iPatax: a Tablet-based Tool for Quantitative Assessment of Cerebellar Ataxia Authors: Takahiro Nagai, Shingo Koide, Tomohiko Ishihara, Masayoshi Tada, Masatoyo Nishizawa, Osamu Onodera Published: 2026 ABSTRACT AbstractThe clinical severity of cerebellar ataxia is still assessed primarily using ordinal rating scales, and continuous quantitative measures are still limited. In this context, we have developed iPatax, an iPad®-based application, as a quantitative and objective tool for assessment of cerebellar motor dysfunction. Finger movements were recorded in 27 controls and 28 patients with cerebellar ataxia during three target-following tasks: non-continuous linear, non-continuous circular, and continuous circular movements. In the controls, the coefficient of variation of velocity (CVV) during non-continuous linear movements and the average distance from the target trajectory (AD) during non-continuous circular movement both improved over repeated trials, consistent with learning-related changes, and were prioritized as candidate metrics. Both metrics differentiated patients with cerebellar ataxia from controls (AUCs = 0.853 and 0.918, respectively). In the patients, CVV during non-continuous linear movement correlated significantly with both the total Scale for the Assessment and Rating of Ataxia (SARA) score and the upper-limb subscore (total SARA: rS = 0.50; upper-limb subscore: rS = 0.51), whereas AD during non-continuous circular movements correlated with the total SARA score but not with the upper-limb subscore (total SARA: rS = 0.44; upper-limb subscore: rS = 0.063). These findings suggest that, as a widely available tablet device, iPatax can capture and quantify clinically relevant temporal and trajectory-related aspects of cerebellar motor dysfunction. This approach may supplement current ordinal rating scales in clinical and research settings.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12311-026-02028-9.Keywords: Cerebellar ataxia, iPatax, Upper-limb function, Quantitative assessment Introduction Cerebellar ataxia is a neurological manifestation caused by cerebellar dysfunction that impairs coordinated movement. In both clinical practice and for evaluation of therapeutic effects, assessment of its presence and severity is essential. In routine practice, severity is still commonly assessed using semiquantitative clinical rating scales such as the Scale for the Assessment and Rating of Ataxia (SARA) and the International Cooperative Ataxia Rating Scale (ICARS) [1, 2]. Although these scales are practical for daily use, they provide only limited objective and quantitative assessment of motor dysfunction in both clinical and research settings [3]. Because they are ordinal and partly dependent on examiner judgment, they may be insufficient for detection of subtle impairment, monitoring of longitudinal change, and evaluation of treatment effects, and may also be affected by ceiling and floor effects. Consequentially, there is still a substantial need for sensitive and reproducible bedside measures for objective quantitation of cerebellar motor dysfunction [4, 5]. To address these limitations, several quantitative functional tests for ataxia have been adopted, including the nine-hole peg test, timed up-and-go test, and 6-min walk test [6–9]. Although these methods provide quantitative data, they mainly reflect motor performance or functional disability, such as movement time or walking distance, and approaches that directly capture the physiological characteristics of cerebellar motor dysfunction remain limited. The cerebellum plays an essential role in motor learning and in the temporal and spatial control of movement [10, 11]. In healthy individuals, repeated movements usually become progressively more stable and accurate with practice, whereas cerebellar dysfunction may impair such stabilization [10]. Abnormalities in temporal control may be manifested as increases in timing variability [12], and abnormalities in spatial control as deviation from an intended trajectory [13]. Therefore, quantitative assessment of these features may provide a more physiologically relevant approach for grasping the degree of cerebellar dysfunction. Currently, simple and accessible methods for continuous assessment of these features, without the need for specialized equipment, are still lacking. To address this gap, we have developed iPatax, an iPad®-based application designed to record finger movements during target-following tasks. In this study, we evaluated multiple kinematic metrics derived from three task conditions to select candidate measures of cerebellar motor dysfunction and to examine their associations with group differences and clinical severity. Materials and Methods Study Participants This study included 27 healthy controls (12 males and 15 females; mean age, 49.2 ± 14.4 years) and 28 patients with cerebellar ataxia (13 males and 15 females; mean age, 53.7 ± 11.5 years), all aged between 30 and 80 years. All of the controls had a SARA score of 0–2. Diagnoses in the patient group included multiple system atrophy cerebellar type (MSA-C), cortical cerebellar atrophy (CCA), Machado-Joseph disease (MJD), spinocerebellar ataxia type 6 (SCA6), spinocerebellar ataxia type 7 (SCA7), spinocerebellar ataxia type 31 (SCA31), dentatorubral-pallidoluysian atrophy (DRPLA), autosomal dominant spinocerebellar ataxia with an unknown genetic mutation (AD-SCA), and paraneoplastic cerebellar degeneration (PCD). Clinical characteristics are summarized in Table 1. The study was approved by the Ethics Committee of Niigata University, Niigata, Japan (2015 − 1522). Written informed consent was obtained from all participants. Table 1Demographic and clinical characteristics of controls and patients with cerebellar ataxiaControlsPatients with cerebellar ataxian (female)27 (15)28 (15)Age, years (mean ± SD)49.2 ± 14.453.7 ± 11.5SARA total score (median [range])0 [0–2]11.5 [4–25]Upper-limb subscore (median [range])0 [0–0.5]4 [1.5–9]Diagnosis, n MSA-C2 CCA5 MJD9 SCA65 SCA71 SCA311 DRPLA1 AD-SCA3 PCD1SARA Scale for the Assessment and Rating of Ataxia, MSA-C multiple system atrophy, cerebellar type, CCA cortical cerebellar atrophy, MJD Machado-Joseph disease, SCA6 spinocerebellar ataxia type 6, SCA7 spinocerebellar ataxia type 7, SCA31 spinocerebellar ataxia type 31, DRPLA dentatorubral-pallidoluysian atrophy, AD-SCA autosomal dominant spinocerebellar ataxia with an unknown genetic mutation, PCD paraneoplastic cerebellar degeneration Demographic and clinical characteristics of controls and patients with cerebellar ataxia SARA Scale for the Assessment and Rating of Ataxia, MSA-C multiple system atrophy, cerebellar type, CCA cortical cerebellar atrophy, MJD Machado-Joseph disease, SCA6 spinocerebellar ataxia type 6, SCA7 spinocerebellar ataxia type 7, SCA31 spinocerebellar ataxia type 31, DRPLA dentatorubral-pallidoluysian atrophy, AD-SCA autosomal dominant spinocerebellar ataxia with an unknown genetic mutation, PCD paraneoplastic cerebellar degeneration iPatax System and Task Procedures iPatax is an iPad®-based application developed for evaluation of ataxia. The application was created using Apple’s integrated development environment Xcode and was run on a 4th-generation iPad® with iOS 10. Participants performed three tracking tasks using the dominant finger: continuous circular movement (C-C), non-continuous circular movement (NC-C), and non-continuous linear movement (NC-L) (Fig. 1A, Video 1). In all tasks, participants tracked a 1.4-cm-diameter target for 1 min, with a fixed motion cycle of 6 s. In the C-C task, the target moved continuously along a circular path with a diameter of 7.39 cm. In the NC-C task, the target moved along the same circular path, with the direction reversed after each lap. In the NC-L task, the target moved back and forth along a 13.79-cm straight line. Finger coordinates were recorded each time movement was detected on the iPad® touchscreen. Selection of learning-related iPatax parameters in controls. (A) Schematic illustration of the three iPatax tasks: non-continuous linear movement (NC-L), non-continuous circular movement (NC-C), and continuous circular movement (C-C). The inset illustrates the distance from the target trajectory during the tracking task. (B) Representative finger velocity trace during a 1-min NC-L task in a control participant. The first 3 s were excluded from the analysis. The remaining period was divided into P1 (3–20 s), P2 (20–40 s), and P3 (40–60 s). (C–E) Changes in the coefficient of variation of velocity (CVV) from P1 to P3 during NC-L (C), NC-C (D), and C-C (E) in controls. (F) Changes in average distance from the target trajectory (AD) from P1 to P3 during NC-C in controls. Points from the same participant are connected by lines, colored blue for decreases and red for increases from P1 to P3. P values were calculated using the Wilcoxon signed-rank test Data Acquisition and Metric Calculation For each 1-min trial, the first 3 s after task onset were excluded from the analysis because of high variability. The remaining interval was divided into three segments: P1 (3–20 s), P2 (20–40 s), and P3 (40–60 s). For each segment, two types of metric were calculated: (1) metrics reflecting temporal variability and (2) metrics reflecting spatial deviation from the target trajectory. For (1), finger velocity was calculated as the displacement between consecutive finger positions divided by the elapsed time. Acceleration was calculated as the change in velocity divided by the elapsed time. For each segment, the coefficient of variation of velocity (CVV) and the coefficient of variation of acceleration (CVA) were calculated by dividing the unbiased standard deviations by their respective means (CV = SD / mean). For (2), the distance from the finger coordinates to the target trajectory (line or circle) was calculated (Fig. 1A). For a line, this was defined as the perpendicular distance from the finger coordinates to the line. For a circle, this was defined as the absolute difference between the distance from the finger coordinates to the center of the circle and the radius. For each segment, the mean distance from the target trajectory was defined as the average distance (AD), and the coefficient of variation of this distance (CVD) was calculated by dividing the unbiased standard deviation by the mean. Pb was defined as the minimum value of each metric across P1–P3, representing the best performance achieved during the trial. Statistical Analysis Statistical analyses and graph generation were performed using R version 4.5.2. For clinical correlation analyses, the upper-limb subscore of the SARA was defined as the sum of the finger chase, nose-finger test, and fast alternating hand movement items. For paired comparisons, such as within-subject comparisons between repeated movements, the Wilcoxon signed-rank test was used. For the repeated-movement comparisons used in the metric-selection step, p values across the 12 candidate metrics were adjusted for multiple testing using the Benjamini–Hochberg procedure to control the false discovery rate, and the corresponding q values were reported. For unpaired comparisons between groups, the Wilcoxon rank-sum test was used. Spearman’s rank correlation coefficient was used for correlation analyses. Multiple regression analysis was used for age-adjusted analyses. Receiver operating characteristic (ROC) analysis was used to assess discrimination between the controls and the patients with cerebellar ataxia. Confidence intervals for the area under the curve (AUC) were estimated using the DeLong method. Results Selection of Learning-Related Ipatax Parameters We first examined changes in kinematic parameters across repeated movements in controls, reasoning that parameters relevant to cerebellar motor function would show learning-related improvement with repetition. Representative finger velocity traces during repeated tracking suggested progressive temporal stabilization over time (Fig. 1B). We therefore evaluated temporal variability using velocity-based measures. The coefficient of variation of velocity (CVV) decreased significantly from P1 to P3 in both non-continuous movements (NC-L CVV, p = 4.2 × 10− 5, q = 5.0 × 10− 4; NC-C CVV, p = 8.6 × 10− 5, q = 5.1 × 10− 4), whereas no significant change was observed in continuous circular movement (C-C CVV, p = 0.971) (Fig. 1C–E). By contrast, the coefficient of variation of acceleration showed less consistent changes across tasks (NC-L CVA, p = 0.0315, q = 0.0757) (Supplementary Fig. 1A–C), supporting CVV as an informative metric of learning-related change in temporal variability. We next evaluated spatial variability using the distance from the target trajectory. Among distance-based measures, average distance (AD) during non-continuous circular movement decreased significantly from P1 to P3 (NC-C AD, p = 0.00146, q = 0.00582), indicating improved spatial accuracy with repetition (Fig. 1F). By contrast, the other distance-based measures showed no significant improvement across repeated movements (Supplementary Fig. 1D–H). This pattern supported NC-C AD as an informative metric of learning-related change in spatial performance. The corresponding P1–P3 improvements in these metrics (NC-L CVV, NC-C CVV, and NC-C AD) were also observed in patients with cerebellar ataxia (Supplementary Fig. 2). Based on the findings in controls, we selected CVV during non-continuous movements and NC-C AD as candidate complementary metrics for subsequent analyses. Differentiation of Patients with Cerebellar Ataxia from Controls We then assessed the effect of age in controls to further refine the candidate metrics. Both NC-L CVV Pb and NC-C CVV Pb, defined as the minimum value across P1–P3, were significantly associated with age (NC-L CVV, rS = 0.49, p = 0.01; NC-C CVV, rS = 0.68, p = 8.5 × 10− 5), whereas NC-C AD Pb was not (NC-C AD, rS = 0.14, p = 0.48) (Fig. 2A, B, Supplementary Fig. 3A). Given its stronger age dependence, NC-C CVV was not prioritized for subsequent analysis. Accordingly, we focused on NC-L CVV and NC-C AD. Differentiation of patients with cerebellar ataxia from controls using complementary metrics. (A, B) Associations of age with NC-L CVV Pb (A) and NC-C AD Pb (B) in controls. Pb was defined as the minimum value across P1–P3. Correlations were assessed using Spearman’s rank correlation coefficient. (C, D) Comparisons of NC-L CVV Pb (C) and NC-C AD Pb (D) between controls (Ctrls) and patients with cerebellar ataxia. Group comparisons were performed using the Wilcoxon rank-sum test. Each data point represents an individual patient, colored by diagnosis. (E) Receiver operating characteristic (ROC) curves for NC-L CVV Pb and NC-C AD Pb for discriminating patients with cerebellar ataxia from controls. The area under the curve (AUC) is shown for each metric We next examined whether NC-L CVV Pb and NC-C AD Pb differed between controls and patients with cerebellar ataxia. NC-L CVV Pb and NC-C AD Pb were both significantly higher in patients than in controls (NC-L CVV, p = 1.9 × 10− 6; NC-C AD, p = 4.4 × 10− 9) (Fig. 2C, D). Multiple regression analysis including group and age as independent variables showed that NC-L CVV Pb remained significantly associated with group after adjustment for age (β = 0.163, 95% CI 0.091–0.234, p = 3.1 × 10− 5). Receiver operating characteristic (ROC) analysis further showed that both metrics discriminated patients with cerebellar ataxia from controls (NC-L CVV Pb, AUC = 0.853, 95% CI: 0.755–0.952; NC-C AD Pb, AUC = 0.918, 95% CI: 0.835–1.000) (Fig. 2E). Association of Complementary Metrics with Clinical Severity in Cerebellar Ataxia Finally, we examined whether the selected metrics were associated with clinical severity in patients with cerebellar ataxia. The diagnosis-colored plots did not suggest any clear diagnosis-specific tendency. NC-L CVV Pb showed a significant positive correlation with both the total SARA score and the upper-limb subscore (total SARA score, rS = 0.50, p = 0.007; upper-limb subscore, rS = 0.51, p = 0.0053) (Fig. 3A, B). By contrast, NC-C AD Pb showed a significant correlation with the total SARA score but not with the upper-limb subscore (total SARA score, rS = 0.44, p = 0.018; upper-limb subscore, rS = 0.063, p = 0.75) (Fig. 3C, D). NC-C CVV Pb was not significantly correlated with either measure of clinical severity (total SARA score, rS = 0.19, p = 0.34; upper-limb subscore, rS = 0.079, p = 0.69) (Supplementary Fig. 3B, C). Association of complementary metrics with clinical severity in patients with cerebellar ataxia. (A, B) Correlations of NC-L CVV Pb with the total SARA score (A) and the upper-limb subscore (B). (C, D) Correlations of NC-C AD Pb with the total SARA score (C) and the upper-limb subscore (D). Pb was defined as the minimum value across P1–P3. Correlations were assessed using Spearman’s rank correlation coefficient. Each data point represents an individual patient, colored by diagnosis Discussion In this study, we were able to show that iPatax provides an objective and quantitative assessment of upper-limb cerebellar ataxia based on finger movements recorded during a simple iPad®-based motor task. In selecting candidate metrics, we used a working hypothesis that parameters showing improvement with repetition in controls would be more likely to reflect cerebellar motor function, given the role of the cerebellum in motor learning and adaptation. On this basis, CVV during non-continuous linear movement and AD during non-continuous circular movement emerged as informative metrics, capturing repetition-related change in the controls and complementary aspects of cerebellar motor dysfunction in the patients. Non-continuous tasks were particularly useful for revealing different aspects of cerebellar ataxia. Compared with continuous movements, movements involving repeated reversals and reinitiation require ongoing adjustment of both timing and trajectory, which may make cerebellar motor irregularities more apparent [14, 15]. In this study, CVV during non-continuous linear movement and AD during non-continuous circular movement were complementary in capturing group differences and associations with clinical severity. Consistent with previous reports of impaired timing during discontinuous movement in patients with cerebellar dysfunction [15, 16], our findings support the possibility that CVV reflects impaired temporal control, as delayed reinitiation at reversal points may cause brief pauses followed by compensatory acceleration, thereby increasing the variability of velocity. By contrast, AD may reflect impaired trajectory control, because overshoot from the intended circular path and impaired movement smoothness would increase the average spatial deviation from the target trajectory. A clinical advantage of iPatax is that it enables rapid and accessible quantitative assessment of cerebellar ataxia on a widely available iPad® platform. Quantitative assessment of upper-limb ataxia has advanced substantially in recent years; dedicated measurement hardware can now yield digital upper-limb outcomes that reflect meaningful longitudinal change and treatment responses [17], alongside the application of other instrumented approaches [18–21]. In this context, iPatax can be implemented on a widely available tablet device in a seated position at the bedside or in the outpatient clinic. Among various touchscreen-based approaches, APP-Coo-Test can quantify upper-limb impairment as the time required to complete sequential dot-touch tasks [22], reflecting overall motor execution speed. iPatax, on the other hand, evaluates how movement unfolds during continuous tracking, having the potential to capture directly the physiological characteristics of cerebellar motor control. These features support the utility of iPatax as a practical quantitative tool for bedside and outpatient assessment of upper-limb cerebellar dysfunction, serving as an adjunct to existing clinical rating methods. Although examination-based scales such as SARA are widely used for assessment of cerebellar ataxia, their ordinal structure and reliance on examiner judgment may limit sensitivity in some settings, and gait- or stance-based items may also be influenced by non-cerebellar factors [23]. In this respect, iPatax offers a complementary approach by providing continuous quantitative measures of upper-limb motor function in a seated position. This study had several limitations. First, the present method is specific to upper-limb motor function and cannot be used to assess truncal ataxia or predominantly lower-limb ataxia. In addition, arm support during testing was not formally standardized, which should be addressed in future studies. Second, its ability to distinguish cerebellar ataxia from other movement disorders remains untested. Third, this was a relatively small single-center study, and the patient cohort was heterogeneous with relatively few severe cases, limiting the generalizability of the findings and the evaluation of disease-specific or severity-specific characteristics. The limited granularity of the ordinal upper-limb subscore in this cohort may also have limited the detectability of associations with that measure. Future studies using larger independent cohorts, including patients without ataxia, are needed to further evaluate this metric-selection approach, determine the test–retest reliability of this method, and assess its usefulness for monitoring disease progression, treatment response, and the effects of repeated task practice in patients with cerebellar ataxia. Conclusion iPatax may serve as a practical and objective complementary tool for quantitative assessment of upper-limb cerebellar ataxia. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material Video 1: Three types of movement test (NC-L, NC-C, C-C) were implemented as an iPad® application. Each test lasts 1 min, but only the initial portion is shown in the video. Supplementary Material 2 Acknowledgements We would like to thank Jun Tokunaga for collecting data from the study subjects. Author Contributions T.N. developed and tested the software. T.N. and S.K. acquired the data. S.K. analyzed the data and drafted the manuscript. T.I. contributed to study organization and data interpretation. M.T. conceived and designed the study and contributed to data interpretation. M.N. contributed to study supervision. O.O. conceived the study and supervised the project. All authors reviewed and approved the final manuscript. Funding Osamu Onodera was funded by Grants-in-Aid from the Research Committee of Ataxia, Health Labour Sciences Research Grant, The Ministry of Health, Labour and Welfare, Japan [grant numbers JPMH20FC1041, JPMH23FC1010 and JPMH26FC1012]. Data Availability All data and analysis codes are available in the Zenodo repository under DOI: 10.5281/zenodo.19626282. Declarations Ethics Approval This study was approved by the Ethics Committee of Niigata University, Niigata, Japan (2015 − 1522). We explained the study to the subjects using an explanatory document approved by the Ethics Committee, and if they agreed, we asked them to sign the consent document. The signed consent documents were kept secure. We confirm that we have read the journal’s position on issues involved in ethical publication and affirm that this work is consistent with those guidelines. Competing interests Osamu Onodera received speaking honoraria from Kyowa Hakko Kirin Co., Ltd., Bristol-Myers Squibb, Ono Pharmaceutical Co., Ltd., Mitsubishi Tanabe Pharm, Takeda, Daiichi-Sankyo, FUJIFILM, SANOFI, and FP-pharm. Osamu Onodera also received consulting fees from Kissei Co., Ltd and Dainihon Sumitomo Co., Ltd.Takahiro Nagai holds stock in Apple Inc., the manufacturer of the iPad® used in this study. The other authors have no additional disclosures to declare. Takahiro Nagai holds stock in Apple Inc., the manufacturer of the iPad® used in this study.The other authors have no additional disclosures to declare. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Takahiro Nagai and Shingo Koide contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material Video 1: Three types of movement test (NC-L, NC-C, C-C) were implemented as an iPad® application. Each test lasts 1 min, but only the initial portion is shown in the video. Supplementary Material 2 Data Availability Statement All data and analysis codes are available in the Zenodo repository under DOI: 10.5281/zenodo.19626282.