0.79	5.46 ± 0.49	0.2948	0.243
Hypothalamus	5.26 ± 0.42	5.66 ± 0.32	0.0097	0.030*
Mammillary nucleus	5.28 ± 0.77	5.82 ± 0.73	0.0491	0.057
Nucleus accumbens	5.25 ± 0.43	5.47 ± 0.38	0.3171	0.244
Parabrachial pigmented nucleus	7.09 ± 0.70	7.87 ± 0.67	0.0113	0.030*
Putamen	6.28 ± 0.69	6.88 ± 0.68	0.0854	0.090
Red nucleus	7.52 ± 0.68	8.37 ± 0.83	0.0097	0.030*
Substantia nigra pars reticulata	7.11 ± 0.73	7.89 ± 0.74	0.0231	0.043*
Subthalamic nucleus	8.78 ± 0.81	9.51 ± 0.81	0.0491	0.057
Substantia nigra pars compacta	6.56 ± 0.69	7.29 ± 0.61	0.0131	0.030*
Ventral pallidum	6.44 ± 0.80	7.24 ± 0.70	0.0051	0.030*
Ventral tegmental area	7.06 ± 0.72	7.95 ± 0.94	0.0264	0.043*

*FDR adjusted P-value < 0.05.FDR, false discovery rate; MPF, macromolecular proton fraction; PD, Parkinson’s disease; SD, standard deviation.

Fig. 3.

Box plots showing macromolecular proton fraction values in the subcortical nuclei. HC; healthy control; MPF, macromolecular proton fraction; PD, Parkinson’s disease.

QSM analysis

Figure 4 shows representative QSM maps for a patient with PD and an age-matched healthy control. Magnetic susceptibility values were significantly elevated in 2 subcortical nuclei of patients with PD (FDR-corrected P < 0.05): the parabrachial pigmented area and substantia nigra pars compacta. Complete magnetic susceptibility measurements and statistical comparisons are presented in Table 3. Figure 5 shows the box plot of magnetic susceptibility values for patients with PD and healthy controls in each subcortical nuclei.

Fig. 4.

A representative example showing an atlas of subcortical nuclei overlaid on a T₁-weighted image, displaying participants’ magnetic susceptibility values. Only subcortical nuclei demonstrating statistically significant differences in MPF and magnetic susceptibility values are shown. The upper rows correspond to a patient with PD, whereas the lower rows correspond to a healthy control. The labeled regions (a)–(g) indicate the hypothalamus, parabrachial pigmented nucleus, red nucleus, substantia nigra pars reticulata, substantia nigra pars compacta, ventral pallidum, and ventral tegmental area, respectively.F, female; MPF, macromolecular proton fraction; PD, Parkinson’s disease.

Table 3.

Magnetic susceptibility values in subcortical nuclei of patients with PD and healthy controls

Subcortical nuclei	Susceptibility (ppm)		P-value	FDR adjusted P-value
	Patients with PD (mean ± SD)	Healthy controls (mean ± SD)
Caudate nucleus	0.071 ± 0.016	0.062 ± 0.016	0.1674	0.224
Extended amygdala	0.068 ± 0.018	0.064 ± 0.012	0.5943	0.624
Globus pallidus external	0.150 ± 0.029	0.148 ± 0.044	> 0.9999	0.919
Globus pallidus internal	0.112 ± 0.028	0.118 ± 0.048	0.6940	0.680
Habenular nuclei	0.045 ± 0.017	0.034 ± 0.014	0.0947	0.224
Hypothalamus	0.041 ± 0.012	0.034 ± 0.010	0.2342	0.265
Mammillary nucleus	0.036 ± 0.015	0.026 ± 0.012	0.2342	0.265
Nucleus accumbens	0.039 ± 0.010	0.033 ± 0.009	0.1531	0.224
Parabrachial pigmented nucleus	0.089 ± 0.029	0.057 ± 0.019	0.0043	0.031*
Putamen	0.078 ± 0.023	0.063 ± 0.020	0.1397	0.224
Red nucleus	0.095 ± 0.024	0.077 ± 0.026	0.0617	0.182
Substantia nigra pars reticulata	0.157 ± 0.034	0.119 ± 0.043	0.0436	0.182
Subthalamic nucleus	0.104 ± 0.038	0.080 ± 0.028	0.0551	0.182
Substantia nigra pars compacta	0.150 ± 0.052	0.096 ± 0.030	0.0014	0.020*
Ventral pallidum	0.155 ± 0.044	0.132 ± 0.054	0.1272	0.224
Ventral tegmental area	0.048 ± 0.022	0.033 ± 0.016	0.1272	0.224

*FDR adjusted P-value < 0.05. FDR, false discovery rate; PD, Parkinson’s disease; SD, standard deviation.

Fig. 5.

Box plots showing magnetic susceptibility values in the subcortical nuclei. HC; healthy control; PD, Parkinson's disease.

Diagnostic performance

ROC analysis demonstrated strong diagnostic performance for the quantitative parameters (Fig. 6). Among MPF-based analyses, the ventral pallidum showed the highest diagnostic accuracy (AUC = 0.82, 95% confidence interval [CI]: 0.66–0.98), followed by the hypothalamus, red nucleus, parabrachial pigmented nucleus, substantia nigra pars compacta, substantia nigra pars reticulata, and ventral tegmental area (Table 4). Additionally, these regions demonstrated clinically meaningful diagnostic performance, despite progressively lower AUC values. In QSM-based analyses, the substantia nigra pars compacta exhibited the highest diagnostic accuracy (AUC = 0.88, 95% CI: 0.73–1.00). This was followed by the parabrachial pigmented nucleus, which also demonstrated notable diagnostic performance (Table 5).

Fig. 6.

ROC curves for MPF and QSM measurements in discriminating patients with PD from healthy controls. Panel (a) shows the ROC curve for the MPF-based analysis, panel (b) shows the ROC curve for the QSM-based analysis, and panel (c) shows the ROC curve for the logistic regression model that combined MPF and magnetic susceptibility values. MPF, macromolecular proton fraction; PD, Parkinson’s disease; QSM, quantitative susceptibility mapping; ROC, receiver operating characteristic.

Table 4.

Significant differences in mean MPF values and diagnostic accuracy (PD vs. HC)

	AUC	95% confidence interval	Cut-off	Sensitivity (%)	Specificity (%)	Youden index
Hypothalamus	0.80	0.64 to 0.96	5.28	60.0	100	60.0
Parabrachial pigmented nucleus	0.79	0.61 to 0.98	7.80	95.0	66.7	61.7
Red nucleus	0.80	0.60 to 1.00	7.97	80.0	77.8	57.6
Substantia nigra pars reticulata	0.77	0.59 to 0.94	6.99	45.0	100	45.0
Substantia nigra pars compacta	0.79	0.61 to 0.97	7.14	85.0	66.7	51.7
Ventral pallidum	0.82	0.66 to 0.98	6.71	80.0	88.9	68.9
Ventral tegmental area	0.72	0.53 to 0.91	7.05	45.0	90.0	35.0

AUC, area under the curve; HC, healthy control; MPF, macromolecular proton fraction; PD, Parkinson’s disease.

Table 5.

Significant differences in mean magnetic susceptibility values and diagnostic accuracy (PD vs. HC)

	AUC	95% confidence interval	Cut-off	Sensitivity (%)	Specificity (%)	Youden index
Parabrachial pigmented nucleus	0.83	0.65 to 1.00	0.058	90.0	66.7	56.7
Substantia nigra pars compacta	0.86	0.71 to 1.00	0.120	75.0	88.9	63.9

AUC, area under the curve; HC, healthy control; PD, Parkinson’s disease.

The Lasso regression analysis with 10-fold cross-validation identified the optimal λ, resulting in the selection of 5 out of the 9 original parameters. Specifically, the selected parameters included MPF values for the substantia nigra pars compacta, substantia nigra pars reticulata, and ventral tegmental area, as well as QSM values for the parabrachial pigmented nucleus and substantia nigra pars compacta. The ROC analysis of the final logistic regression model yielded an AUC of 0.93 (95% CI: 0.750–1.000). At the optimal cut-off point, the model achieved a sensitivity of 100%, specificity of 88.9%, and Youden index of 0.89.

Discussion

This study demonstrates that MPF mapping can quantitatively detect pathological changes in the subcortical nuclei of patients with PD. The observed reduction in MPF across multiple subcortical structures aligns with histopathological evidence of widespread neurodegeneration extending beyond the substantia nigra pars compacta,^27–29 highlighting the importance of MPF as an indicator of neurodegeneration.

Because MPF strongly correlates with myelin content,^16,30,31 it has been employed to evaluate myelin in cortical and white matter regions.³² Our focus on the subcortical nuclei was motivated by the disease-specific pathology of PD, which predominantly affects these regions. While myelin is more abundant in white matter, the subcortical nuclei contain functionally substantial myelin components that are vulnerable to neurodegenerative processes. Previous quantitative MRI studies have demonstrated the clinical relevance of assessing gray matter myelin in neurodegenerative conditions,^16,32 supporting our methodological approach. Studies using R₁ values as an indicator of myelin content have demonstrated a significant reduction in myelin within subcortical regions of patients with PD.³³ In particular, the R₁ value was notably decreased in pathways originating from the basal ganglia (especially the substantia nigra and putamen), as well as in the brainstem and limbic system. Accordingly, the decrease in subcortical MPF found in this study likely reflects microstructure changes, e.g., reduced myelin content, as reported in previous investigations. While MPF strongly correlates with myelin content, the observed reductions in subcortical regions may reflect broader microstructural alterations beyond demyelination. Previous neurite orientation dispersion and density i maging (NODDI) studies in PD have demonstrated decreased intracellular volume fraction and increased orientation dispersion (OD) in affected brain regions,^34,35 indicating neuronal loss and microstructural disorganization. These changes, including reduced neuronal density, altered dendritic architecture, and protein aggregation, could collectively contribute to the decreased MPF values observed here.

In regions with high iron accumulation, elevated magnetic susceptibility values could theoretically cause local magnetic field variations that might influence MPF measurements. However, MPF has been demonstrated to be insensitive to iron content and magnetic susceptibility effects in neural tissues.^16,32 Unlike relaxation-based myelin measures that are confounded by iron deposition, MPF provides iron-insensitive quantitative assessment of macromolecular content.³² This iron insensitivity has been validated in studies of iron-rich subcortical gray matter structures in multiple sclerosis and other neurodegenerative conditions.³⁶

The relationship between MPF and neuromelanin deserves careful consideration, particularly in substantia nigra-related regions. Neuromelanin-sensitive MRI contrast partially relies on magnetization transfer effects and T₁ shortening, similar to the physical principles underlying MPF mapping. In PD, the loss of neuromelanin-containing dopaminergic neurons in the substantia nigra pars compacta could theoretically contribute to the observed MPF reduction in these regions. However, the following factors suggest that our MPF findings primarily reflect microstructural changes beyond neuromelanin loss alone. (1) MPF reductions were observed across multiple subcortical nuclei, including neuromelanin-lacking regions, such as the ventral pallidum and hypothalamus; (2) The spatial distribution of MPF changes extended beyond well-established neuromelanin-rich regions, such as the substantia nigra and locus coeruleus; and (3) previous validation studies have demonstrated the primary sensitivity of MPF to myelin and macromolecular content rather than specific pigments.^16,30,31 Nevertheless, in neuromelanin-rich regions, such as the substantia nigra pars compacta, the observed MPF reduction likely represents a composite effect of multiple pathological processes, including neuronal loss, myelin degradation, and neuromelanin depletion. Future studies combining MPF mapping with dedicated neuromelanin-sensitive imaging could help clarify these overlapping contributions.

Quantitative susceptibility measurements revealed significantly elevated values in the basal ganglia and midbrain regions, consistent with findings of previous QSM studies documenting iron accumulation in PD.^7,37,38 The spatial correspondence between regions exhibiting reduced MPF and increased susceptibility values suggests an underlying pathophysiological relationship. This finding aligns with research demonstrating iron-mediated neurodegeneration through oxidative stress mechanisms and protein aggregation.^13,39

ROC analysis showed that MPF mapping achieves diagnostic accuracy with AUC values exceeding 0.80, particularly in the ventral pallidum, hypothalamus, and red nucleus. Moreover, the diagnostic performance of a logistic regression model combining MPF and QSM measurements was superior to either measurement alone. These results are comparable to or exceed the diagnostic performance of existing quantitative MRI techniques,^10,12,40,41 and they suggest that combining MPF and magnetic susceptibility values is an effective predictor for PD diagnosis. The complementary nature of these measurements indicates potential diagnostic synergy, consistent with recent multimodal imaging approaches.^42,43

The pattern of MPF reduction across subcortical nuclei indicates more extensive involvement of brain structures than previously recognized. This finding supports the concept of PD as a multisystem disorder.^44,45 Additionally, the concurrent increase in magnetic susceptibility likely reflects the complex interplay between myelin degradation and iron dysregulation, potentially representing different stages of the neurodegeneration process.^11,46

The ability of MPF mapping to detect alterations in multiple subcortical structures suggests potential utility for early diagnosis, addressing a critical need in PD management. The widespread changes in MPF align with the diverse clinical manifestations of PD, including motor and non-motor symptoms.⁴⁷ The involvement of hypothalamic and limbic structures, e.g., the ventral pallidum and ventral tegmental area, may relate to non-motor manifestations, suggesting that MPF mapping could be valuable for monitoring disease progression across multiple domains.

Some limitations of this study warrant acknowledgment. First, the imbalance in sample size between the PD group (n = 20) and healthy control group (n = 9) represents a study limitation. While this disparity may have reduced statistical power for detecting subtle between-group differences, our use of non-parametric statistical tests helped mitigate potential issues with variance estimation. The observed large effect sizes and significant differences despite the smaller control group suggest that our main findings are robust. Future studies with larger, balanced cohorts would further strengthen these preliminary findings. Second, it should be noted that our cohort included predominantly patients with advanced-stage PD (Hoehn and Yahr stages 3–5), which limits the direct translation of our findings to early disease detection. However, this preliminary investigation successfully demonstrated that MPF mapping can differentiate patients with PD from healthy controls with high accuracy. Future studies should focus on patients with early-stage PD (Hoehn and Yahr stages 1–2) with adequate sample sizes to validate the utility of MPF mapping for prodromal disease detection. Third, owing to the limited sample size and predominance of advanced-stage PD cases, we were unable to assess the relationship between disease severity (Hoehn and Yahr stage) and MPF or QSM quantitative imaging parameters. Additionally, we did not assess Movement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale scores in this study, which represents a limitation for comprehensive clinical characterization. Future studies with larger cohorts spanning the full spectrum of disease severity are needed to determine whether MPF and QSM values correlate with clinical progression. All patients with PD were evaluated in their “on medication” state, which may have influenced our neuroimaging findings. Previous studies have suggested that dopaminergic medication can affect functional and structural MRI measurements,^48–51 although the specific impact on MPF and QSM remains unclear. Similar to findings in diffusion tensor imaging studies,⁵² the influence of medication state on MPF and QSM requires further investigation.

Conclusion

MPF mapping revealed significant alterations in multiple subcortical nuclei, providing novel insights into the pathophysiology of advanced stages of PD. When used in combination with magnetic susceptibility measurements, MPF shows strong potential as a quantitative diagnostic biomarker for PD. Further research should aim to validate these results in larger cohorts including those with early-stage PD and clarify the relationships between these imaging markers and clinical outcomes.