Abstract
Parkinson’s disease (PD) exhibits marked clinical heterogeneity, which poses challenges for diagnosis, prognosis, and therapeutic precision, especially for early-stage PD patients. Existing subtyping approaches often rely on subjective clinical scales and single-modality data, which limits their sensitivity in capturing subtle but clinically relevant differences across patients. To reveal clinically meaningful PD subtypes, we propose a data-driven multimodal framework that integrates resting-state electroencephalography (EEG) and dual-task gait features using mutual cross-attention (MCA) fusion. Forty idiopathic early-stage PD patients were enrolled in a prospective study. EEG biomarkers were encoded via a convolutional neural network for the prediction of motor severity (MDS-UPDRS-III), while dual-task gait features were derived to capture subtle motor dysfunctions. The MCA enabled bidirectional attention-guided integration of EEG and gait features, which were then clustered using an unsupervised method. The analysis revealed three distinct subtypes, with dual-task-based fusion providing superior clinical separation. Subtype I was characterized by pronounced motor deficits; Subtype II showed moderate symptoms with relatively preserved quality of life; and Subtype III presented mild motor impairments but exhibited poorer cognitive and psychosocial outcomes. Feature contribution analyses highlighted central beta and theta EEG activity, along with dual-task gait metrics (e.g., stride length during turning), as key drivers of subtype differentiation. Longitudinal follow-up demonstrated subtype-specific rehabilitation responses, with Subtype II showing an insufficient response compared to other subtypes. In conclusion, this study enables digital phenotyping of PD with prognostic implications for personalized rehabilitation strategies and accelerates precision medicine.
Similar content being viewed by others
Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. The underlying code for this study is available at https://github.com/deyuwang126/mutual_crossattention_PDclustering.
References
Bloem, B. R., Okun, M. S. & Klein, C. Parkinson’s disease. Lancet 397, 2284–2303 (2021).
Farrow, S. L., Cooper, A. A. & O’Sullivan, J. M. Redefining the hypotheses driving Parkinson’s diseases research. npj Parkinson’s. Dis. 8, 45 (2022).
Deng, X. et al. Parkinson’s disease subtypes: approaches and clinical implications. Parkinsonism Relat. Disord. 130, 107208 (2025).
Stebbins, G. T. et al. How to identify tremor-dominant and postural instability/gait difficulty groups with the Movement Disorder Society Unified Parkinson’s disease rating scale: comparison with the Unified Parkinson’s disease rating scale. Mov. Disord. 28, 668–670 (2013).
Eisinger, R. S. et al. Motor subtype changes in early Parkinson’s disease. Parkinsonism Relat. Disord. 43, 67–72 (2017).
von Coelln, R. et al. The inconsistency and instability of Parkinson’s disease motor subtypes. Parkinsonism Relat. Disord. 88, 13–18 (2021).
Erro, R. et al. Comparing postural instability and gait disorder, and akinetic-rigid subtyping of Parkinson disease and their stability over time. Eur. J. Neurol. 26, 1212–1218 (2019).
Fereshtehnejad, S.-M., Zeighami, Y., Dagher, A. & Postuma, R. B. Clinical criteria for subtyping Parkinson’s disease: biomarkers and longitudinal progression. Brain 140, 1959–1976 (2017).
Endo, M. et al. Data-driven discovery of movement-linked heterogeneity in neurodegenerative diseases. Nat. Mach. Intell. 6, 1034–1045 (2024).
Yassine, S. et al. Identification of Parkinson’s disease subtypes from resting-state electroencephalography. Mov. Disord. 38, 1451–1460 (2023).
Brendel, M., Su, C., Hou, Y., Henchcliffe, C. & Wang, F. Comprehensive subtyping of Parkinson’s disease patients with similarity fusion: a case study with BioFIND data. npj Parkinson’s. Dis. 7, 83 (2021).
Inguanzo, A. et al. MRI subtypes in Parkinson’s disease across diverse populations and clustering approaches. npj Parkinson’s. Dis. 10, 159 (2024).
Dadu, A. et al. Identification and prediction of Parkinson’s disease subtypes and progression using machine learning in two cohorts. npj Parkinson’s. Dis. 8, 172 (2022).
Yang, Y. et al. Validation of a spatiotemporal gait model using inertial measurement units for early-stage parkinson’s disease detection during turns. IEEE Trans. Biomed. Eng. 69, 3591–3600 (2022).
Meng, L. et al. Inertial-based gait metrics during turning improve the detection of early-stage Parkinson’s disease patients. IEEE Trans. Neural Syst. Rehabil. Eng. 31, 1472–1482 (2023).
Russo, M. et al. A cluster analysis for Parkinson’s disease phenotyping with gait parameters. In Proc. IEEE International Conference on Metrology for eXtended Reality, Artificial Intelligence and Neural Engineering (MetroXRAINE) 882–887 (IEEE, 2023).
Inguanzo, A. et al. Hierarchical cluster analysis of multimodal imaging data identifies brain atrophy and cognitive patterns in Parkinson’s disease. Parkinsonism Relat. Disord. 82, 16–23 (2021).
Lawton, M. et al. Developing and validating Parkinson’s disease subtypes and their motor and cognitive progression. J. Neurol. Neurosurg. Psychiatry 89, 1279 (2018).
Wang, B. et al. Similarity network fusion for aggregating data types on a genomic scale. Nat. Methods 11, 333–337 (2014).
Zhang, J. et al. Multi-modal cross-attention network for Alzheimer’s disease diagnosis with multi-modality data. Comput. Biol. Med. 162, 107050 (2023).
Golovanevsky, M., Eickhoff, C. & Singh, R. Multimodal attention-based deep learning for Alzheimer’s disease diagnosis. J. Am. Med. Inform. Assoc. 29, 2014–2022 (2022).
Zhao, Y. & Gu, J. Feature fusion based on mutual-cross-attention mechanism for EEG emotion recognition. In Proc. Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 (eds Marius, G.L. et al.) 276-285 (Springer Nature Switzerland, 2024).
Zhang, R., Sheng, J., Zhang, Q., Wang, J. & Wang, B. A review of multimodal fusion–based deep learning for Alzheimer’s disease. Neuroscience 576, 80–95 (2025).
Tang, C., Xi, M., Sun, J., Wang, S. & Zhang, Y. MACFNet: detection of Alzheimer’s disease via multiscale attention and cross-enhancement fusion network. Comput. Methods Prog. Biomed. 254, 108259 (2024).
Renee, M. & Hendricks, M. T. K. A systematic review of parkinson’s disease cluster analysis research. Aging Dis. 12, 1567–1586 (2021).
Schalkamp, A.-K., Rahman, N., Monzón-Sandoval, J. & Sandor, C. Deep phenotyping for precision medicine in Parkinson’s disease. Dis. Models Mechan. 15, dmm049376 (2022).
Vaswani, A. et al. Attention is all you need. In Proc. 31st International Conference on Neural Information Processing Systems 6000–6010 (Curran Associates Inc., 2017).
Yan, F., Guo, Z., Iliyasu, A. M. & Hirota, K. Multi-branch convolutional neural network with cross-attention mechanism for emotion recognition. Sci. Rep. 15, 3976 (2025).
Kim, D. H. et al. Characterization of idiopathic Parkinson’s disease subgroups using quantitative gait analysis and corresponding subregional striatal uptake visualized using 18F-FP-CIT positron emission tomography. Gait Posture 82, 167–173 (2020).
Mirelman, A. et al. Gait impairments in Parkinson’s disease. Lancet Neurol. 18, 697–708 (2019).
Belvisi, D. et al. The pathophysiological correlates of Parkinson’s disease clinical subtypes. Mov. Disord. 36, 370–379 (2021).
Fereshtehnejad, S.-M. et al. New clinical subtypes of Parkinson disease and their longitudinal progression: a prospective cohort comparison with other phenotypes. JAMA Neurol. 72, 863–873 (2015).
Liu, J. et al. Common and specific effects in brain oscillations and motor symptoms of tDCS and tACS in Parkinson’s disease. Cell Rep. Med. 6, 102044 (2025).
Radcliffe, E. M. et al. Oscillatory beta dynamics inform biomarker-driven treatment optimization for Parkinson’s disease. J. Neurophysiol. 129, 1492–1504 (2023).
Wang, Q., Meng, L., Pang, J., Zhu, X. & Ming, D. Characterization of EEG data revealing relationships with cognitive and motor symptoms in Parkinson’s disease: a systematic review. Front. Aging Neurosci. 10, 587396 (2020).
Meng, L. et al. The inertial-based gait normalcy index of dual task cost during turning quantifies gait automaticity improvement in early-stage Parkinson’s rehabilitation. J. Neuroeng. Rehabil. 21, 166 (2024).
Meng, L. et al. Inertial-based dual-task gait normalcy index at turns: a potential novel gait biomarker for early-stage Parkinson’s disease. IEEE Trans. Neural Syst. Rehabil. Eng. 33, 687–695 (2025).
Fasano, A., Canning, C. G., Hausdorff, J. M., Lord, S. & Rochester, L. Falls in Parkinson’s disease: a complex and evolving picture. Mov. Disord. 32, 1524–1536 (2017).
Disbrow, E. A. et al. Cortical oscillatory dysfunction in Parkinson disease during movement activation and inhibition. PLoS ONE 17, e0257711 (2022).
Sosnik, R., Fahoum, F., Katzir, Z., Mirelman, A. & Maidan, I. Key shifts in frontoparietal network activity in Parkinson’s disease. npj Parkinson’s. Dis. 11, 2 (2025).
Caviness, J. N. et al. Differential spectral quantitative electroencephalography patterns between control and Parkinson’s disease cohorts. Eur. J. Neurol. 23, 387–392 (2016).
Sarter, M., Avila, C., Kucinski, A. & Donovan, E. Make a left turn: cortico-striatal circuitry mediating the attentional control of complex movements. Mov. Disord. 36, 535–546 (2021).
Johansson, M. E., Toni, I., Kessels, R. P. C., Bloem, B. R. & Helmich, R. C. Clinical severity in Parkinson’s disease is determined by decline in cortical compensation. Brain 147, 871–886 (2024).
Mancini, M. et al. Digital gait biomarkers in Parkinson’s disease: susceptibility/risk, progression, response to exercise, and prognosis. npj Parkinson’s. Dis. 11, 51 (2025).
Zarkali, A., Thomas, G. E. C., Zetterberg, H. & Weil, R. S. Neuroimaging and fluid biomarkers in Parkinson’s disease in an era of targeted interventions. Nat. Commun. 15, 5661 (2024).
Bai, X. et al. The association of motor reserve and clinical progression in Parkinson’s disease. NeuroImage Clin. 44, 103704 (2024).
Yoon, Y. J. et al. Occipital hypoperfusion and motor reserve in Parkinson’s disease: an early-phase 18F-FP-CIT PET study. npj Parkinson’s. Dis. 10, 221 (2024).
Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J. Comput. Appl. Math. 20, 53–65 (1987).
Acknowledgements
This work was supported by the National Natural Science Foundation of China (Grant No. 82372083).
Author information
Authors and Affiliations
Contributions
L.M.: Conceptualization, funding acquisition, methodology, writing—original draft, writing—review and editing. D.W.: Formal analysis, methodology, visualization, writing—original draft. Y.S. and J.P.: Project administration, investigation, data curation. X.Z. and D.M.: Supervision, resources.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Wang, D., Shi, Y., Pang, J. et al. Data-driven subtyping of early Parkinson’s disease via mutual cross-attention fusion of EEG and dual-task gait features. npj Parkinsons Dis. (2026). https://doi.org/10.1038/s41531-026-01258-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41531-026-01258-2
