31P MR spectroscopy detected altered midbrain energy metabolism (lower Pi/PCr, higher ATP/Pi) in early-stage PD versus mimics, correlated with nonmotor/autonomic symptoms, and achieved high diagnostic accuracy (AUC 0.90) outperforming iron and neuromelanin MRI.

This provides a noninvasive, quantitative biomarker of mitochondrial/energy dysfunction with strong translational value for early differential diagnosis, patient stratification, and as a potential target-engagement or efficacy readout for metabolism-focused PD therapies.

Background Whether phosphorus 31 (31P) MR spectroscopy outperforms or complements conventional hydrogen 1 (1H) MRI biomarkers in the differential diagnosis of early-stage Parkinson disease (PD) remains unclear. Purpose To evaluate 31P MR spectroscopy and its integration with conventional 1H MRI for discriminating early-stage PD from mimics. Materials and Methods This prospective study consecutively enrolled participants with early-stage PD, participants with PD mimics, and controls (November 2023 to October 2024). Participants underwent 31P MR spectroscopy for inorganic phosphate (Pi), phosphocreatine (PCr), and adenosine triphosphate (ATP); quantitative susceptibility mapping for iron; and neuromelanin-sensitive MRI. Intergroup imaging differences were assessed using multivariable general linear models, and partial correlations with clinical scores were analyzed. A penalized logistic regression classifier evaluated discrimination performance of energy metabolites, alone and combined with iron and/or neuromelanin. Results Seventy-two participants with early-stage PD (mean age, 60.1 years ± 6.8 [SD]; 44 male participants), 34 with PD mimics (mean age, 62.1 years ± 7.9; 18 male participants), and 46 controls (mean age, 56.3 years ± 9.6; 33 female controls) were included. Early-stage PD showed a decreased Pi/PCr ratio (mean, 0.47 [95% CI: 0.44, 0.50] vs 0.57 [95% CI: 0.52, 0.63]) and an increased total ATP/Pi ratio (mean, 4.89 [95% CI: 4.57, 5.24] vs 4.06 [95% CI: 3.50, 4.68]) in the left midbrain compared with participants with PD mimics (both Bonferroni-corrected P < .05). The Pi/PCr ratio correlated with nonmotor symptom scores (r = 0.42; 95% CI: 0.12, 0.69; P = .008) and autonomic symptom scores (r = 0.60; 95% CI: 0.34, 0.80; P < .001) in PD. Energy metabolites outperformed both iron (area under the receiver operating characteristic curve [AUC], 0.90 vs 0.50; P < .001) and neuromelanin (AUC, 0.90 vs 0.64; P = .04) in differentiating PD from mimics. There was no evidence that combining iron and/or neuromelanin improved AUC over energy metabolites alone (0.90 vs 0.90 with the addition of iron [P > .99] vs 0.91 with the addition of neuromelanin [P = .70] vs 0.93 with the addition of both [P = .42]). Conclusion 31P MR spectroscopy revealed disrupted midbrain energy homeostasis in early-stage PD and effectively helped differentiate it from mimics. © RSNA, 2026 Supplemental material is available for this article.