Abstract
Neurodegenerative diseases manifest different motor and cognitive signs and symptoms that are highly heterogeneous. Parsing these heterogeneities may lead to an improved understanding of underlying disease mechanisms; however, current methods are dependent on clinical assessments and an arbitrary choice of behavioural tests. Here we present a data-driven subtyping approach using video-captured human motion and brain functional connectivity from resting-state functional magnetic resonance imaging. We applied our framework to a cohort of individuals at different stages of Parkinson’s disease. The process mapped the data to low-dimensional measures by projecting them onto a canonical correlation space that identified three Parkinson’s disease subtypes: subtype I was characterized by motor difficulties and poor visuospatial abilities; subtype II exhibited difficulties in non-motor components of activities of daily living and motor complications (dyskinesias and motor fluctuations) and subtype III was characterized by predominant tremor symptoms. We conducted a convergent validity analysis by comparing our approach to existing and widely used approaches. The compared approaches yielded subtypes that were adequately well-clustered in the motion-brain representation space we created to delineate subtypes. Our data-driven approach, contrary to other forms of subtyping, derived biomarkers predictive of motion impairment and subtype memberships that were captured objectively by digital videos.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 digital issues and online access to articles
$119.00 per year
only $9.92 per issue
Buy this article
- Purchase on SpringerLink
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
We provided a toy dataset of select motion encoding outputs and fMRI subnetworks with our code release62. To protect study participant privacy, we are unable to release the clinical data or the gait examination videos. The NTU RGB+D dataset54 used to pretrain the motion encoder model is available at https://rose1.ntu.edu.sg/dataset/actionRecognition/.
Code availability
We have released code for subtype analysis and predicting motor impairment using FC data62. The previously published code for the GaitForeMer motion encoder is available at https://github.com/markendo/GaitForeMer.
References
Foltynie, T., Brayne, C. & Barker, R. A. The heterogeneity of idiopathic Parkinson’s disease. J. Neurol. 249, 138–145 (2002).
Post, B., Speelman, J. D., De Haan, R. J. & Group, C.-S. Clinical heterogeneity in newly diagnosed Parkinson’s disease. J. Neurol. 255, 716–722 (2008).
Wang, D. D. & Choi, J. T. Brain network oscillations during gait in Parkinson’s disease. Front. Hum. Neurosci. 14, 568703 (2020).
Tessitore, A. et al. Resting-state brain connectivity in patients with Parkinson’s disease and freezing of gait. Parkinsonism Relat. Disord. 18, 781–787 (2012).
Crémers, J., D’Ostilio, K., Stamatakis, J., Delvaux, V. & Garraux, G. Brain activation pattern related to gait disturbances in Parkinson’s disease. Mov. Disord. 27, 1498–1505 (2012).
Endo, M. et al. Gaitforemer: self-supervised pre-training of transformers via human motion forecasting for few-shot gait impairment severity estimation. In Proc. Medical Image Computing and Computer Assisted Intervention–MICCAI 2022: 25th International Conference, Singapore, September 18–22 (eds Wang, L., Dou, Q., Fletcher, P.T., Speidel, S. & Li, S.) 130–139 (Springer, 2022).
Kocabas, M., Athanasiou, N. & Black, M. J. VIBE: video inference for human body pose and shape estimation. In Proc. 2020 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 5253–5263 (IEEE, 2020).
Nerrise, F., Zhao, Q., Poston, K. L., Pohl, K. M. & Adeli, E. An explainable geometric-weighted graph attention network for identifying functional networks associated with gait impairment. In Proc. Medical Image Computing and Computer Assisted Intervention–MICCAI 2023: 26th International Conference, Vancouver, Canada, October 8–12 (eds Greenspan, H. et al.) 723–733 (Springer, 2023).
Van Rooden, S. M. et al. The identification of Parkinson’s disease subtypes using cluster analysis: a systematic review. Mov. Disord. 25, 969–978 (2010).
Lu, M. et al. Quantifying Parkinson’s disease motor severity under uncertainty using mds-updrs videos. Med. Image Anal. 73, 102179 (2021).
Eisinger, R. S. et al. Motor subtype changes in early Parkinson’s disease. Parkinsonism Relat. Disord. 43, 67–72 (2017).
Thenganatt, M. A. & Jankovic, J. Parkinson disease subtypes. JAMA Neurol. 71, 499–504 (2014).
von Coelln, R. et al. The inconsistency and instability of Parkinson’s disease motor subtypes. Parkinsonism Relat. Disord. 88, 13–18 (2021).
Zhang, X. et al. Data-driven subtyping of Parkinson’s disease using longitudinal clinical records: a cohort study. Sci. Rep. 9, 797 (2019).
Belvisi, D. et al. The pathophysiological correlates of Parkinson’s disease clinical subtypes. Mov. Disord. 36, 370–379 (2021).
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 Parkinsons Dis. 7, 83 (2021).
Kipf, T. N. & Welling, M. Semi-supervised classification with graph convolutional networks. In Proc. International Conference on Learning Representations https://openreview.net/pdf?id=SJU4ayYgl (2016).
Müller-Oehring, E. M., Schulte, T., Pfefferbaum, A. & Sullivan, E. V. Disruption of cerebellar-cortical functional connectivity predicts balance instability in alcohol use disorder. Drug Alcohol Depend. 235, 109435 (2022).
McKimm, E. Cerebellar Neuropathology Influences Cerebellar-Prefrontal Cortex Pathways: A Comprehensive Approach as Relevant to Autism Spectrum Disorders. PhD thesis, Univ. Memphis (2016).
Frommann, I. et al. Distinct patterns of cognitive impairment in multiple system atrophy patients of cerebellar and parkinsonian predominance. Basal Ganglia 2, 91–96 (2012).
Zwicker, J. G., Missiuna, C., Harris, S. R. & Boyd, L. A. Brain activation associated with motor skill practice in children with developmental coordination disorder: an fMRI study. Int. J. Dev. Neurosci. 29, 145–152 (2011).
Basaia, S. et al. Cerebro-cerebellar motor networks in clinical subtypes of Parkinson’s disease. npj Parkinsons Dis. 8, 113 (2022).
Hoffland, B. et al. A gait paradigm reveals different patterns of abnormal cerebellar motor learning in primary focal dystonias. Cerebellum 13, 760–766 (2014).
O’Callaghan, C. et al. Cerebellar atrophy in Parkinson’s disease and its implication for network connectivity. Brain 139, 845–855 (2016).
King, L. A. et al. Cognitively challenging agility boot camp program for freezing of gait in parkinson disease. Neurorehabil. Neural Repair 34, 417–427 (2020).
Nasreddine, Z. S. et al. The Montreal cognitive assessment, MoCA: a brief screening tool for mild cognitive impairment. J. Am. Geriatr. Soc. 53, 695–699 (2005).
Benton, A. L., Varney, N. R. & Hamsher, K. D. Visuospatial judgment: a clinical test. Arch. Neurol. 35, 364–367 (1978).
Rodriguez, M., Rodriguez-Sabate, C., Morales, I., Sanchez, A. & Sabate, M. Parkinson’s disease as a result of aging. Aging Cell 14, 293–308 (2015).
Selikhova, M. et al. A clinico-pathological study of subtypes in Parkinson’s disease. Brain 132, 2947–2957 (2009).
Delis, D. C., Kramer, J. H., Kaplan, E. & Ober, B. A. California Verbal Learning Test 2nd edn. (Psychological Corporation, 2000).
Smith, A. Symbol Digit Modalities Test (Western Psychological Services, 1973).
Heijmans, M. et al. Monitoring Parkinson’s disease symptoms during daily life: a feasibility study. npj Parkinsons Dis. 5, 21 (2019).
Mehanna, R., Moore, S., Hou, J. G., Sarwar, A. I. & Lai, E. C. Comparing clinical features of young onset, middle onset and late onset Parkinson’s disease. Parkinsonism Relat. Disord. 20, 530–534 (2014).
Pagano, G., Ferrara, N., Brooks, D. J. & Pavese, N. Age at onset and Parkinson disease phenotype. Neurology 86, 1400–1407 (2016).
Wickremaratchi, M. M. et al. The motor phenotype of Parkinson’s disease in relation to age at onset. Mov. Disord. 26, 457–463 (2011).
Wagner, M. L., Fedak, M. N., Sage, J. I. & Mark, M. H. Complications of disease and therapy: a comparison of younger and older patients with Parkinson’s disease. Annal. Clin. Lab. Sci. 26, 389–395 (1996).
Benedict, R. H., Schretlen, D., Groninger, L., Dobraski, M. & Shpritz, B. Revision of the brief visuospatial memory test: studies of normal performance, reliability, and validity. Psychol. Assess. 8, 145 (1996).
Hooper, E. H. Hooper Visual Organization Test (VOT) (Western Psychological Services, 1983).
Spreen, O. & Benton, A. L. Neurosensory Center Comprehensive Examination for Aphasia (University of Victoria, Neuropsychology Laboratory, 1977).
Rofes, A. et al. What drives task performance during animal fluency in people with Alzheimer’s disease? Front. Psychol. 11, 1485 (2020).
Bowie, C. R. & Harvey, P. D. Administration and interpretation of the trail making test. Nat. Protoc. 1, 2277–2281 (2006).
Stroop, J. R. Studies of interference in serial verbal reactions. J. Exp. Psychol. 18, 643 (1935).
Kaplan, E., Goodglass, H. & Weintraub, S. Boston Naming Test (Lea and Febiger, 1983).
Murray, M. P., Sepic, S. B., Gardner, G. M. & Downs, W. J. Walking patterns of men with parkinsonism. Am. J. Phys. Med. Rehab. 57, 278–294 (1978).
Disbrow, E. et al. Resting state functional connectivity is associated with cognitive dysfunction in non-demented people with Parkinson’s disease. J. Parkinsons Dis. 4, 453–465 (2014).
Zhong, J. et al. Levodopa imparts a normalizing effect on default-mode network connectivity in non-demented Parkinson’s disease. Neurosci. Lett. 705, 159–166 (2019).
Karunanayaka, P. R. et al. Default mode network differences between rigidity- and tremor-predominant Parkinson’s disease. Cortex 81, 239–250 (2016).
Hou, Y. et al. Default-mode network connectivity in cognitively unimpaired drug-naive patients with rigidity-dominant Parkinson’s disease. J. Neurol. 264, 152–160 (2017).
Brown, C. A. The Default Mode Network and Executive Function: Influence of Age, White Matter Connectivity, and Alzheimer’s Pathology (Univ. Kentucky, 2017).
Heinonen, J. et al. Default mode and executive networks areas: association with the serial order in divergent thinking. PLoS ONE 11, e0162234 (2016).
Crockett, R. A., Hsu, C. L., Best, J. R. & Liu-Ambrose, T. Resting state default mode network connectivity, dual task performance, gait speed, and postural sway in older adults with mild cognitive impairment. Front. Aging Neurosci. 9, 423 (2017).
Wu, T. & Hallett, M. The cerebellum in Parkinson’s disease. Brain 136, 696–709 (2013).
Goetz, C. G. et al. Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS): scale presentation and clinimetric testing results. Mov. Disord. 23, 2129–2170 (2008).
Shahroudy, A., Liu, J., Ng, T.-T. & Wang, G. NTU RGB+ D: a large scale dataset for 3D human activity analysis. In Proc. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 1010–1019 (IEEE, 2016).
Whitfield-Gabrieli, S. & Nieto-Castanon, A. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connect. 2, 125–141 (2012).
Joachims, T. Transductive learning via spectral graph partitioning. In Proc. 20th International Conference on Machine Learning (ICML-03) (eds Fawcett, T. & Mishra, N.) 290–297 (AAAI Press, 2003).
Thompson, B. Canonical Correlation Analysis: Uses and Interpretation Vol. 47 (Sage, 1984).
Lloyd, S. Least squares quantization in pcm. IEEE Trans. Inf. Theory 28, 129–137 (1982).
Arthur, D. & Vassilvitskii, S. K-means++: the advantages of careful seeding. In Proc. Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms (ed Gabow, H.) 1027–1035 (SIAM, 2007).
Chiang, M. M.-T. & Mirkin, B. Intelligent choice of the number of clusters in k-means clustering: an experimental study with different cluster spreads. J. Classif. 27, 3–40 (2010).
Torres, G. J., Basnet, R. B., Sung, A. H., Mukkamala, S. & Ribeiro, B. M. A similarity measure for clustering and its applications. Int. J. Electr. Comput. Syst. Eng. 3, 164–170 (2009).
Endo, M. et al. markendo/pd-subtyping-motion-brain. Zenodo https://doi.org/10.5281/zenodo.12555317 (2024).
Acknowledgements
This research was supported in part by National Institutes of Health grant nos. AA010723 (E.V.S.), AA017347 (E.V.S., E.A.), AG047366 (V.W.H., K.L.P., E.A.), MH113406 (K.M.P.) and AG066515 (V.W.H.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This study was also supported by the Stanford School of Medicine Department of Psychiatry and Behavioral Sciences Jaswa Innovator Award (E.A.) and the Stanford Institute for Human-Centered Artificial Intelligence GCP Cloud Credit (E.A.).
Author information
Authors and Affiliations
Contributions
M.E. was responsible for methodology, investigation, software, visualization and writing of the original draft. F.N. was responsible for the methodology. Q.Z. was responsible for the methodology and for reviewing and editing the paper. E.V.S. was responsible for the methodology and for reviewing and editing the paper. L.F.-F. was responsible for the methodology. V.W.H. conceptualized the project. K.M.P. was responsible for the methodology and for reviewing and editing the paper. K.L.P. conceptualized the project and was responsible for the methodology and for reviewing and editing the paper. E.A. conceptualized the project and was responsible for project administration, supervision, writing of the original draft, and reviewing and editing the paper.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Machine Intelligence thanks Carl Yang, Jing Zhang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Full visualization of subtype characteristics.
The displayed variables exhibited statistically significant (P < 0.05) levels of expression between subtypes (N = 30) as measured by the two-sided Chi-square test for categorical variables and the Kruskal-Wallis test for continuous variables. For each box, the central line indicates the subtype median value, and the top and bottom edges indicate the 75th and 25th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range, and data points beyond the whiskers are represented using the ♦ symbol. Each subtype mean value is overlayed as a gold × symbol, and dashed gold lines connect subtype mean values across subtypes. Overall, Subtype I had the highest disease severity with more severe problems relating to motor aspects of experiences of daily living (MDS-UPDRS Part-II), higher observed motor difficulties (MDS-UPDRS Part-III excluding tremor), and lower visuospatial abilities (JLO). Subtype II exhibited more difficulties in non-motor aspects of experiences of daily living (MDS-UPDRS Part-I) and higher motor complications (MDS-UPDRS Part -V). Subtype III exhibited the lowest disease severity despite not being the youngest cohort. Generally, individuals in this subtype exhibited tremor symptoms, particularly for rest tremor amplitude of upper extremities and consistency of rest tremor. Yet, some individuals within the cohort had no tremor.
Extended Data Fig. 2 Relationship between joint movement/rotation and PD gait impairment severity.
Visualization of motion metrics that exhibit a significant difference (two-sided t-test P < 0.05) between individuals without gait impairment (MDS-UPDRS 3.10 score 0, N = 9) and individuals with gait impairment (MDS-UPDRS 3.10 scores 1-3, N = 45). For each metric, the left plot shows the difference in motion metric expression between individuals without gait impairment and individuals with gait impairment. The right plot depicts the difference in motion metric expression with respect to gait impairment severity. On each box, the central line indicates the median value, and the top and bottom edges indicate the 75th and 25th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range, and data points beyond the whiskers are represented using the ♦ symbol. For plots with statistically significant (P < 0.05) metric values between groups, each group mean value is overlayed as a × symbol, and dashed lines connect the mean group values. Motion metric trends that exist between individuals without gait impairment versus with gait impairment don’t necessarily hold true and are sometimes even reversed when observing individuals with mild gait impairment versus more severe gait impairment.
Extended Data Fig. 3 Analyzing disease duration and onset age.
(A) Visualization of disease duration and onset age across subtypes (N = 30). For each box, the central line indicates the subtype median value, and the top and bottom edges indicate the 75th and 25th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range, and data points beyond the whiskers are represented using the ♦ symbol. For the onset age variable which is significantly different between groups (P < 0.05 from Kruskal-Wallis test), each subtype mean value is overlayed as a gold × symbol, and dashed gold lines connect subtype mean values across subtypes. While disease duration is not significantly different across the three subtypes, onset age is significant and exhibits a pattern similar to the age variable. (B) Visualization of disease duration and age onset in our subtyping space. CCX signifies the motion component of our space. Each oval represents a single individual since there are multiple data points per individual.
Extended Data Fig. 4 Visualization of extracted measurements from gait videos with respect to cohort (PD vs. CTRL) and gait impairment score within the PD cohort.
For each metric, the left plot compares metric values across the Healthy/Control group and the PD group. The right plot compares the metric values across different levels of gait impairment severity (MDS-UPDRS 3.10) among the PD group. On each box, the central line indicates the median value, and the top and bottom edges indicate the 75th and 25th percentiles, respectively. The whiskers extend to 1.5 times the interquartile range, and data points beyond the whiskers are represented using the ♦ symbol. For plots with statistically significant metric values between groups (P < 0.05 from two-sided t-test for left plots and P < 0.05 from Kruskall-Wallis for right plots), each group mean value is overlayed as a × symbol, and dashed lines connect the mean group values. For plots without statistical significance, the overall metric mean across all groups is drawn as a dashed line.
Supplementary information
Supplementary Information
Supplementary Figs. 1–7, Discussion on brain pathways’ link to motor impairment and Supplementary Table 1.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Endo, M., Nerrise, F., Zhao, Q. et al. Data-driven discovery of movement-linked heterogeneity in neurodegenerative diseases. Nat Mach Intell 6, 1034–1045 (2024). https://doi.org/10.1038/s42256-024-00882-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s42256-024-00882-y
This article is cited by
-
A multiscale approach for biomedical machine learning
Nature Machine Intelligence (2024)