ABSTRACT
Objective
Progressive supranuclear palsy (PSP) with ventricular dilatation and idiopathic normal pressure hydrocephalus (iNPH) may present with overlapping symptoms, particularly when gait disturbance predominates. This study compared midbrain qualitative signs and quantitative metrics to identify the most accurate marker for differentiating PSP from iNPH.
Methods
We retrospectively enrolled 32 patients with PSP, 33 with shunt-predictive iNPH, and 35 age-matched healthy controls, all of whom fulfilled established clinical criteria. High-resolution isotropic magnetization prepared rapid acquisition gradient echo (1.0 mm voxel size) was used to evaluate the hummingbird sign (HBS), morning glory sign, midbrain-to-pons area ratio (M/P), Magnetic Resonance Parkinsonism Index (MRPI), and its updated version, MRPI 2.0. Binary logistic regression analysis was performed to assess the predictive value of midbrain metrics and qualitative features.
Results
All imaging markers differed significantly among PSP, iNPH, and control subjects (p<0.01). The M/P ratio, MRPI, and MRPI 2.0 effectively distinguished all group pairs, with the M/P ratio demonstrating fair accuracy [area under the curve (AUC)=0.81; accuracy 74%]. Both MRPI versions achieved acceptable performance (AUC<0.78). The HBS demonstrated good discriminatory performance in differentiating PSP from iNPH (AUC=0.82; accuracy 81%) and emerged as the sole independent predictor in logistic regression [AUC=0.87; odds ratio (OR): 32; p<0.01]. The MRPI 2.0 outperformed all qualitative features (AUC=0.94; OR: 18; p<0.01) in distinguishing iNPH from healthy individuals.
Conclusion
Quantitative metrics for Parkinsonism showed stepwise differences across PSP, iNPH, and control groups. The HBS outperformed all quantitative measures in distinguishing PSP from iNPH when a predefined point-based system was applied.
INTRODUCTION
Progressive supranuclear palsy (PSP) is a primary tauopathy characterized by abnormal tau deposition leading to progressive neurodegeneration (1). Clinically, PSP presents with postural instability, gait disturbance, cognitive impairment, urinary symptoms, and frontal executive dysfunction; vertical supranuclear gaze palsy is a hallmark feature (2). However, early gaze palsy may be subtle, resulting in low sensitivity of the National Institute of Neurological Disorders and Stroke-Society for PSP (NINDS-SPSP) diagnostic criteria despite high specificity (3). This diagnostic challenge is particularly evident in Parkinsonism-predominant PSP from other parkinsonian disorders (4). Consequently, magnetic resonance imaging (MRI) plays an important role in early diagnosis, with qualitative imaging signs, such as the hummingbird sign (HBS) and the morning glory sign (MGS), proposed as suggestive markers of the PSP spectrum (5).
Idiopathic normal pressure hydrocephalus (iNPH), which has an estimated annual incidence of 5.5 per 100,000 and is considerably more common than PSP, is thought to result from impaired cerebrospinal fluid (CSF) absorption and altered brain compliance (6, 7). Clinically, iNPH is characterized by gait instability, urinary incontinence, and frontal executive dysfunction (8). MRI features such as disproportionately enlarged subarachnoid space hydrocephalus (DESH), narrowed callosal angle, ventricular enlargement, and periventricular hyperintensities support the diagnosis (9). iNPH may be misdiagnosed as PSP due to overlapping clinical features, particularly parkinsonism and gait disturbance. Early differentiation often requires prolonged clinical surveillance or invasive CSF testing, and coexistence of PSP pathology and iNPH-related CSF disturbance has been reported (10). An accurate distinction is essential because treatment strategies differ substantially between the two conditions (11, 12).
MRI assessment of PSP traditionally relies on qualitative midbrain atrophy features, including mesencephalic beaking and loss of lateral tegmental convexity (13). However, these signs may be subtle or absent in early or Parkinsonism-predominant cases, limiting the sensitivity of qualitative assessments. Consequently, several midbrain morphometric indices were proposed, including the MR Parkinsonism Index (MRPI), the midbrain-to-pons area ratio (M/P), and the more recent MRPI 2.0. Although initial reports suggested improved performance of MRPI 2.0 in distinguishing PSP from idiopathic Parkinsonism, subsequent studies have questioned its superiority in differentiating Parkinsonism-predominant PSP from other Parkinsonian disorders (14, 15). Notably, the application of quantitative midbrain indices, including MRPI 2.0, for differentiating PSP from iNPH has received relatively limited attention in the literature. Therefore, the objective of this study was to investigate the diagnostic value of midbrain morphometric indices, including MRPI 2.0, when combined with classical qualitative MRI features, in differentiating the PSP spectrum from iNPH.
METHODS
Study Group Selection
This retrospective study was carried out at a single center between January 2018 and December 2024. This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Non-Interventional Clinical Research Ethics Committee of Koç University under reference (approval no: 2025.318.IRB2.145, date: 11.07.2025). Patient data were reviewed using the institution’s digital medical record system. Inclusion criteria for the PSP group were the fulfillment of the clinical diagnostic criteria for probable PSP spectrum, according to the neuropathologic guidelines of the NINDS-SPSP, and the availability of isotropic T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) imaging at diagnosis (3). All clinical diagnoses of PSP were confirmed by a neurologist specializing in movement disorders (Eser Buluş). Initially, 41 patients with probable PSP were identified; however, only 32 (14 females, 18 males) had corresponding MRI available at the time of diagnosis and were included in the study. Similarly, 46 patients presenting with the full clinical triad suggestive of iNPH—gait disturbance, urinary incontinence, and cognitive impairment—were identified in the digital medical records. However, 13 of these did not demonstrate the supportive imaging features required for a diagnosis of shunt-predictive probable iNPH, as outlined in the criteria endorsed by the Japanese Society of Normal Pressure Hydrocephalus (16). Those imaging requisites include disproportionately enlarged subarachnoid spaces, an Evans index >0.3, and a steeply narrowed callosal angle [non-compartmental analysis (NCA) <90º]. The remaining 33 patients (14 females, 19 males) fulfilled the complete clinical triad and all required imaging features for shunt-predictive probable iNPH. Among these, 15 patients underwent ventriculoperitoneal shunt surgery. All included iNPH patients had pre-intervention MRI available, including isotropic T1-weighted MPRAGE sequences. Exclusion criteria for both groups included the absence of MRI at the time of diagnosis, the presence of space-occupying lesions or large areas of encephalomalacia, indicative of prior major trauma, large-vessel occlusion, or hemorrhage. However, no participants from either group met these exclusion criteria. For the control group, 35 age-matched healthy subjects (15 females, 20 males) were selected from individuals who attended the neurology outpatient clinics during the same period and presented with symptoms indicative of a non-neurodegenerative condition, including tension-type headache and migraine-type headache, or who attended for routine check-up. Control subjects were included only if their brain MRI examinations were normal, with no evidence of localized or diffuse abnormalities. Figure 1 illustrates the participant selection process with a flow diagram.
The Magnetic Resonance Parkinsonism Indices
All planimetric measurements were ultimately performed using isotropic T1-weighted MPRAGE images acquired in the sagittal orientation, with an isotropic voxel size of 1.0 mm, bandwidth of 180 Hz/pixel, repetition time/echo time of 2400/3.7 ms, field of view of 250×250×160 mm, matrix size of 205×256×256, and a flip angle of 8°. The isotropic MPRAGE datasets were subsequently reformatted into coronal-oblique, sagittal, and axial planes to enable measurement of midbrain area (M), pons area (P), lengths of the superior cerebellar peduncle (SCP) and middle cerebellar peduncle (MCP), third ventricular width (V3), and the left-to-right width of the frontal horns of the lateral ventricles (FH). The MRPI was calculated using the formula: MRPI=(P/M)×(MCP/SCP). The MRPI 2.0 was derived by incorporating ventricular dimensions into that formula: MRPI 2.0=MRPI×(V3/FH). All measurements were performed manually by a fellowship-trained neuroradiologist with 7 years of experience in the field. Figure 2 illustrates the planimetric measurements used to calculate MRPI and MRPI 2.0 in patients with PSP, patients with iNPH, and control subjects. The M/P, a conventionally recommended metric for differentiating PSP from other atypical Parkinsonian syndromes, was also calculated (17). All planimetric measurements were repeated after 4 weeks to assess the intra-rater agreement.
Qualitative Parameters
The HBS was independently evaluated by two experienced neuroradiologists, Yunus Emre Şentürk (6 years of neuroimaging experience) and Ahmet Peker (5 years of neuroimaging experience), according to the ordinal scoring system described by Kim et al. (18). This system evaluates four features: the shape of the third ventricle, the contour of the beak, the configuration of the midbrain, and the overall sagittal appearance of the midbrain-to-pons interface. Each feature was scored on a scale from 0 to 2. A total composite score >5 was considered indicative of a positive HBS, as this threshold yielded the best performance compared with other thresholds when accounting for the Kim et al. (18) series. Similarly, the MGS was independently evaluated by two raters (Yunus Emre Şentürk and Ahmet Peker) on axially reformatted MR images, in accordance with the method outlined by Adachi et al. (19). An imaginary horizontal line was drawn posterior to the cerebral aqueduct. A second line was then extended from the lateral margin of the cerebral peduncle to the point where the first line intersected the tegmentum. The MGS was considered positive when the lateral tegmental border of the midbrain was located medial to this second line. The HBS and MGS statuses of all cases were re-evaluated by the central rater after a minimum interval of four weeks to assess intra-rater agreement.
Statistical Analysis
All statistical analyses were performed using SPSS version 28 (IBM Corp., Armonk, NY, USA). The Shapiro-Wilk test was used to assess the normality of all continuous variables. Parameters following a normal distribution, such as the M/P, MRPI, and MRPI 2.0, were compared across groups using one-way analysis of variance, with post-hoc pairwise comparisons using Tukey’s honestly significant difference test. The Cohen’s kappa analysis was performed for all patients with PSP spectrum and iNPH to assess the inter- and intrarater reliability of HBS and MGS between the outcomes of the central (Yunus Emre Şentürk) and the co-rater (Ahmet Peker). Intrarater reliability was assessed using a two-way mixed-effects intraclass correlation coefficient for absolute agreement. Variables that demonstrated significant differences were subsequently included in a backward stepwise logistic regression (BSLR) model to evaluate the discriminatory power of individual planimetric features for differentiating PSP from iNPH, and iNPH from control subjects. A p-value of <0.05 was considered statistically significant.
RESULTS
Demographics
Baseline characteristics and symptoms on admission are presented in Table 1. Each group’s mean age resembled that of the others, while the patients with PSP had a wider age range, with a mean age of 73±10 years. The gender distribution was similar across groups, as shown in Table 1.
The Quantitative Parkinsonism Features
The Evans index exceeded 0.3 in all iNPH patients and in 48% of those with PSP. The M/P ratio, MRPI, and MRPI 2.0 differed among groups based on analysis of variance statistics (p<0.001, for each analysis). Table 2 summarizes the distribution of means for each midbrain metric parameter. Post-hoc pairwise comparisons revealed notable mean differences in all three midbrain metrics between PSP and iNPH, and between iNPH and control subjects (p<0.001). Figure 3 presents box plots of the distribution of each midbrain metric across groups. These findings indicate a clear stepwise separation of the three groups, with iNPH values consistently falling between those of PSP and control subjects. All quantitative parkinsonism parameters demonstrated excellent intra-rater reliability, with kappa (κ) values of 0.94 for the M/P ratio, 0.92 for MRPI, and 0.91 for MRPI 2.0 (all p<0.001).
The Qualitative Assessment of the Midbrain
The presence of the HBS in each group was evaluated using the visual scoring system proposed by Kim et al. (18), in which a composite score of ≥5, agreed upon by both raters, was considered HBS-positive. HBS was identified in 29 patients with PSP (90.6%) and 9 patients with iNPH (27.3%); none of the control subjects exhibited this sign (p<0.001). In contrast, the MGS was less frequently observed, with 15 PSP patients (47.0%) and 4 iNPH patients (12.1%) classified as MGS-positive (p=0.003). None of the control subjects had a positive MGS.
The central rater identified HBS positivity in 39 (60%) cases, while the co-rater identified 36 (55%) cases. Interrater reliability for HBS was almost perfect (κ=0.85; p<0.001), with almost perfect intrarater agreement (κ=0.89; p<0.001). For the MGS, the central rater and co-raters identified 19 (29%) and 16 (24%) positive cases, respectively. Interrater agreement was substantial (κ=0.77; p<0.001), and intrarater agreement was almost perfect (κ=0.85; p<0.001).
Performances of Qualitative and Quantitative Parameters
The receiver operating characteristic analysis for both quantitative parameters is summarized in Table 3 and Figure 3. The M/P demonstrated good discriminatory performance between the PSP and iNPH groups, with an area under the curve (AUC) of 0.81 and an accuracy of 0.74. Although slightly lower than the M/P, MRPI and MRPI 2.0 also demonstrated fairly good performance, with AUCs of 0.77 and 0.74, and accuracies of 0.77 and 0.72, respectively. Overall, the midbrain morphometrics demonstrated superior performance in differentiating iNPH from control subjects. Among these, the MRPI and MRPI 2.0 demonstrated excellent discriminatory power, with AUCs of 0.93 and 0.92 and accuracies of 0.88 and 0.87, respectively. The M/P also performed well, though slightly below the threshold for excellence, achieving an AUC of 0.90 and an accuracy of 0.82. In the qualitative assessment of midbrain shape, the HBS demonstrated good diagnostic utility for distinguishing PSP from iNPH (AUC HBS: 0.82; AUC MGS: 0.67). In addition, HBS and MGS demonstrated poor efficacy in distinguishing iNPH from control subjects (AUC HBS: 0.62; AUC MGS: 0.55).
Logistic Regression Analysis
To evaluate the combined predictive value of quantitative midbrain metrics and qualitative imaging signs in distinguishing iNPH from PSP, a BSLR was performed. Collinearity was identified among the morphometric parameters because MRPI values are derived from the M/P. MRPI 2.0 was therefore retained due to its inclusion of ventricular dimensions. The final model combining MRPI 2.0, HBS, and MGS showed strong performance (Nagelkerke R2=0.41; AUC=0.87; accuracy=81%). HBS was the strongest independent predictor of PSP [odds ratio (OR): 32; p<0.001), followed by MGS (OR: 9; p=0.021). The MRPI 2.0 was not an independent predictor (p=0.97).
A second BSLR model differentiating iNPH from controls similarly excluded the M/P and MRPI due to collinearity. MRPI 2.0 emerged as the only significant predictor (OR: 18.4), favoring an iNPH diagnosis over controls (Nagelkerke R2=0.52; AUC=0.93; accuracy=87%).
DISCUSSION
The current study demonstrates that midbrain morphometric measurements are useful for differentiating the PSP spectrum from iNPH. Among quantitative indices, the M/P showed the highest overall accuracy in favoring PSP, whereas both MRPI and MRPI 2.0 demonstrated favorable but limited performance. Specifically, an M/P <0.2 yielded high sensitivity (84%) but poor specificity, whereas MRPI and MRPI 2.0 exhibited nearly 90% specificity at the expense of low sensitivity. These complementary performance patterns suggest that combining quantitative parameters may be more informative than relying on a single metric.
Among the qualitative parameters, only the HBS achieved good performance. When all quantitative and qualitative factors were combined, diagnostic performance improved, and the HBS emerged as a hallmark predictor distinguishing the PSP spectrum from iNPH. A challenge in interpreting HBS is the lack of a universally accepted verification method for equivocal cases. In this study, the systematic approach of Kim et al. (18) was applied, resulting in three clinically overt PSP cases being classified as HBS-negative. This stricter definition, favoring definite midbrain atrophy, likely contributed to the enhanced diagnostic performance of HBS in the combined regression model. Consistent with prior work by Kim et al. (18), who reported AUC values of 0.72-0.76 for HBS in PSP versus idiopathic parkinsonism, our cohort demonstrated an even higher AUC of 0.82 for distinguishing PSP from iNPH. Therefore, the diagnostic performance of HBS appears superior to the midbrain metric markers when either an informative scoring system is used or a scoring system is applied systematically.
Among quantitative parameters, the simpler M/P slightly outperformed the more complex MRPI and MRPI 2.0, despite MRPI 2.0 incorporating ventricular measurements to enhance discrimination. In our cohort, MRPI 2.0 showed poor sensitivity, which was likely influenced by ventricular enlargement observed in 48% of PSP cases, in which the Evans index exceeded 0.3. This ventriculomegaly may reduce the discriminatory advantage of the ventricular component of MRPI 2.0. Consistent with our results, Constantinides et al. (20) reported that the M/P outperformed MRPI, with both parameters showing comparable moderate-to-good diagnostic performance. As in their study, the combined use of M/P and MRPI failed to produce high specificity (>80%), likely due to collinearity between variables. Another series also questioned the performance of the MRPI, reporting a cutoff >13 with high specificity (84%) but low sensitivity (29%) (10). In contrast, our higher cut-off of 18.9 improved sensitivity but remained insufficient for standalone clinical decision-making compared to previous reports. A small study of 19 PSP and 17 iNPH cases reported considerable overlap between PSP and iNPH using MRPI 2.0 without providing diagnostic performance measurements (21). In contrast, the clearer stepwise differences observed across all quantitative metrics among the PSP, iNPH, and control groups are likely attributable to the larger sample size of our study.
Although MRPI 2.0 was designed to improve differentiation by incorporating ventricular dimensions, our findings indicate performance comparable to the original MRPI in distinguishing PSP from iNPH. Ventricular enlargement resembling iNPH has been reported in up to 38.2% of PSP cases, particularly in the Parkinsonism-predominant subtype. iNPH has been identified as one of the most frequent mimics of PSP (22, 23). This overlap likely limits the added value of MRPI 2.0 in this context. Nonetheless, MRPI 2.0 has demonstrated excellent performance in distinguishing PSP from Parkinson’s disease in larger series (24). MRPI 2.0 values were significantly higher in iNPH patients than in controls and outperformed HBS in distinguishing the two groups in our cohort. While classical markers such as NCA, DESH, and periventricular hyperintensities remain sufficient for diagnosing probable iNPH, MRPI 2.0 also achieved a sensitivity of 94% and specificity of 80% at a cutoff of 2.5, suggesting its potential utility as a supportive imaging marker in diagnosing probable iNPH.
Study Limitations
Our cohort has several limitations. First, the study design was retrospective and performed at a single center. Second, although all cases met established diagnostic criteria, complete exclusion of overlapping PSP and iNPH pathology is not possible because shunt-responsive PSP has been reported in the current literature (25). Third, the low sample size limited the performance of multivariable regression models; however, overfitting was minimized by restricting the number of variables included. Finally, PSP subtypes were not analyzed separately, although prior studies indicate no significant differences in MRPI among those PSP subtypes (15).
CONCLUSION
Quantitative midbrain indices, including MRPI 2.0 and previous morphometric parameters, showed a stepwise distribution among PSP, iNPH, and healthy control subjects. MRPI 2.0 did not outperform the original MRPI, and the simpler M/P ratio showed slightly better discriminatory power for distinguishing PSP from iNPH. Of all imaging features, the HBS remained the most reliable marker and an independent predictor of PSP versus iNPH.
