Introduction

Progressive supranuclear palsy (PSP) is a rapidly progressive neurodegenerative disorder characterized by the accumulation of 4-repeat tau protein in the brain, leading to typical clinical features including vertical supranuclear gaze palsy, postural instability, and cognitive impairment1. Despite well-defined diagnostic criteria, differential diagnosis between PSP and Parkinson’s disease (PD) remains challenging, particularly for atypical variants of PSP such as PSP-Parkinsonism (PSP-P)2,3. This highlights the urgent need to develop objective, easily accessible biomarkers to improve diagnostic accuracy and enable early intervention4,5.

Bradykinesia is a core feature of both PD and PSP, which is defined as slowness of movement and a decrease in amplitude or velocity (or progressive hesitations/halts) as movements are continued6. It is now recognized as a complex motor impairment encompassing bradykinesia (reduced movement velocity), hypokinesia (reduced movement amplitude), sequence effect (progressive reduction in amplitude or velocity), hesitations (irregularities in movement)7,8. The finger-tapping task is widely used in clinical practice to assess bradykinesia9. However, most studies have focused on PD. Only Ling et al. and Djuric-Jovicic et al. have reported the absence of sequence effect as a key finger-tapping feature of PSP compared to PD, using wearable devices10,11. Results regarding the finger tapping amplitude, rhythm and velocity between PSP and PD remain controversial10. Furthermore, the wearable devices could affect the finger movement to some extent and are not convenient for outpatient work. Recent advances in computer vision enable precise quantification of movement parameters (amplitude, velocity, duration) using simple video recordings12,13.

To date, no study has explored the relationship between finger-tapping impairments and clinical symptoms or the underlying neural mechanisms in PSP11. The clinical significance of finger-tapping in PSP remains controversial14,15,16. In PD patients, neuroimaging studies indicated cortico-striatal and cerebellar-thalamic circuits affect the finger-tapping movement. Cerebellum and brainstem selectively atrophy in PSP17,18,19,20. However, there was no direct study combined the bradykinesia or finger tapping of PSP patients with the specific patterns of brain atrophy. Thus, an integrated analysis of the association between finger-tapping parameters and clinical characteristics, as well as PSP-specific brain atrophy patterns, holds significant clinical promise.

Herein, we employed video-based kinematic analysis to quantify and compare finger-tapping parameters, such as finger-tapping angle, angular velocity, cycle duration, coefficient of variation (CV), and slope, among PSP, PD patients, and healthy controls (HC). We further assess the diagnostic utility of finger-tapping parameters for differentiating PSP from PD, as well as their association with clinical features in PSP patients. Furthermore, using T1-weighted magnetic resonance imaging (MRI), we compared the volumes of various brain regions between PSP patients and HCs. To elucidate the mechanism of finger-tapping impairments in PSP, we assessed the correlations between brain volumes and kinematic parameters. This study aims to provide novel insights into PSP-specific motor dysfunction, develop a novel digital biomarker based on finger-tapping video, and uncover potential pathophysiological mechanisms underlying bradykinesia in PSP.

Results

Demographic features and clinical characteristics

The demographic features, motor and non-motor symptoms evaluation of all participants are listed in Table 1. Thirty-one PSP patients (18 males/13 females; median age=66 years; 31right-handed), thirty-one PD patients (15 males/16 females; median age=65 years; 31right-handed), and thirty HCs (19 males/11 females; median age=66 years; 29 right-handed/1 left-handed) were recruited. Disease duration was similar between PSP and PD (3.0 (2.0, 5.5) vs 5.0(2.0, 7.5); P = 0.071). UPDRS motor scores, MMSE, MoCA and PDQ-39 scores (assessed in the “OFF” condition) were closely between PSP and PD patients.

Table 1 Demographic feature, motor and non-motor symptoms evaluation
Full size table

Finger-tapping kinematics differ among PSP, PD patients and HCs

First, to eliminate potential bias related to hand dominance, we analyzed bilateral finger-tapping data from all participants. Although finger-tapping performance of the dominant hand was superior to that of the non-dominant hand, no significant differences were observed (Supplemental Tables 1–3). Thus, kinematic parameters of the dominant hand were used for all subsequent analyses. No significant differences in finger-tapping kinematics were observed across different PSP subtypes (Supplemental Fig. 1)

Results showed that the average finger-tapping angle of the PSP group (19.70 ± 8.89°) was significantly smaller than that of the PD group (32.84 ± 10.28°) and the HC group (46.09 ± 11.53°; Fig. 1A). Similar findings were observed for the maximum finger-tapping angle across all tapping cycles (Fig. 1C). The average finger-tapping angular velocity of the PSP group (78.78 ± 34.48°/s) was significantly lower than that of the HC group (158.22 ± 35.02°/s; Fig. 1B) and modestly lower than that of the PD group (104.92 ± 32.30°/s).

Among time-related parameters, the PSP group exhibited significantly shorter average finger-tapping cycle durations (0.25(0.22,0.29) s) compared with the PD (0.31(0.26,0.41) s) and HC groups (0.29(0.25, 0.34) s), while no significant difference was observed between PD and HC groups (P = 0.13; Fig. 1D). Compared with PD patients (2.95 ± 1.03 Hz), PSP patients (3.67 ± 0.94 Hz) exhibited a significantly higher average finger-tapping frequency, completing more finger-tapping cycles per unit time. However, no significant differences were observed between the PSP and HC groups (3.37 ± 0.84 Hz; P = 0.531) or PD and HC groups (P = 0.195; Fig. 1E).

CVs for finger-tapping angle, cycle duration and finger-tapping angular velocity were similar between PSP and PD groups, but were significantly higher in both groups compared to HCs (Table 2; Fig. 1G–I).

Fig. 1: Finger-tapping kinematic parameters among PSP, PD and HCs.
Fig. 1: Finger-tapping kinematic parameters among PSP, PD and HCs.
Full size image

A Difference of average finger-tapping angle among PSP, PD and HC; B difference of average finger-tapping angular velocity among PSP, PD and HC; C difference of maximum finger-tapping angle among PSP, PD and HC; D difference of finger-tapping frequency among PSP, PD and HC; E difference of average finger-tapping cycle duration among PSP, PD and HC; F difference of finger-tapping angles-cycle numbers slope among PSP, PD and HC; G–I difference of coefficient of variation (CV) of finger-tapping angle, angular velocity and cycle duration among PSP, PD and HC.

Table 2 Analysis of finger-tapping kinematic parameters among patients with PSP, PD and HC
Full size table

The slope of the finger-tapping angle was significantly steeper in PD patients (−0.18(−0.46, −0.36) °/cycle) compared with PSP patients (−0.06 (−0.23,0.05) °/cycle), while no significant difference was observed between the PSP and HC groups (−0.04 (−0.22,0.05) °/cycle); P = 0.584; Fig. 1F). This finding indicates that PSP patients do not exhibit a progressive reduction in finger-tapping amplitude during continuous finger-tapping, a hallmark feature of PD patients. The slopes of angular velocity and cycle duration did not differ significantly among PSP, PD, and HC groups (Table 2).

To evaluate the diagnostic performance of finger-tapping kinematics for PSP-PD differentiation, receiver operating characteristic (ROC) curve analysis was performed for all parameters with significant between-group differences (Fig. 2). Results showed that both the average and maximum finger-tapping angle had excellent diagnostic ability to distinguish PSP from PD (both AUROC = 0.83, P < 0.001, Table 3). Similar diagnostic performance was observed in ROC curve analysis for differentiating PSP-P from PD (AUROC = 0.81, P < 0.001; Supplemental Table 4). Due to the high collinearity among parameters, a combined diagnostic model could not be constructed.

Table 3 Diagnostic performance of finger-tapping kinematic parameters for PSP
Full size table

Finger-tapping kinematic parameters correlate with motor function scale scores in PSP patients

To further investigate the relationship between finger-tapping kinematic parameters and clinical features in PSP patients, we selected parameters with significant differences between PSP and PD for correlation analysis. Results showed that worse of finger-tapping kinematic parameters in PSP patients were associated with more severe motor disability and poorer quality of life.

In PSP patients, the average finger-tapping angles were significantly negatively correlated with total UPDRS scores (Pearson’s r = −0.575), UPDRS II scores (Pearson’s r = −0.531), UPDRS III scores (Spearman’s r = −0.644), PSPRS scores (Pearson’s r = −0.608), NMSS scores (Pearson’s r = −0.420), and PDQ-39 scores (Pearson’s r = −0.602) (Fig. 3A–F). However, no significant associations were found with age, disease duration, H-Y stage, UPDRS I scores, FOG-Q scores, MMSE scores, or MoCA scores. Finger-tapping angular velocities showed similar correlation patterns (Fig. 3G–K). The maximum finger-tapping angles had significant negative correlations with both UPDRS III scores (Pearson’s r = −0.448; P = 0.032) and PDQ-39 scores (Spearman’s r = −0.530, P = 0.009) (Supplemental Fig. 2A, B).

Fig. 2
Fig. 2
Full size image

ROC curve of average finger-tapping angle, average cycle duration, average finger-tapping angular velocity, maximum finger-tapping angle, finger-tapping frequency and angle-cycle number slope to differentiate PSP and PD.

Among time-related parameters, the slope of finger-tapping angle was independently negatively correlated with MoCA scores (Spearman’s r = -0.504) (Fig. 3L) and was also correlated with delayed response function scores (Spearman’s r = -0.465, P = 0.025) (Supplemental Fig. 2C). The finger-tapping frequency and average finger-tapping cycle durations exhibited no significant correlations with any of the aforementioned clinical features.

To identify the specific clinical symptoms associated with finger-tapping parameters, we performed additional correlation analyses between these finger-tapping kinematic parameters and individual items of the UPDRS III and PSPRS. Beyond right-hand finger-tapping test subscores, average finger-tapping angles, average angular velocities, and maximum finger-tapping angles were significantly correlated with balance-related subscores (UPDRS III Items 9, 12, 13; PSPRS Items 25–28; Fig. 3M, N).

Fig. 3: Association between finger-tapping kinematic parameters and clinical feature.
Fig. 3: Association between finger-tapping kinematic parameters and clinical feature.
Full size image

A–L Correlation analyses between average finger-tapping angle, angular velocity, slope of angle and PSPRS, UPDRS, PDQ-39, MoCA and NMSS scores. M Spearman correlation analysis between average finger-tapping angle, angular velocity, maximum finger-tapping angle and all items in UPDRS-III. N Spearman correlation analysis between average finger-tapping angle, angular velocity, maximum finger-tapping angle and all items in PSPRS.

Finger-tapping kinematic parameters correlate with regional brain volumes in PSP patients

Relative to HCs, PSP patients exhibited significant volumetric atrophy in 27 brain regions, including the left transverse temporal gyrus, left putamen, bilateral basal ganglia, bilateral nucleus accumbens (NAc), bilateral inferior parietal lobules, brainstem, and bilateral cerebellum (Supplemental Fig. 3).

Subsequent correlation analysis revealed significant associations between finger-tapping kinematic parameters and volumes of multiple brain regions in PSP patients. Volumes of the bilateral superior temporal gyrus (STG) and bilateral NAc were significantly positively correlated with the average finger-tapping angle (Fig. 4A–D). Volumes of bilateral STG, bilateral NAc, left superior lobe and left putamen were positively correlated with average angular velocity (Fig. 4E–H). Bilateral amygdala, right NAc, left STG and right putamen volumes were positively associated with maximum finger-tapping angle (Fig. 4I–L). Notably, the slope of finger-tapping angle exhibited significant negative correlations with right cerebellar cortical volume (Spearman’s r = −0.671), right cerebellar white matter volume (Spearman’s r = −0.720), and brainstem volume (Spearman’s r = −0.622) (Fig. 4M–O). No significant correlations were observed between cortical or subcortical brain volumes and average finger-tapping cycle duration, finger-tapping frequency, or CVs of finger-tapping angle, angular velocity, and cycle duration (Fig. 4P).

Fig. 4: Association between finger-tapping kinematic parameters and brain volumes.
Fig. 4: Association between finger-tapping kinematic parameters and brain volumes.
Full size image

A–H Correlation analyses between average finger-tapping angle, angular velocity and bilateral superior temporal gyri, nucleus accumbens volumes. I–L Correlation analyses between maximum finger-tapping angle and bilateral amygdala, right nucleus accumbens, left superior temporal gyri and right putamen volumes. M–O Correlation analyses between slope of finger tapping angle and right cerebellar cortical volume, right cerebellar white matter volume, and brainstem volume. P Spearman correlation analyses between finger-tapping kinematic parameters and atrophied regional brain volumes in progressive supranuclear palsy.

Discussion

The present study quantitatively characterized finger-tapping kinematic parameters in PSP patients, PD patients, and HCs using a video-based motion capture approach. The core finding is that, relative to PD patients and HCs, PSP patients exhibit reduced finger-tapping amplitude and velocity, increased motor irregularity, and an absence of the sequence effect during repetitive finger-tapping. Finger-tapping amplitude was identified as the most effective parameter for differentiating PSP from PD. Additionally, finger-tapping kinematic parameters were closely correlated with reduced volumes of key brain regions (cerebellum, brainstem, NAc, and STG) and the severity of motor dysfunction in PSP patients. These findings support the utility of video-based finger-tapping kinematic analysis as a practical, non-invasive, and objective digital biomarker for the clinical diagnosis and motor function assessment of PSP.

We confirmed previous findings that PSP patients lack the sequence effect during repetitive finger-tapping, using a simple approach (video-based motion analysis) and a larger sample size10,11,23. Our results showed that the slope of angle was -0.04°/cycle in HCs and -0.06°/cycle in PSP patients. This slight reduction was primarily attributed to physiological fatigue. In contrast, the slope of angle was -0.18°/cycle in PD patients, which is indicative of a primary pathological process. Djuric-Jovicic et al. reported that the finger-tapping amplitude and velocity were comparable between PSP and PD patients10. In contrast, our findings showed that both finger-tapping amplitude and velocity were significantly reduced in the PSP group relative to PD and HC groups, which is consistent with the findings of Ling et al.11. Notably, while the absence of sequence effect represents a distinctive finding of PSP patients, finger-tapping amplitude emerged as the most robust parameter for differentiating both PSP and PSP-P from PD. These results suggest that bradykinesia is more severe in PSP, consistent with the core clinical criterion of akinesia (a more severe motor impairment than the hypokinesia that is common in PD) according to the 2017 MDS PSP diagnostic criteria3. Overall, compared with PD, the bradykinesia of PSP patients is more severe, with reduced movement amplitude, slower velocity, and absence of sequence effect. These characteristics are consistent across the PSP-RS, PSP-P, and PSP-PGF subtypes, indicating that the bradykinesia of PSP is related to its common pathological changes.

Ling et al. further analyzed the correlations between finger tapping kinematics and UPDRS scores, H-Y stage, age and disease duration but reported no significant associations11. This is likely due to their small sample size (only 9 PSP patients). The lack of a correlation between H-Y stage and finger-tapping parameters may be attributed to the fact that most PSP patients were classified as H-Y stage 3–4 due to the early-onset postural instability. Our results showed significant associations between finger-tapping amplitude, velocity and clinical markers of disease severity such as PSPRS and UPDRS scores. A significant relationship was also identified between finger-tapping kinematic parameters and another core clinical feature of PSP--postural instability, which may stem from overlapping neural substrates (e.g., the STG) that modulate both balance and finger movements24. Patients with smaller finger-tapping amplitudes and slower velocity exhibited more severe motor symptoms, suggesting that video-based motion analysis has potential to serve as a marker for the severity of the disease. Given that the PSPRS and UPDRS scales require specialized training, are time-consuming to administer, and poorly capture subtle clinical changes21, automated video analysis may serve as a novel tool for disease assessment and follow-up in busy outpatient clinics25,26. Our standardized video recording protocol also provides a practical reference for future research.

Bradykinesia in PSP may be underpinned by distinct neural network alterations relative to PD. Levodopa improves finger-tapping amplitude and velocity in PD patients but has no significant effect on the sequence effect, indicating pathological alterations beyond the dopaminergic pathway27. Studies proved that the cerebellum is involved in the regulation of movement timing28. PD patients exhibit compensatory cerebellar volume increases, whereas PSP patients show prominent cerebellar and brainstem atrophy29,30. In PD patients, the sequence effect is significantly correlated with abnormal activation of the cerebellum and changes in cerebellar network connections31,32. We also observed that a subset of PSP patients exhibited a mild sequence effect, and these patients had larger cerebellar and brainstem volumes relative to other PSP patients. These findings suggest that the sequence effect is related to the cerebellar and brainstem-related motor circuits. Impairments in finger-tapping amplitude and velocity in PD are associated with dysfunction in multiple brain regions, including the primary motor cortex, basal ganglia, supplementary motor area, and primary sensory cortex8. The correlation analysis showed that PSP patients had positive correlations between finger movement amplitude, velocity and the volumes of the amygdala, STG, and NAc. Prior study has confirmed that the NAc modulates the electrical activity of the sensorimotor cortex to affect finger movements33. Thus, alterations in finger-tapping amplitude and velocity in PSP patients may involve impairments in sensorimotor integration and dysfunction in associated sensorimotor cortical networks34. This mechanism has rarely been reported in PSP, highlighting the need for further brain network research.

This study has several limitations, including a relatively small sample size, single-center recruitment, and a cross-sectional design. We were unable to assess the diagnostic efficacy of combining multiple bradykinesia-related features for distinguishing PSP from PD. Future multicenter, longitudinal studies with larger cohorts that include more PSP subtypes are needed to validate our findings and characterize dynamic changes in finger-tapping movements in PSP patients. Furthermore, standardized finger-tapping paradigms should be developed, and head-to-head comparative studies between wearable device-based and video-based kinematic analysis are warranted. This study only performed a basic correlational analysis between regional brain volumes (derived from T1-weighted MRI) and finger-tapping parameters in PSP patients. Further fMRI studies could provide deeper insights into neural mechanisms of bradykinesia in PSP.

In conclusion, this study identifies a unique finger-tapping profile in PSP patients, characterized by reduced amplitude and velocity, shorter cycle duration, and absence of sequence effect. Reduced finger-tapping amplitude effectively distinguishes PSP from PD. Finger-tapping kinematic parameters correlate with motor and balance function and are associated with atrophy of the NAc, STG, cerebellum, and brainstem. Video-based finger-tapping kinematic analysis thus represents a practical, non-invasive digital biomarker with significant clinical potential for the diagnosis, severity assessment, and longitudinal follow-up of PSP patients.

Methods

Participants

A total of 31 PSP (16 PSP-RS, 9 PSP-P and 6 PSP-PGF) patients and 31 PD patients matched for age and sex were recruited from the Department of Neurology, Qilu Hospital of Shandong University in this study. All PSP patients fulfilled the 2017 International Parkinson and Movement Disorder Society (MDS) PSP diagnostic criteria, and all PD patients fulfilled the 2015 MDS diagnostic criteria for PD. Patients with severe tremor or dystonia that affected finger-tapping movements were excluded. Age- and sex- matched HCs were recruited from the patients’ family or volunteers. All HCs had no history of Parkinson’s or other neurological disorders. Common exclusion criteria included severe medical comorbidities, severe cognitive impairment precluding performance of the finger-tapping, and other diseases affecting finger movements. All patients discontinued the medication for more than 12 hours and were tested during “OFF” condition.

The Unified Parkinson′s Disease Rating Scale (UPDRS), Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Parkinson’s Disease Quality of Life Scale (PDQ-39) and Non-motor Symptom Scale (NMSS) were performed in all patients. The Progressive Supranuclear Palsy Rating Scale (PSPRS) and Freezing Gait Questionnaire (FOG-Q) were performed in PSP patients. Among all participants, 15/31 PSP patients and 28/30 HCs underwent cranial MRI T1-weighted imaging.

This study was conducted with the written informed consent of all participants and approved by the Ethics Committee of Qilu Hospital of Shandong University (KLYY-202311(XZ)-010-2).

Finger-tapping video

A camera, placed on a tripod, was used to record standard video (1920 × 1080 pixel resolution, 30 frames per second) in a bright environment21. All participants were instructed to sit facing the camera and repeatedly tap their index finger to thumb as rapidly and widely as possible for at least 12 seconds starting upon the “begin” cue and ending at the “stop” signal. When recording the video, the fully exposed palm should face the camera, with the thumb and index finger parallel to the video plane, and the third, fourth and fifth digits naturally extended. The left and right hands were recorded separately.

Kinematic parameters

All finger-tapping videos were analyzed using Python 3.8.20, with key libraries including OpenCV 4.11.0.86 (for drawing hand ROIs), MediaPipe 0.10.11 (for detecting hand key points), NumPy 1.24.4 (for coordinate acquisition and calculation), and SciPy 1.10.1 (for identifying the finger-tapping cycle)22.

To minimize the influence of confounding factors such as arm elevation, hand repositioning, or adjustments of the hand position, only the 2-12 second segment of each finger-tapping video was analyzed. This approach ensured that the evaluated movement represented stable, continuous finger tapping. Firstly, a region of interest (ROI) of hand was manually delineated. Subsequently, keypoints corresponding to the wrist, index fingertip, and thumb tip were automatically identified through Media Pipe. A two-dimensional cartesian coordinate system was established using the wrist key point as the origin. The positive y-axis direction was defined as vertically ascending, and the positive x-axis direction was defined as horizontally towards the right (picture perspective). The coordinates of these three hand key points were acquired within this system and then underwent normalization to mitigate the influence of variations in video scaling (Fig. 5A).

A finger-tapping cycle was defined as the interval between two consecutive instances corresponding to the minimum distance separating the thumb tip and index finger tip points (Fig. 5B). The finger-tapping angle (°), angular velocity(°/s), cycle duration(s) of each cycle were calculated. For each participant, the average, maximum, and coefficient of variation (CV) were calculated for each kinematic parameter across all identified cycles. Additionally, finger-tapping frequency (Hz) was computed (Fig. 5C).

Furthermore, to assess the sequence effect of finger-tapping, the slope of the linear regression line between finger-tapping angles, angular velocities, cycle durations versus cycle numbers were calculated for each participant.

Fig. 5: Identification of finger key points, finger-tapping cycles, and calculation of kinematic parameters.
Fig. 5: Identification of finger key points, finger-tapping cycles, and calculation of kinematic parameters.
Full size image

A Key point recognition of the hand. B, C Detection of finger-tapping cycles and calculation of finger-tapping kinematic parameters.

Magnetic resonance image acquisition

MRI was performed with an eight-channel head coil and a Siemens verio 3.0 T MR scanner. Minimizing head movement was achieved by using tight and snug foam cushioning. All of the participants received earplugs to help with scanner noise reduction. The 3D T1-weighted images were obtained using a magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence (repetition time/echo time = 2000/2.3 ms; inversion time = 900 ms; flip angle = 9°; matrix = 256 × 256; slice thickness = 1 mm, no gap; 192 slices).

Brain volume measurements

Freesurfer version 6.0 (https://surfer.nmr.mgh.harvard.edu/) was utilized to parcellate the whole brain into 95 distinct anatomical regions, and then calculate the volume of each region in each subject. All procedures were inspected by experienced radiologists to confirm that each region was correctly parcellated.

Statistical analysis

Data analysis was performed using SPSS version 27.0 (IBM Crop., Armonk, NY). Categorical variables are presented as frequencies and percentages. Continuous variables are presented as mean and standard deviation (SD) if normally distributed, or as median and interquartile range (IQR) if non-normally distributed. The Chi-square test was employed to compare sex differences. One-way analysis of variance (ANOVA) or the Kruskal-Wallis test was used to compare continuous variables across the three groups, depending on the normality of the data distribution. Post-hoc comparisons were performed using the Bonferroni correction method. Differences in brain volumetric measures between PSP and HC were compared using the independent Student’s t-test or the Mann-Whitney U test, according to data normality. Spearman’s and Pearson’s tests were conducted to investigate the associations between finger tapping kinematic parameters and age, disease duration, clinical scale scores, and brain volumes, based on the normality of the data distribution. And the FDR correction was applied to avoid false positive. Diagnostic accuracy was evaluated using the area under the receiver operating characteristic curve (AUC). Results were considered significant at a level of P < 0.05.