MRI sequences for detecting neuromelanin

The detection and estimation of signals at SN or LC differ depending on the MRI sequence used. In the research works reviewed, signals from SN and LC attributed to NM were detected using one or more of the following sequences: (a) 2D or 3D T1-weighted fast spin echo (FSE) and turbo spin echo (TSE), (b) 2D or 3D single-echo or multi-echo gradient echo (GRE) sequences with or without magnetization transfer (MT), (c) Delayed Alternating Refocusing Pulse Sequence (DANTE) T1 with Sampling Perfection with Application optimized Contrasts using different flip angle Evolutions (SPACE) and (d) T1-SPACE sequences.

In the reviewed articles, T1-FSE and TSE images were the most common anatomical MRI sequences used to visualize NM-rich structures and quantify the signal intensity. ^{Reference Kashihara, Shinya and Higaki12–Reference Moon, Park and Yun39} These sequences provide high contrast due to the T1 shortening effect and high isotropic resolution that is useful for volumetric assessments. The disadvantages include a high specific absorption rate (SAR), limited slice coverage, longer scan times and higher motion sensitivity. ^{Reference Balchandani and Naidich40,Reference He, Chen, LeWitt, Yan and Haacke41}

GRE sequences were used to detect SN and LC signals in a minority of studies. ^{Reference Oshima, Fushimi and Okada1,Reference Gao, Gaurav and Ziegner42–Reference Lakhani, Zhou and Tao46} GRE sequences are sensitive to differences in magnetic susceptibility, provide high contrast images and are suitable for visualizing NM, iron and other deposits. The use of off-resonance MT pulse further enhances the signal contrast. The multi-echo 3D GRE sequence with MT allows simultaneous imaging of NM-like substances and iron in the SN and optimizes contrast at lower flip angles and offers larger coverage of deep gray matter structures. ^{Reference Liu, Li and He47} Susceptibility-weighted imaging and quantitative susceptibility mapping (QSM) are other sequences that can indirectly assess NM through alterations in magnetic susceptibility. ^{Reference Liu, Buch and Chen48}

DANTE T1-SPACE is another specialized sequence for detection of SN and LC signals. ^{Reference Oshima, Fushimi and Okada1,Reference Langley, Hwang, Hu and Huddleston49} The DANTE suppression technique improves the image quality and suppresses motion artifacts, whereas SPACE offers high-resolution isotropic 3D images. Compared to T1-SPACE alone, DANTE-T1 SPACE sequence demonstrated increased signal intensity of the SN and is more suited for NM studies. ^{Reference Oshima, Fushimi and Okada1}

While studies have sparingly compared these sequences against each other, He et al. compared T1-TSE with GRE sequences and found the GRE sequence to reduce SAR while achieving comparable signal contrast. ^{Reference He, Chen, LeWitt, Yan and Haacke41} Hence, for comprehensive imaging of NM-rich areas, GRE with MT may be a better choice as it provides better coverage, less motion sensitivity and more detailed information on both NM and iron content, which is relevant for neurodegenerative disease research and diagnostics.

Quantification of neuromelanin

In the articles reviewed, irrespective of the sequence used to evaluate SN and LC, the methods of quantification involved (a) measurement of signal-to-noise (SNR) ratio, contrast-to-noise ratio (CNR) and contrast ratio (CR) or (b) morphometric estimation of SN and LC signals as volume or area in specific regions of interest (ROI). These measurement techniques were applied to a standard single or multiple axial slice(s) and were performed voxel-wise using manual tracing, atlas-based or template-based segmentation techniques.

In ROI-based techniques, areas with high signal intensity in SN and LC were traced manually or using automated or semi-automated software’s such as FreeSurfer or ITK-SNAP, and the signal intensity in the ROI is compared to a reference region (e.g., cerebral peduncles, tegmentum). ^{Reference Pyatigorskaya, Magnin and Mongin11,Reference Gao, Gaurav and Ziegner42,Reference Gaurav, Yahia-Cherif and Pyatigorskaya50–Reference Pyatigorskaya, Yahia-Cherif and Gaurav52} The relative difference in the signals is quantified as CNR, SNR or CR. Signal intensity can also be measured across several ROI and the mean signal may be used to represent signal intensity levels. Disadvantages of this method include inter-rater variability due to the subjective nature of manual tracing and its cumbersome nature, especially when analyzing large sample sets. Furthermore, the use of single slice or few slices may not cover the entire anatomy of SN or LC, leading to inconsistencies across studies. The presence of advanced neurodegeneration of catecholaminergic neurons in PD can also lead to biased estimates by potentially excluding severely impacted areas. ^{Reference Sulzer, Cassidy and Horga53}

In voxel-wise analysis, the MRI sequence is normalized to standard templates such as the Montreal Neurological Institute space or predefined templates such as PD25 or CTRL17 and then compared. This approach allows for consistent ROI placement and analysis is performed voxel-wise to derive SN morphometric measures such as volume or area. This increases the reliability and reproducibility across studies. Voxel-wise analysis also improves classification and correlation with severity of PD compared to other methods. ^{Reference Sulzer, Cassidy and Horga53}

As discussed above, numerous studies have reported that NM-MRI sequences quantify NM in the SN or LC. However, it remains uncertain whether the observed MRI signal changes can be attributed solely to reduced NM. Watanabe et al. ^{Reference Watanabe, Tan, Wang, Martinez-Hernandez and Frahm54} noted that MRI sequences with strong MT effects primarily reflect the water content and proton density, rather than NM content alone. They suggested that T1 hyperintensity changes in the SN and LC may be a reflection of the presence of noradrenergic neuronal cell bodies rather than accumulated NM.

Factors affecting neuromelanin and its estimation

Accurate estimation of NM content in the SN using MRI depends on multiple biological, pathological and technical factors. A key pathological consideration is the relationship between NM loss and iron accumulation. With neuronal loss in PD, the iron previously bound to NM is released and accumulates in the SN. This inverse relationship is supported by several studies showing elevated iron content in PD patients. ^{Reference Isaias, Trujillo and Summers23,Reference Takahashi, Watanabe and Tanaka33,Reference He, Chen, LeWitt, Yan and Haacke41} Takahashi et al. ^{Reference Takahashi, Watanabe and Tanaka33} used QSM to demonstrate a correlation between decreased NM signals and increased iron levels. Conversely, Reimao et al. ^{Reference Reimão, Ferreira and Nunes55} reported no significant association between NM-MRI signal and iron in early PD, where NM signals were reduced but iron content did not differ significantly from that of healthy controls. These discrepancies suggest that iron deposition might be non-linearly related to loss of signals at NM-rich areas or indicate at least partially independent mechanisms.

In addition to pathological changes, heterogeneous spatial signal distribution within the SN significantly affects estimation accuracy. Studies have documented that signals are lower in the dorsal SN and that disease-related loss typically begins in this region. ^{Reference Liu, Wang and Liu28,Reference Biondetti, Santin and Valabrègue56} Schwarz et al. ^{Reference Schwarz, Xing, Tomar, Bajaj and Auer15} reported that the greatest NM volume reduction occurred in the posterior SN, followed by the anterior SN and LC. Even in healthy individuals, NM-MRI studies show that posterior SN has lower baseline signal intensity across the lifespan, ^{Reference Xing, Sapuan, Dineen and Auer57} indicating a physiological vulnerability to depletion in this subregion. Importantly, nigrosome-1 – rich in dopaminergic neurons – has emerged as a focal point of NM loss in PD. Langley et al. ^{Reference Langley, Huddleston, Crosson, Song, Factor and Hu58} demonstrated that NM reduction occurs in regions corresponding to nigrosome-1 and nigrosome-2, whereas Biondetti et al. ^{Reference Biondetti, Santin and Valabrègue56} observed loss primarily in nigrosome-2, suggesting inter-individual variability. Overall, nigrosome-1 consistently exhibits the most significant NM depletion, and this loss aligns with iron accumulation patterns, ^{Reference He, Chen, LeWitt, Yan and Haacke41,Reference Biondetti, Santin and Valabrègue56,Reference Langley, Huddleston, Sedlacik, Boelmans and Hu59,Reference Langley, He and Huddleston60} supporting its role as a reliable biomarker for PD.

Technical aspects of NM-MRI also impact estimation accuracy. Factors such as scanner field strength, image resolution, motion artifacts and sequence parameters (e.g., magnetization transfer contrast, repetition time, echo time) can alter signal contrast. As highlighted by He et al., ^{Reference He, Chen, LeWitt, Yan and Haacke41} NM imaging contrast depends on proper sequence selection and anatomical localization, particularly in relation to the dorsolateral SNc and nigrosome-1. Langley et al. ^{Reference Langley, Huddleston, Crosson, Song, Factor and Hu58} further emphasized the use of multimodal imaging – combining NM-MRI and QSM – to improve localization of nigrosomal degeneration. Moreover, image analysis techniques – such as the method of ROI definition, signal normalization and partial volume correction – are critical in reducing variability and enhancing reproducibility. Together, these biological, pathological and imaging-related factors underscore the need for standardized protocols to ensure reliable NM quantification in PD.

Differentiation of Parkinson’s disease and healthy controls

SN: In most studies, the volume of SN hyperintensity was cross-sectionally evaluated between PD and healthy controls as the outcome of interest. Patients with PD had lower volume of NM-MRI signal compared to the healthy group. ^{Reference Pyatigorskaya, Magnin and Mongin11,Reference Kashihara, Shinya and Higaki12,Reference Liu, Yang and Liu16,Reference Isaias, Trujillo and Summers23,Reference Takahashi, Watanabe and Tanaka31,Reference Takahashi, Watanabe and Tanaka33,Reference Jokar, Jin and Huang43,Reference Ariz, Martínez, Alvarez, Fernández-Seara and Castellanos45,Reference Langley, Huddleston, Sedlacik, Boelmans and Hu59–Reference Gaurav, Valabrègue and Yahia-Chérif67} Similarly, SN width and SN area were also demonstrably lower in PD. ^{Reference Schwarz, Xing, Tomar, Bajaj and Auer15,Reference Le Berre, Kamagata and Otsuka18,Reference Wang, Yan, Zhang, Liu, Luo and Chang22,Reference Fabbri, Reimão and Carvalho37,Reference Reimão, Ferreira and Nunes55,Reference Reimão, Pita Lobo and Neutel68,Reference Schwarz, Rittman, Gontu, Morgan, Bajaj and Auer69,Reference Hartono, Chen and Welton70} When quantitated as CR, CNR, signal ratio or SNR ratio, the PD group had a lower ratio compared to healthy controls. ^{Reference Pyatigorskaya, Magnin and Mongin11,Reference Takahashi, Watanabe and Tanaka14,Reference Wang, Yan, Zhang, Liu, Luo and Chang22,Reference Isaias, Trujillo and Summers23,Reference Zhou, Wang and Cheng25,Reference Fabbri, Reimão and Carvalho37,Reference Ariz, Martínez, Alvarez, Fernández-Seara and Castellanos45,Reference Gaurav, Valabrègue and Yahia-Chérif67,Reference Schwarz, Rittman, Gontu, Morgan, Bajaj and Auer69,Reference Ohtsuka, Sasaki and Konno71–Reference Wiesman, Madge and Fon79} Hence, signal intensity, whether inferred from SN morphometry or from NM ratio, is significantly lower in PD compared to controls, pointing towards its diagnostic potential. Two studies evaluated the longitudinal changes of SN signal intensity in PD. In a study by Matsuura and colleagues, the area of NM-MRI signal in SN and CR was demonstrated to progressively decrease across 2.3 years of follow-up. ^{Reference Matsuura, Maeda and Tabei36} Gaurav et al. also recorded a decrease in signal volume in SN in 109 patients with PD during 2 years of follow-up. ^{Reference Gaurav, Yahia-Cherif and Pyatigorskaya50} These studies suggest a longitudinal decline in PD.

LC: When signal intensity in LC was compared, the CNR and CR were significantly lower in PD compared to controls in all studies except for Schwarz et al. who did not find any differences between PD and healthy controls. ^{Reference Madelung, Meder and Fuglsang9,Reference Isaias, Trujillo and Summers23–Reference Liu, Wang and Liu28,Reference Wang, Li and Huang34,Reference Zhou, Guo and Wu38,Reference Xing, Sapuan, Dineen and Auer57,Reference Cao, Han and Tang65,Reference Schwarz, Rittman, Gontu, Morgan, Bajaj and Auer69,Reference Wolters, Heijmans and Priovoulos78–Reference Langley, Hwang, Huddleston and Hu81} This may be because of a relatively small group of PD patients (PD:10, healthy controls:12) with a wide range of duration of disease (0.33 to 10 years) evaluated in their study.

Representative examples of MRI images demonstrating signal intensity in the SN and LC of healthy control and PD are given in Figure 2.

Figure 2.

Neuromelanin selective 3T MRI of a healthy individual at (A) the substantia nigra and (B) locus coeruleus. Substantia nigra (C) and locus coeruleus (D) of a patient with Parkinson’s disease showing decreased intensity of NM at both regions compared to the healthy control.

Relationship of neuromelanin with clinical parameters

Significant relationship between NM-MRI signals and clinical variables in PD has been reported by studies. The modified H&Y staging is reported to correlate negatively with NM-MRI signal, volume, area and SNR. ^{Reference Kashihara, Shinya and Higaki12,Reference Takahashi, Watanabe and Tanaka14,Reference Matsuura, Maeda and Tabei36,Reference Fabbri, Reimão and Carvalho37} Furthermore, disease duration also correlates with SN volume and area, CNR, CR and SNR. ^{Reference Kashihara, Shinya and Higaki12,Reference Wang, Wu, Cai, Yu, Chen and Wang26,Reference Liu, Wang and Liu28,Reference Matsuura, Maeda and Tabei36,Reference Ariz, Martínez, Alvarez, Fernández-Seara and Castellanos45,Reference Langley, Hwang, Hu and Huddleston49,Reference Gaurav, Yahia-Cherif and Pyatigorskaya50,Reference Prasad, Stezin and Lenka76} Compared to early PD, later stages of PD also show lower NM-MRI signal parameters at the SN compared to early PD. ^{Reference Takahashi, Watanabe and Tanaka14,Reference Liu, Wang and Liu28,Reference Fabbri, Reimão and Carvalho37,Reference Miyoshi, Ogawa and Kitao82} However, when comparing two earlier stages of PD, no significant difference was identified by Ohtsuka et al. ^{Reference Ohtsuka, Sasaki and Konno71} Another study comparing NM-MRI signals from de novo PD and early PD (3–5 years after motor onset) at the SN also did not identify any significant difference. ^{Reference Reimão, Ferreira and Nunes55,Reference Reimão, Pita Lobo and Neutel68} These results suggest a significant inverse relationship between NM-MRI signals and clinical stage of PD. However, the difference may not be pronounced in early clinical stages.

The NM-MRI signals also correlate with severity of motor symptoms. The Unified Parkinson’s disease rating scale (UPDRS) scores negatively correlate with signal intensity volume, area, CR, CNR and SNR. ^{Reference Wang, Wu, Cai, Yu, Chen and Wang26,Reference Fabbri, Reimão and Carvalho37,Reference Huddleston, Chen and Hwang44,Reference Ariz, Martínez, Alvarez, Fernández-Seara and Castellanos45,Reference Gaurav, Yahia-Cherif and Pyatigorskaya50,Reference Gaurav, Valabrègue and Yahia-Chérif67,Reference Prasad, Stezin and Lenka76,Reference Biondetti, Gaurav and Yahia-Cherif83} The L-dopa equivalent daily dose also correlates with SN area. ^{Reference Fabbri, Reimão and Carvalho37} At LC, CNR also correlates with UPDRS and disease duration. ^{Reference Liu, Wang and Liu28,Reference Nobileau, Gaurav and Chougar85}

The NM-MRI signals are also associated with clinical laterality. Castellanos et al. reported the single intensity volume of SN to be lower in the hemisphere contralateral to the side with more severe motor symptoms. ^{Reference Castellanos, Fernández-Seara and Lorenzo-Betancor32} Similar relationship was identified using CR by Xiang et al., ^{Reference Xiang, Gong and Wu77} Prasad et al. ^{Reference Prasad, Saini, Yadav and Pal75} and Wang et al. ^{Reference Wang, Li and Huang34}

Neuromelanin in motor and non-motor symptoms of Parkinson’s disease

Few studies have reported the NM-MRI signals to be associated with motor and non-motor symptoms of PD.

A study by Huddleston and colleagues compared signal intensity in SN between PD with and without freezing of gait (FOG) and healthy controls. ^{Reference Huddleston, Chen and Hwang44} While CR of entire SN and lateral-ventral tier of SN was significantly lower in PD with FOG compared with healthy controls, there was no significant difference in PD with and without FOG. They also showed that the severity of FOG measured using the FOG questionnaire significantly correlated with SN volume. ^{Reference Huddleston, Chen and Hwang44} In contrast, Wang et al. reported PD with FOG to have decreased volume and surface area at SN and LC compared to PD without FOG. These studies had similar demographic and clinical characteristics except for the MR sequence. ^{Reference Wang, Wu, Cai, Yu, Chen and Wang26}

Only few studies evaluated NM-MRI signals in subtypes of PD. Wang et al. evaluated postural instability gait difficulty motor subtype of PD and reported significant decline in the NM-MRI signal from lateral SN compared to tremor-dominant PD patients. ^{Reference Wang, Yan, Zhang, Liu, Luo and Chang22} However, Xiang et al. reported a difference in NM-MRI signals from medial SN between the two groups. ^{Reference Xiang, Gong and Wu77}

Among the non-motor symptoms, Palermo et al. evaluated difference in signal intensity volumes in PD with and without orthostatic hypotension but did not find any significant differences. ^{Reference Palermo, Galgani and Bellini80} In PD with RBD, Nobileau et al. demonstrated that SN signal intensity was lower compared to PD without RBD. ^{Reference Nobileau, Gaurav and Chougar85} The LC signal intensity and LC to pons ratio are also reported to be lower in the PD with RBD. ^{Reference Pyatigorskaya, Yahia-Cherif and Valabrègue51,Reference Sommerauer, Fedorova and Hansen86–Reference Horsager, Andersen and Knudsen88} NM-MRI signal from SN and LC is known to correlate with cognitive dysfunction in PD. ^{Reference Zhou, Guo and Wu38,Reference Huddleston, Chen and Hwang44} The severity of non-motor symptoms evaluated using Non-Motor Symptoms Scale score also showed a negative correlation with CNR at LC. ^{Reference Liu, Wang and Liu28}

Although at times conflicting, these results likely reflect the dopaminergic and non-dopaminergic origins of different non-motor and motor symptoms in PD. However, more research is necessary to ascertain the significance of these findings.

Neuromelanin in prodromal and preclinical Parkinson’s disease

Changes in signals of SN and LC in NM-MRI are also reported in prodromal and preclinical stages of PD such as iRBD and in carriers of genetic variations in PD genes.

SN: Patients with iRBD are reported to have volume, signal intensity, SNR and CNR values which is intermediate between healthy controls and patients with PD. ^{Reference Biondetti, Santin and Valabrègue56,Reference Biondetti, Gaurav and Yahia-Cherif83,Reference Nobileau, Gaurav and Chougar85–Reference Mangone, Houot and Gaurav90} The volume of SN signal intensity in hyposmic prodromal PD is also lower than that of healthy controls. ^{Reference Langley, Hwang, Huddleston and Hu81}

In a study by Gao et al., preclinical stage of LRRK2 had a signal intensity volume and CNR of SN intermediate between PD and healthy controls. ^{Reference Gao, Gaurav and Ziegner42} However, on comparing LRRK2 and GBA carriers, Langley and colleagues reported no difference in volume. ^{Reference Langley, Hwang, Hu and Huddleston49} Considering that research works by Castellanos et al. and Martinez et al. support loss of volume of signal intensity of SN in idiopathic PD and PD related to Parkin and LRRK2 mutations to be similar, it may be assumed that loss of signal intensity in SN is an ongoing process starting at preclinical and prodromal stages and worsening significantly in clinical stages of PD. ^{Reference Castellanos, Fernández-Seara and Lorenzo-Betancor32,Reference Martínez, Ariz and Alvarez91}

LC: Reflecting the results in SN, Knudsen et al. reported that patients with iRBD had significantly lower LC: pons signal intensity ratio. ^{Reference Knudsen, Fedorova and Hansen92}

While there is a paucity of literature on longitudinal changes in NM-rich structures across different stages of evolution of PD, there is converging evidence of declining trajectory of signal intensity in SN and LC from preclinical, prodromal and clinical stages. It may be hypothesized that there is a likely threshold at which the trajectories of normal people and those who will develop PD start diverging. Further studies are needed to test this hypothesis. However, if proven, these signals could be used as potential biomarkers in PD. However, much needs to be elucidated on NM changes in PD; variability in signals, differences in trajectory of signal loss and threshold of signal at which patients develop prodromal and symptomatic PD are key research questions.

Differential diagnosis of Parkinson’s disease

SN and LC signal intensity changes are reported not only in PD but also in other neurodegenerative parkinsonism such as MSA, PSP and in a few other neurological disorders.

SN: In most studies that evaluated morphometry of signal intensity of SN in PD and other neurodegenerative disorders, volume was lower in PD compared to MSA, PSP, Alzheimer’s disease (AD), vascular parkinsonism (VP), normal pressure hydrocephalus (NPH) and ET. ^{Reference Ohtsuka, Sasaki and Konno19,Reference Matsuura, Maeda and Yata29,Reference Wang, Huang and Li30,Reference Reimão, Pita Lobo and Neutel35,Reference Moon, Park and Yun39,Reference Lakhani, Zhou and Tao46,Reference Pyatigorskaya, Yahia-Cherif and Gaurav52,Reference Wang, Huang and Haacke93–Reference Matsusue, Fujihara and Tanaka95} Notable exceptions include a study by Kashihara and colleagues who concluded that volume of signal intensity of SN is reduced in PSP, MSA and CBD to the same extent as PD. ^{Reference Kashihara, Shinya and Higaki17} This observation is also supported partially by studies by Matsusue et al. (PD vs atypical parkinsonism), Simoes et al. (PD vs MSA) and Ikezawa et al. (PD vs dystonia). ^{Reference Simões, Castro Caldas and Grilo20,Reference Matsusue, Fujihara and Tanaka95,Reference Ikezawa, Yokochi and Okiyama96} These differences may be related to differences in sample size, disease characteristics such as duration and severity of disease and MRI protocols. In contrast to NM morphometry in SN, NM-MRI signals measured as SNR, CNR or CR are consistently lower in PD compared to other neurological disorders such as MSA, PSP, AD, VP, NPH and ET. ^{Reference Simões, Castro Caldas and Grilo20,Reference Wang, Huang and Li30,Reference Wang, Huang and Haacke93–Reference Matsusue, Fujihara and Tanaka95} Notable exception was a study by Lakhani et al., which did not report any difference in CR’s between PD and ET. ^{Reference Lakhani, Zhou and Tao46}

LC: Lower CR was reported in PD compared to PSP and MSA. ^{Reference Matsuura, Ii and Maeda13,Reference Ohtsuka, Sasaki and Konno19,Reference Matsuura, Maeda and Yata29} In ET, studies have demonstrated mixed results; Wang et al. reported lower CNR in PD compared to ET. ^{Reference Wang, Huang and Haacke93} In contrast, Liu et al. reported CNR of LC to be similar in PD and ET, but lower than in healthy controls. ^{Reference Liu, Zhou and Fang21} Comparing CR in AD and PD, Moon et al. identified the CR at LC to be lower in PD. ^{Reference Moon, Park and Yun39}

Only one study evaluated changes in signal intensity of SN in drug-induced parkinsonism. In this study by Matsuura and colleagues, ^{Reference Matsuura, Ii and Maeda13} the area of signal from SN and CR at LC was lower in PD compared to drug-induced parkinsonism. This possibly indicates a relatively unaffected SN and LC in drug-induced parkinsonism compared to PD. However, more studies are required to confirm this observation. Furthermore, concurrent comparison of NM content using MRI and dopamine transporter scans may also help differentiate these disorders using only MRI.