Quantitative Analysis Versus Visual Assessment of Neuromelanin MR Imaging for the Diagnosis of Parkinson’s disease

INTRODUCTION

Recent MR studies have described a reduction of neuromelanin signal in the substantia nigra (SN) of Parkinson’s disease (PD) patients [1–3] with high diagnostic accuracy even in early disease stages [4, 5]. These findings reflect the known pathological features of PD with dopaminergic neuronal loss in the SN pars compacta and consequent reduction of neuromelanin content [6]. This neuronal loss occurs asymmetrically and primarily in the ventrolateral and caudal region of the SN with relative preservation of the dorsal and rostral region [7], and MR-sensitive neuromelanin studies have been able to identity these specific disease changes [8].

PD diagnosis in the early disease stage can be challenging but has important therapeutic and prognostic consequences [9]. For this reason, neuromelanine sensitive-MR imaging may have the potential to become an alternative to the more expensive and risky nuclear medicine techniques [10] currently used in the diagnostic evaluation of early stage PD.

Previous neuromelanin MR studies have used automated post-processing [5] or manual ROI tracing of the SN region [4, 8] but these methods are time consuming or need additional post-processing software. Since this technique may have a significant impact on clinical patient management, an easy and widely available mechanism for image analysis is needed to enable the routine use of these MR images.

In this study we examined neuromelanin-sensitive MR imaging in “de novo” untreated PD patients and in PD patients with a 2–5 year disease duration, comparing quantitative width measurement with visual image assessment by experienced neuroradiologists to ascertain their ability to detect SN neuromelanin changes and to differentiate between PD early-stage patients and controls.

To the best of our knowledge this is the first study in early stage PD patients comparing different neuromelanin MR imaging evaluation methods, in particular using visual assessment or manual widthmeasurements.

PATIENTS AND METHODS

RESULTS

MR imaging quality allowed the identification of the SN high signal area and width measurements on all subjects. We analyzed 12 patients with “de novo” PD, 10 patients with 2–5 year disease duration and 10 healthy controls whose clinical characteristics are shown in Table 1. No significant differences were observed in age and sex among the 3 groups, and there were also no significant differences in H&Y stage and UPDRS scores between the “de novo” and 2–5 years disease duration PD patients. All PD patients had mild disease severity with H & Y of 1-2 and mean UPDRS motor score of 29.12 (±12.92). Symptoms at onset were predominantly unilateral with half of the “de novo” PD patients presenting criteria for asymmetric disease on UPDRS part III [12].

Quantitative width measurements of SN T1 high signal are shown on Table 2. The median global averaged width of the T1 high signal in the SN was significantly decreased in both PD groups when compared with healthy controls with a p-value<0.001 and 0.002 respectively (Fig. 2). This pattern of width differences between the groups was more pronounced in the lateral, rather than the central, SN segments (Table 2 and Fig. 2) and the medial SN width was only decreased significantly when comparing the 2–5 year PD group with the controls (p-value = 0.002). We found no significant differences between left and right SN width for all the segments (p-value = 0.714). A substantial overlap was observed between the “de novo” and the 2–5 year PD groups, with no significant differences in the width measurements.

ROC analysis is detailed in Table 3 and indicated good area under the curve (AUC) values for the lateral segment width in both PD patients groups. The sensitivity and specificity of the width measurement in the lateral part of the SN was 88% and 65% respectively, when the cut-off value for SN width was set at 1.86 mm for discriminating the “de novo” PD group vs. controls and were both 75% , with a width cut-off value of 1.97 mm, for discriminating between the 2–5 year PD group and controls (Table 3).

The visual assessment results of the SN high signal area for the two raters are shown in Table 4. The neuromelanin size/signal was decreased in the vast majority of PD patients (only rated as normal in 2 PD “de novo” patients). Marked asymmetry was found in almost all the 2–5 year PD patient’s but only in half of the “de novo” PD group. Raters independently classified as “probable/possible PD” 17/22 of PD patients (rater 1) and 20/22 (rater 2); 6/10 and 8/10 of the controls were rated as “not PD”. No PD patient was classified as “not PD” and no healthy control was classified as “probable PD”. Agreement between observers was assessed with Cohen’s Kappa coefficient showing a moderate agreement for subject’s classification with a coefficient of 0.45 (0.25–0.66) and fair agreement for the size/signal of the SN T1 high signal area evaluation with a coefficient of 0.39 (0.059–0.73)

DISCUSSION

Visual assessment of neuromelanin-sensitive MR images by experienced neuroradiologists was able to detect changes in the SN of early-stage PD patients with a discrimination power similar to that of aquantitative measurement of T1 high signal width. The sensitivity, specificity and AUC of the SN neuromelanin width measurements for the distinction of early stage PD patients from healthy controls were comparable to the PD patient’s identification rates obtained by visual assessment.

Neuroradiologists were also able to detect a greater reduction of neuromelanin in the lateral part of the SN in PD patients and a less pronounced change in the central and medial segments, in accordance with the quantitative width measurements and reflecting the known pathological gradient of neuronal depletion in PD [13]. There was a moderate and fair inter-rater agreement for subject’s classification and size/signal of the T1 high signal area evaluation respectively, with no misclassifications of PD patients as “not PD” and no healthy control classified as “probable PD” for either rater. Although the quantitative width measurements were not significantly different between the “de novo” and the 2–5 year PD groups, on visual inspection the 2–5 year PD patients were classified with greater certainty as “probable PD” that the “de novo” PD patients. We think that these findings may reflect the more global analysis provided by visual inspection as it takes into account several parameters such as size, signal and morphological data, allowing the identification of a disease pattern. These data are in agreement with neuropathology findings in PD which correlate the decrease of SN neuromelanin with clinical severity [7]. However this decrease was not reflected in our quantitative measurements which were similar to previous reports of SN neuromelanin signal intensity changes [4] and transcranial sonography [14] in PD suggesting a stability of the neuromelanin content with disease duration. The different imaging and post-processing methods used in these studies may explain the inconsistent findings in neuromelanin patterns with disease duration [2, 15] and additional research on neuromelanin MRI signal changes throughout PD evolution is needed.

PD motor symptoms are usually asymmetric at the time of clinical presentation and throughout the course of the disease [11, 16]. We were not able to find a significant asymmetry in the SN neuromelanin nor a correlation with the clinically most affected side in our PD patients either with visual inspection or with quantitative analysis. This may be due to the small number of subjects in our samples which limits the power of the study.

Previous neuromelanin MR studies have used quantitative measurements with automated analysis or manual ROI tracing [4,5,8, 4,5,8]. These post-processing techniques are time consuming or need post-processing software which limits clinical implementation of neuromelanin imaging. We used manual width measurements that, although operator-dependent, are fast and widely available and additionally allow the study of the different segments of the SN. Visual image inspection was also performed, since it can give an even faster and easier evaluation of the T1 high signal in the SN. In our study, visual inspection was able to identify a pattern of SN neuromelanin that allowed a good discrimination between PD patients and controls with high agreement with quantitative measurements. Additionally, the sensitivity and specificity obtained for these MRI neuromelanin studies are only slightly lower than those obtained in [¹²³I]- FP-CIT SPECT studies in identifying PD patients [10, 17]. Finally, neuromelanin sensitive MRI has a relatively lower cost compared with PET or SPECT imaging and can be applied to all compliant patients with few involuntarymovements and no MRI contraindications, unlike transcranial sonography where up to 18% of individuals have insufficient temporal acoustic windows [18]. The different neuromelanin imaging evaluation methodologies used in our research can also be used to study other conditions, namely atypical parkinsonian syndromes, a future line of research that can have a significant impact in clinical patient management.

Our study has several limitations, mainly the small number of patients. The imaging technique requires a 3.0 Tesla MR machine which has an impact on availability and the neuromelanin sequence has a long acquisition time, a relatively low spatial resolution and is prone to in-plane signal inhomogeneity. In addition, we used a fully manual, operator dependent image analysis technique that can have a high variability. The intra and inter-rater reproducibility of MR neuromelanin measurements and visual inspection have not been fully established, and so the reproducibility of these parameters needs to be addressed in further studies, especially concerning visual inspection, in order to allow a wider implementation of this imagingtechnique.

However, we obtained a good agreement in visual assessment between the two raters and the findings are consistent with quantitative data and with previous studies using different methods and techniques [2, 5].

Motion artifacts were not a problem for visual inspection or width measurement in our PD patients groups, but these were early stage H&Y 1-2, with few involuntary movements. Image artifacts may be an issue in advanced-stage PD patients interfering with image evaluation.

Neuromelanin MR imaging can have a significant impact on early stage PD diagnosis, however there are some mismatches between neuromelanin MR imaging changes and the clinical diagnosis of our subjects, a fact that may be intrinsically related to neuromelanin alterations in PD. Degenerated nigral cells may still contain neuromelanin granules influencing MRI signal measurements and the relationship between the extent of dopamine cell loss and PD’s clinical manifestations remains currently unclear with previous dopaminergic imaging studies showing that some PD patients with mild symptoms have no evidence of dopaminergic dysfunction [19].

Using a fast, easy, highly available assessment of SN neuromelanin with visual inspection or width measurement, we were able to detect a disease pattern change in early stage PD patients. Visual inspection of MR sensitive neuromelanin imaging of the SN may become an attractive tool in clinical practice for PD diagnosis, but additional studies are needed to validate this method and to determine how this imaging technique can affect the management of PD patients.

CONCLUSION

Visual inspection of neuromelanin-sensitive MRI techniques has similar results compared to a quantitative width analysis of SN changes in early-stage PD, allowing the identification of a specific disease pattern and a good distinction of PD patient’s from healthy controls. These findings can have an impact on a wider use of neuromelanin MR imaging in clinical practice for PD patients evaluation.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.