A multimodal mutual information-guided feature selection framework for predicting rehabilitation response in Parkinson's disease with postural instability and gait disorder

Abstract

Postural instability and gait disorder (PIGD) subtype of Parkinson's disease (PD) is marked by heterogeneous motor and cognitive impairments, making rehabilitation response difficult to predict. Identifying robust multimodal predictors is essential for precision rehabilitation. This study aimed to identify key multimodal features associated with response to motor-cognitive interactive rehabilitation and to develop a generalizable prediction framework. Twenty-one PD patients with PIGD completed a motor-cognitive interactive rehabilitation program. Multimodal data, including demographics, clinical scales, gait parameters, magnetic resonance imaging (MRI), and EEG, were collected across 14 feature modalities. A multimodal sequential forward selection framework based on mutual information (MSFSF-MI) was proposed where predictive stability of selected feature sets was assessed across five machine learning models (support vector machine, RBF, random forest, stochastic gradient boosting, and XGB). Multimodal feature subsets derived by the proposed framework consistently outperformed unimodal models across classifiers. Cross-model analyses highlighted functional connectivity, cortical thickness, low-frequency power spectral density, and phase–amplitude coupling as reproducible predictors, forming a key feature set mainly from MRI and EEG domains. This study identified predictive and potentially robust multimodal neural features of PD rehabilitation response. The introduced nested prediction framework demonstrates strong potential for future generalization, providing a methodological foundation for personalized neurorehabilitation strategies.

Keywords

feature selection framework / machine learning prediction / multimodal data integration / Parkinson's disease / postural instability and gait disorder

References

[1] Kwon KY, Joo BE, You J, Kim RO. Impact of motor features on non-motor symptoms in patients with de novo Parkinson's disease: cognition, depression, anxiety, fatigue, and dysautonomia. Geriatr Gerontol Int. 2025.

[2] Michels J, van der Wurp H, Kalbe E, et al. Long-term cognitive decline related to the motor phenotype in Parkinson's disease. J Parkinsons Dis. 2022;12(3):905-916. https://doi.org/10.3233/jpd-212787

[3] Lin YP, Lin II, Chiou WD, et al. Optimizing rehabilitation strategies in Parkinson's disease: a comparison of dual cognitive-walking treadmill training and single treadmill training. Sci Rep. 2024;14(1):25210. https://doi.org/10.1038/s41598-024-75422-0

[4] Wong PL, Hung CW, Yang YR, et al. Effects of motor and cognitive complex training on obstacle walking and brain activity in people with Parkinson's disease: a randomized controlled trial. Eur J Phys Rehabil Med. 2024;60(4):611-620. https://doi.org/10.23736/s1973-9087.24.08261-3

[5] Meng L, Wang D, Shi Y, et al. Enhanced brain functional connectivity and activation after 12-week Tai Chi-based action observation training in patients with Parkinson's disease. Front Aging Neurosci. 2023;15:1252610. https://doi.org/10.3389/fnagi.2023.1252610

[6] Bloem BR, Okun MS, Klein C. Parkinson's disease. Lancet. 2021;397(10291):2284-2303. https://doi.org/10.1016/s0140-6736(21)00218-x

[7] Ye H, Robak LA, Yu M, Cykowski M, Shulman JM. Genetics and pathogenesis of Parkinson's syndrome. Annu Rev Pathol. 2023;18(1):95-121. https://doi.org/10.1146/annurev-pathmechdis-031521-034145

[8] Vijayakumari AA, Saadatpour L, Floden D, Fernandez H, Walter BL. Neuroanatomical heterogeneity drives divergent cognitive and motor trajectories in Parkinson's disease subtypes. J Neurol Sci. 2025;468:123335. https://doi.org/10.1016/j.jns.2024.123335

[9] Okai D, Askey-Jones S, Samuel M, David AS, Brown RG. Predictors of response to a cognitive behavioral intervention for impulse control behaviors in Parkinson's disease. Mov Disord. 2015;30(5):736-739. https://doi.org/10.1002/mds.26108

[10] Sun J, Xing F, Feng J, et al. Differential symptom cluster responses and predictors to repetitive transcranial magnetic stimulation treatment in Parkinson's disease: a retrospective study. Heliyon. 2024;10(12):e32799. https://doi.org/10.1016/j.heliyon.2024.e32799

[11] Gates P, Ridgel AL. Body mass index and exercise effort influences changes in motor symptoms after high-cadence dynamic cycling in Parkinson's disease. Front Rehabil Sci. 2022;3:858401. https://doi.org/10.3389/fresc.2022.858401

[12] Tsai SY, Tai CH, Lee YY. Exploring potential predictors of treadmill training effects in people with parkinson disease. Arch Phys Med Rehabil. 2024;105(3):525-530. https://doi.org/10.1016/j.apmr.2023.09.008

[13] Meng D, Jin Z, Chen K, et al. Quality of life predicts rehabilitation prognosis in Parkinson's disease patients: factors influence rehabilitation prognosis: factors influence rehabilitation prognosis. Brain Behav. 2022;12(5):e2579. https://doi.org/10.1002/brb3.2579

[14] Perskaudas R, Myers CE, Interian A, et al. Reward and punishment learning as predictors of cognitive behavioral therapy response in Parkinson's disease comorbid with clinical depression. J Geriatr Psychiatr Neurol. 2024;37(4):282-293. https://doi.org/10.1177/08919887231218753

[15] Wang X, Hu R, Wang T, et al. Resting-state electroencephalographic signatures predict treatment efficacy of tACS for refractory auditory hallucinations in schizophrenic patients. IEEE J Biomed Health Inform. 2025;29(3):1886-1896. https://doi.org/10.1109/jbhi.2024.3509438

[16] Zhou Y, Xie H, Li X, et al. Predicting upper limb motor recovery in subacute stroke patients via fNIRS-measured cerebral functional responses induced by robotic training. J NeuroEng Rehabil. 2024;21(1):226. https://doi.org/10.1186/s12984-024-01523-6

[17] Gupta R, Kumari S, Senapati A, Ambasta RK, Kumar P. New era of artificial intelligence and machine learning-based detection, diagnosis, and therapeutics in Parkinson's disease. Ageing Res Rev. 2023;90:102013. https://doi.org/10.1016/j.arr.2023.102013

[18] Alrawis M, Al-Ahmadi S, Mohammad F. Bridging modalities: a multimodal machine learning approach for Parkinson’s disease diagnosis using EEG and MRI data. Appl Sci. 2024;14(9):3883. https://doi.org/10.3390/app14093883

[19] Miah ASM, Suzuki T, Shin J. A methodological and structural review of Parkinson’s disease detection across diverse data modalities. IEEE Access. 2025;13:98931-98975. https://doi.org/10.1109/access.2025.3575023

[20] Cheng J, Sun J, Yao K, Xu M, Cao Y. A variable selection method based on mutual information and variance inflation factor. Spectrochim Acta Mol Biomol Spectrosc. 2022;268:120652. https://doi.org/10.1016/j.saa.2021.120652

[21] Fu Y, Jiang L, Detre J, Wang Z. Alzheimer's disease classification using mutual information generated graph convolutional network for functional MRI. J Alzheimers Dis. 2025:13872877251350306.

[22] Aatila M, Lachgar M, Hamid H, Kartit A. Keratoconus severity classification using features selection and machine learning algorithms. Comput Math Methods Med. 2021;2021:9979560. https://doi.org/10.1155/2021/9979560

[23] Younis H, Anwar MW, Khan MUG, Sikandar A, Bajwa UI. A new sequential forward feature selection (SFFS) algorithm for mining best topological and biological features to predict protein complexes from protein-protein interaction networks (PPINs). Interdiscip Sci. 2021;13(3):371-388. https://doi.org/10.1007/s12539-021-00433-8

[24] Meng L, Shi Y, Zhao H, Wang D, Zhu X, Ming D. 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. 2024;21(1):166. https://doi.org/10.1186/s12984-024-01456-0

[25] Stebbins GT, Goetz CG, Burn DJ, Jankovic J, Khoo TK, Tilley BC. 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. 2013;28(5):668-670. https://doi.org/10.1002/mds.25383

[26] Goetz CG, Fahn S, Martinez-Martin P, et al. Movement disorder Society-sponsored revision of the unified Parkinson's Disease Rating Scale (MDS-UPDRS): process, format, and clinimetric testing plan. Mov Disord. 2007;22(1):41-47. https://doi.org/10.1002/mds.21198

[27] Goetz CG, Tilley BC, Shaftman SR, 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. 2008;23(15):2129-2170. https://doi.org/10.1002/mds.22340

[28] Shulman LM, Gruber-Baldini AL, Anderson KE, Fishman PS, Reich SG, Weiner WJ. The clinically important difference on the unified Parkinson's disease rating scale. Arch Neurol. 2010;67(1):64-70. https://doi.org/10.1001/archneurol.2009.295

[29] Mishra B, Sudheer P, Rajan R, et al. Bridging the gap between statistical significance and clinical relevance: a systematic review of minimum clinically important difference (MCID) thresholds of scales reported in movement disorders research. Heliyon. 2024;10(5):e26479. https://doi.org/10.1016/j.heliyon.2024.e26479

[30] Yang Y, Chen L, Pang J, Huang X, Meng L, Ming D. Validation of a spatiotemporal gait model using inertial measurement units for early-stage Parkinson's disease detection during turns. IEEE Trans Biomed Eng. 2022;69(12):3591-3600. https://doi.org/10.1109/tbme.2022.3172725

[31] Meng L, Pang J, Yang Y, Chen L, Xu R, Ming D. Inertial-based gait metrics during turning improve the detection of early-stage Parkinson's disease patients. IEEE Trans Neural Syst Rehabil Eng. 2023;31:1472-1482. https://doi.org/10.1109/tnsre.2023.3237903

[32] Meng L, Huang X, Yang Y, Pang J, Chen L, Ming D. An inertial sensor-based algorithm for turning detection during gait. Annu Int Conf IEEE Eng Med Biol Soc. 2021;2021:6372-6375.

[33] Lanzone J, Colombo MA, Sarasso S, et al. EEG spectral exponent as a synthetic index for the longitudinal assessment of stroke recovery. Clin Neurophysiol. 2022;137:92-101. https://doi.org/10.1016/j.clinph.2022.02.022

[34] Kanai S, Oguri M, Okanishi T, et al. Predictive modeling based on functional connectivity of interictal scalp EEG for infantile epileptic spasms syndrome. Clin Neurophysiol. 2024;167:37-48. https://doi.org/10.1016/j.clinph.2024.08.016

[35] Daume J, Kamiński J, Schjetnan AGP, et al. Control of working memory by phase-amplitude coupling of human hippocampal neurons. Nature. 2024;629(8011):393-401. https://doi.org/10.1038/s41586-024-07309-z

[36] V KR, Rajagopalan SS, Bhardwaj S, et al. Machine learning detects EEG microstate alterations in patients living with temporal lobe epilepsy. Seizure. 2018;61:8-13. https://doi.org/10.1016/j.seizure.2018.07.007

[37] Nieto-Castanon A, Whitfield-Gabrieli S. CONN Functional Connectivity Toolbox: RRID SCR_009550, Release 222022.

[38] Gaser C, Dahnke R, Thompson PM, Kurth F, Luders E, the Alzheimer's Disease Neuroimaging Initiative. The Alzheimer's disease neuroimaging initiative. CAT: a computational anatomy toolbox for the analysis of structural MRI data. GigaScience. 2024;13:giae049. https://doi.org/10.1093/gigascience/giae049

[39] Rolls ET, Huang C.-C, Lin C.-P, Feng J, Joliot M. Automated anatomical labelling atlas 3. Neuroimage. 2020;206:116189. https://doi.org/10.1016/j.neuroimage.2019.116189

[40] Destrieux C, Fischl B, Dale A, Halgren E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. Neuroimage. 2010;53(1):1-15. https://doi.org/10.1016/j.neuroimage.2010.06.010

[41] Al-Sarem M, Saeed F, Alkhammash EH, Alghamdi NS. An aggregated mutual information based feature selection with machine learning methods for enhancing IoT Botnet attack detection. Sensors (Basel). 2021;22(1):185. https://doi.org/10.3390/s22010185

[42] Park H, Youm C, Cheon SM, et al. Using machine learning to identify Parkinson's disease severity subtypes with multimodal data. J NeuroEng Rehabil. 2025;22(1):126. https://doi.org/10.1186/s12984-025-01648-2

[43] Mirelman A, Volkov J, Salomon A, et al. Digital mobility measures: a window into real-world severity and progression of Parkinson's disease. Mov Disord. 2024;39(2):328-338. https://doi.org/10.1002/mds.29689

[44] Panthong R, Srivihok A. Wrapper feature subset selection for dimension reduction based on ensemble learning algorithm. Procedia Comput Sci. 2015;72:162-169. https://doi.org/10.1016/j.procs.2015.12.117

[45] Luo L, Ye L, Luo M, Huang D, Peng H, Yang F. Methods of forward feature selection based on the aggregation of classifiers generated by single attribute. Comput Biol Med. 2011;41(7):435-441. https://doi.org/10.1016/j.compbiomed.2011.04.005

[46] Gupta R, Yadav AK, Jha SK, Pathak PK. Composition of feature selection techniques for improving the global horizontal irradiance estimation via machine learning models. Therm Sci Eng Prog. 2024;48:102394. https://doi.org/10.1016/j.tsep.2024.102394

[47] Billot A, Lai S, Varkanitsa M, et al. Multimodal neural and behavioral data predict response to rehabilitation in chronic Poststroke aphasia. Stroke. 2022;53(5):1606-1614. https://doi.org/10.1161/strokeaha.121.036749

[48] Lee YW, Choi JW, Shin E.-H. Machine learning model for predicting malaria using clinical information. Comput Biol Med. 2021;129:104151. https://doi.org/10.1016/j.compbiomed.2020.104151

[49] Zhou E, Wang W, Ma S, et al. Prediction of anxious depression using multimodal neuroimaging and machine learning. Neuroimage. 2024;285:120499. https://doi.org/10.1016/j.neuroimage.2023.120499

[50] Herbers C, Zhang R, Erdman A, Johnson MD. Distinguishing features of Parkinson's disease fallers based on wireless insole plantar pressure monitoring. npj Parkinson's Dis. 2024;10(1):67. https://doi.org/10.1038/s41531-024-00678-2

[51] Vergara JR, Estévez PA. A review of feature selection methods based on mutual information. Neural Comput Appl. 2014;24(1):175-186. https://doi.org/10.1007/s00521-013-1368-0

[52] Mielniczuk J. Information theoretic methods for variable Selection-A review. Entropy. 2022;24(8):1079. https://doi.org/10.3390/e24081079

[53] Mwangi B, Tian TS, Soares JC. A review of feature reduction techniques in neuroimaging. Neuroinformatics. 2014;12(2):229-244. https://doi.org/10.1007/s12021-013-9204-3

[54] Doherty T, Dempster E, Hannon E, et al. A comparison of feature selection methodologies and learning algorithms in the development of a DNA methylation-based telomere length estimator. BMC Bioinf. 2023;24(1):178. https://doi.org/10.1186/s12859-023-05282-4

[55] Suresh S, Newton DT, Everett TH, Lin G, Duerstock BS. Feature selection techniques for a machine learning model to detect autonomic dysreflexia. Front Neuroinf. 2022;16:901428. https://doi.org/10.3389/fninf.2022.901428

[56] Gazzaley A, Rissman J, Cooney J, et al. Functional interactions between prefrontal and visual association cortex contribute to top-down modulation of visual processing. Cerebr Cortex. 2007;17(suppl 1(0 1)):i125-i135. https://doi.org/10.1093/cercor/bhm113

[57] Küchenhoff S, Sorg C, Schneider SC, et al. Visual processing speed is linked to functional connectivity between right frontoparietal and visual networks. Eur J Neurosci. 2021;53(10):3362-3377. https://doi.org/10.1111/ejn.15206

[58] He Y, Shao X, Liu C, et al. Diverse frontoparietal connectivity supports semantic prediction and integration in sentence comprehension. J Neurosci. 2025;45(5):e1404242024. https://doi.org/10.1523/jneurosci.1404-24.2024

[59] Fuchs TA, Benedict RHB, Bartnik A, et al. Preserved network functional connectivity underlies cognitive reserve in multiple sclerosis. Hum Brain Mapp. 2019;40(18):5231-5241. https://doi.org/10.1002/hbm.24768

[60] Anthony M, Lin F. A systematic review for functional neuroimaging studies of cognitive reserve across the cognitive aging spectrum. Arch Clin Neuropsychol. 2018;33(8):937-948. https://doi.org/10.1093/arclin/acx125

[61] Sasabayashi D, Takahashi T, Takayanagi Y, et al. Resting state hyperconnectivity of the default mode network in schizophrenia and clinical high-risk state for psychosis. Cerebr Cortex. 2023;33(13):8456-8464. https://doi.org/10.1093/cercor/bhad131

[62] Azarias FR, Almeida G, de Melo LF, Rici REG, Maria DA. The journey of the default mode network: development, function, and impact on mental health. Biology. 2025;14(4):395. https://doi.org/10.3390/biology14040395

[63] Kharitonova M, Martin RE, Gabrieli JD, Sheridan MA. Cortical gray-matter thinning is associated with age-related improvements on executive function tasks. Dev Cogn Neurosci. 2013;6:61-71. https://doi.org/10.1016/j.dcn.2013.07.002

[64] Basavaraju R, Feng X, France J, Huey ED, Provenzano FA. Depression is associated with preserved cortical thickness relative to apathy in frontotemporal dementia. J Geriatr Psychiatr Neurol. 2022;35(1):78-88. https://doi.org/10.1177/0891988720964258

[65] Kirschen RM, Leaver AM. Hearing function moderates age-related differences in brain morphometry in the HCP aging cohort. Hum Brain Mapp. 2024;45(16):e70074. https://doi.org/10.1002/hbm.70074

[66] Tan E, Troller-Renfree SV, Morales S, et al. Theta activity and cognitive functioning: integrating evidence from resting-state and task-related developmental electroencephalography (EEG) research. Dev Cogn Neurosci. 2024;67:101404. https://doi.org/10.1016/j.dcn.2024.101404

RIGHTS & PERMISSIONS

2026 The Author(s). Journal of Intelligent Medicine published by John Wiley & Sons Australia, Ltd on behalf of Tianjin University.