pith. sign in

arxiv: 2601.22516 · v1 · submitted 2026-01-30 · 💻 cs.LG · cs.AI

SCOPE-PD: Explainable AI on Subjective and Clinical Objective Measurements of Parkinson's Disease for Precision Decision-Making

Md Mezbahul Islam , John Michael Templeton , Masrur Sobhan , Christian Poellabauer , Ananda Mohan Mondal This is my paper

Pith reviewed 2026-05-16 09:31 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords Parkinson's diseaseExplainable AIRandom ForestSHAPMDS-UPDRSMultimodal dataPrecision medicinePPMI cohort
0
0 comments X

The pith

Combining subjective and objective Parkinson's assessments lets Random Forest predict the disease at 98.66 percent accuracy while naming tremor, bradykinesia, and facial expression as the strongest signals.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds SCOPE-PD, a framework that merges patient-reported symptoms and clinician-scored tests with objective measures to train machine learning models for Parkinson's prediction. On PPMI cohort data, Random Forest outperforms other algorithms when both data types are used together, reaching 98.66 percent accuracy. SHAP explanations then rank tremor, bradykinesia, and facial expression as the top three contributors from the MDS-UPDRS assessment. The approach seeks to cut diagnostic delays caused by subjective judgment alone and to deliver interpretable, patient-specific risk estimates instead.

Core claim

SCOPE-PD integrates subjective MDS-UPDRS assessments with objective clinical measurements collected from the PPMI study into a single multimodal dataset. Multiple machine learning algorithms are trained on this dataset; the Random Forest model attains the highest accuracy of 98.66 percent when both subjective and objective features are combined. SHAP analysis applied to the best model identifies tremor, bradykinesia, and facial expression as the three highest-ranking contributors from the MDS-UPDRS test to the Parkinson's disease prediction outcome.

What carries the argument

Random Forest classifier trained on the combined subjective-objective feature set and interpreted with SHAP value rankings to expose the most influential MDS-UPDRS items.

If this is right

  • Clinicians gain concrete guidance to prioritize tremor, bradykinesia, and facial expression during early assessment.
  • Prediction performance improves measurably when subjective and objective data streams are fused rather than used in isolation.
  • SHAP-derived explanations enable individualized risk reports that can be shown directly to patients and care teams.
  • The same multimodal workflow can be reused for other movement disorders that collect comparable subjective and objective scores.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • External validation on cohorts outside PPMI would be needed before claiming the accuracy generalizes to routine clinical use.
  • Adding continuous sensor streams such as smartphone tapping tests could further strengthen the objective component without increasing subjectivity.
  • The ranked features could serve as targets for designing shorter screening batteries or focused monitoring protocols in future trials.

Load-bearing premise

The PPMI cohort is treated as representative of broader Parkinson's populations so that the 98.66 percent accuracy and feature rankings will hold for new patients without site-specific bias or overfitting.

What would settle it

Retraining or testing the same Random Forest pipeline on an independent Parkinson's dataset collected at a different clinical site or in a different demographic group and obtaining accuracy well below 90 percent.

read the original abstract

Parkinson's disease (PD) is a chronic and complex neurodegenerative disorder influenced by genetic, clinical, and lifestyle factors. Predicting this disease early is challenging because it depends on traditional diagnostic methods that face issues of subjectivity, which commonly delay diagnosis. Several objective analyses are currently in practice to help overcome the challenges of subjectivity; however, a proper explanation of these analyses is still lacking. While machine learning (ML) has demonstrated potential in supporting PD diagnosis, existing approaches often rely on subjective reports only and lack interpretability for individualized risk estimation. This study proposes SCOPE-PD, an explainable AI-based prediction framework, by integrating subjective and objective assessments to provide personalized health decisions. Subjective and objective clinical assessment data are collected from the Parkinson's Progression Markers Initiative (PPMI) study to construct a multimodal prediction framework. Several ML techniques are applied to these data, and the best ML model is selected to interpret the results. Model interpretability is examined using SHAP-based analysis. The Random Forest algorithm achieves the highest accuracy of 98.66 percent using combined features from both subjective and objective test data. Tremor, bradykinesia, and facial expression are identified as the top three contributing features from the MDS-UPDRS test in the prediction of PD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript introduces SCOPE-PD, an explainable AI framework that combines subjective and objective clinical assessments from the PPMI dataset to predict Parkinson's disease. Multiple ML models are evaluated, with Random Forest reported to achieve 98.66% accuracy on the combined feature set; SHAP analysis identifies tremor, bradykinesia, and facial expression (from MDS-UPDRS) as the top contributing features for individualized predictions.

Significance. If the reported performance is shown to hold under subject-level partitioning and external validation, the work would provide a useful multimodal, interpretable approach to PD prediction that addresses limitations of subjective-only methods. The emphasis on SHAP-based explanations supports clinical translation by highlighting actionable features.

major comments (1)
  1. [Abstract / Methods] Abstract and Methods: The central performance claim of 98.66% accuracy with Random Forest on combined features is presented without any description of the train-test partitioning strategy, cross-validation procedure, class-balance handling, or statistical testing. PPMI is a longitudinal cohort with repeated visits per subject; without explicit confirmation of subject-level (rather than visit-level) splitting, the result cannot be verified as free of leakage and therefore cannot support the generalization claim.
minor comments (1)
  1. [Abstract] Abstract: The number of subjects, total features, and class distribution are not stated, making it difficult to contextualize the reported accuracy and feature rankings.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The comment highlights an important gap in the description of our experimental protocol, which we have addressed through targeted revisions to improve clarity and verifiability.

read point-by-point responses

  1. Referee: [Abstract / Methods] Abstract and Methods: The central performance claim of 98.66% accuracy with Random Forest on combined features is presented without any description of the train-test partitioning strategy, cross-validation procedure, class-balance handling, or statistical testing. PPMI is a longitudinal cohort with repeated visits per subject; without explicit confirmation of subject-level (rather than visit-level) splitting, the result cannot be verified as free of leakage and therefore cannot support the generalization claim.

    Authors: We agree that the original submission did not sufficiently detail the partitioning strategy, cross-validation, class-balance handling, or statistical testing. In the revised manuscript we will add an explicit subsection in Methods that states: (i) subject-level splitting was performed so that all visits belonging to the same PPMI participant appear exclusively in the training or test partition; (ii) stratified 5-fold cross-validation was used on the training subjects; (iii) class imbalance was mitigated by class-weighting within the Random Forest implementation; and (iv) 95 % bootstrap confidence intervals together with McNemar’s test were computed for model comparisons. These additions directly confirm the absence of leakage and support the reported generalization performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical ML pipeline on external PPMI data

full rationale

The paper reports standard supervised learning results: multiple ML models are trained on PPMI subjective+objective features, Random Forest is selected for highest accuracy (98.66%), and SHAP is applied post-hoc for interpretability. No equations, derivations, or claims reduce a 'prediction' to a fitted parameter by construction, nor rely on self-citation load-bearing uniqueness theorems. Performance metrics are measured on held-out data splits from an external cohort; the process contains no self-definitional or renaming steps that collapse the result to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework depends on the representativeness of the PPMI cohort and standard assumptions of supervised classification; no new entities or free parameters are introduced beyond typical ML hyperparameters.

axioms (1)
  • domain assumption The PPMI dataset is representative of the target PD population for training and evaluation
    All reported performance rests on this external data source without additional validation cohorts mentioned.

pith-pipeline@v0.9.0 · 5544 in / 1244 out tokens · 59649 ms · 2026-05-16T09:31:56.177217+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. STEP-PD: Stage-Aware and Explainable Parkinson's Disease Severity Classification Using Multimodal Clinical Assessments

    cs.LG 2026-04 unverdicted novelty 4.0

    STEP-PD applies XGBoost to multimodal PPMI clinical data for three-class PD severity staging with 94.14% accuracy and SHAP explanations highlighting a shift from motor to balance impairments.

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages · cited by 1 Pith paper

  1. [1]

    Frontiers in aging neuroscience16, 1498756 (2025)

    Luo, Y., Qiao, L., Li, M., Wen, X., Zhang, W., Li, X.: Global, regional, national epidemiology and trends of parkinson’s disease from 1990 to 2021: findings from the global burden of disease study 2021. Frontiers in aging neuroscience16, 1498756 (2025)

    work page 1990

  2. [2]

    Cellular and Molecular Neurobiology 45(1), 1–16 (2025)

    Yaribash, S., Mohammadi, K., Sani, M.A.: Alpha-synuclein pathophysiology in neurodegenerative disorders: A review focusing on molecular mechanisms and treatment advances in parkinson’s disease. Cellular and Molecular Neurobiology 45(1), 1–16 (2025)

    work page 2025

  3. [3]

    Drugs-Real World Outcomes11(1), 1–11 (2024)

    Chaudhuri, K.R., Azulay, J.-P., Odin, P., Lindvall, S., Domingos, J., Alobaidi, A., Kandukuri, P.L., Chaudhari, V.S., Parra, J.C., Yamazaki, T.,et al.: Economic burden of parkinson’s disease: A multinational, real-world, cost-of-illness study. Drugs-Real World Outcomes11(1), 1–11 (2024)

    work page 2024

  4. [4]

    Journal of Neurology268(9), 3444–3455 (2021)

    Siciliano, M., Trojano, L., De Micco, R., Sant’Elia, V., Giordano, A., Russo, A., Passamonti, L., Tedeschi, G., Chiorri, C., Tessitore, A.: Correlates of the dis- crepancy between objective and subjective cognitive functioning in non-demented patients with parkinson’s disease. Journal of Neurology268(9), 3444–3455 (2021)

    work page 2021

  5. [5]

    In: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp

    Islam, M.M., Templeton, J.M., Poellabauer, C., Mondal, A.M.: Redone-pd: Reflections of dopamine-related gene mutations on neurocognitive functions in healthy controls and parkinson’s disease. In: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 6113–6120 (2024). IEEE

    work page 2024

  6. [6]

    Procedia computer science132, 1788–1794 (2018)

    Grover, S., Bhartia, S., Yadav, A., KR, S.,et al.: Predicting severity of parkinson’s disease using deep learning. Procedia computer science132, 1788–1794 (2018)

    work page 2018

  7. [7]

    Scientific reports12(1), 14036 (2022)

    Templeton, J.M., Poellabauer, C., Schneider, S.: Classification of parkinson’s disease and its stages using machine learning. Scientific reports12(1), 14036 (2022)

    work page 2022

  8. [8]

    Current research in translational medicine, 103541 (2025) 14

    Esan, A.O., Olawade, D.B., Soladoye, A.A., Omodunbi, B.A., Adeyanju, I.A., Aderinto, N.: Explainable ai for parkinson’s disease prediction: A machine learning approach with interpretable models. Current research in translational medicine, 103541 (2025) 14

    work page 2025

  9. [9]

    Computer methods and programs in biomedicine136, 79–88 (2016)

    Pereira, C.R., Pereira, D.R., Silva, F.A., Masieiro, J.P., Weber, S.A., Hook, C., Papa, J.P.: A new computer vision-based approach to aid the diagnosis of parkinson’s disease. Computer methods and programs in biomedicine136, 79–88 (2016)

    work page 2016

  10. [10]

    Advances in neural information processing systems30(2017)

    Lundberg, S.M., Lee, S.-I.: A unified approach to interpreting model predictions. Advances in neural information processing systems30(2017)

    work page 2017

  11. [11]

    Cancers17(21), 3454 (2025)

    Sobhan, M., Islam, M.M., Trepka, M.J., Holt, G.E., Dimitroff, C.J., Mondal, A.M.: Tilda-x: Transcriptome-informed lung cancer disparities via explainable ai. Cancers17(21), 3454 (2025)

    work page 2025

  12. [12]

    Expert Systems with Applications265, 126084 (2025)

    Tran, V.Q., Byeon, H.: Explainable hybrid tabular variational autoencoder and feature tokenizer transformer for depression prediction. Expert Systems with Applications265, 126084 (2025)

    work page 2025

  13. [13]

    The Lancet Neurology14(10), 1002–1009 (2015)

    Nalls, M.A., McLean, C.Y., Rick, J., Eberly, S., Hutten, S.J., Gwinn, K., Sutherland, M., Martinez, M., Heutink, P., Williams, N.M.,et al.: Diagno- sis of parkinson’s disease on the basis of clinical and genetic classification: a population-based modelling study. The Lancet Neurology14(10), 1002–1009 (2015)

    work page 2015

  14. [14]

    sleep14(6), 540–545 (1991)

    Johns, M.W.: A new method for measuring daytime sleepiness: the epworth sleepiness scale. sleep14(6), 540–545 (1991)

    work page 1991

  15. [15]

    Best Practices in Nursing Care to Older Adults4(1), 1–2 (2012)

    Greenberg, S.A.: The geriatric depression scale (gds). Best Practices in Nursing Care to Older Adults4(1), 1–2 (2012)

    work page 2012

  16. [16]

    Movement dis- orders: official journal of the Movement Disorder Society23(15), 2129–2170 (2008)

    Goetz, C.G., Tilley, B.C., Shaftman, S.R., Stebbins, G.T., Fahn, S., Martinez- Martin, P., Poewe, W., Sampaio, C., Stern, M.B., Dodel, R.,et al.: Movement disorder society-sponsored revision of the unified parkinson’s disease rating scale (mds-updrs): scale presentation and clinimetric testing results. Movement dis- orders: official journal of the Movemen...

    work page 2008

  17. [17]

    Sleep9(2), 293–308 (1986)

    Schenck, C.H., Bundlie, S.R., Ettinger, M.G., Mahowald, M.W.: Chronic behav- ioral disorders of human rem sleep: a new category of parasomnia. Sleep9(2), 293–308 (1986)

    work page 1986

  18. [18]

    Movement disorders: official journal of the Movement Disorder Society19(11), 1306–1312 (2004)

    Visser, M., Marinus, J., Stiggelbout, A.M., Van Hilten, J.J.: Assessment of auto- nomic dysfunction in parkinson’s disease: the scopa-aut. Movement disorders: official journal of the Movement Disorder Society19(11), 1306–1312 (2004)

    work page 2004

  19. [19]

    Spielberger, C.D.: State-trait anxiety inventory for adults (1983)

    work page 1983

  20. [20]

    Movement disorders27(2), 242–247 (2012)

    Weintraub, D., Mamikonyan, E., Papay, K., Shea, J.A., Xie, S.X., Siderowf, A.: Questionnaire for impulsive-compulsive disorders in parkinson’s disease–rating 15 scale. Movement disorders27(2), 242–247 (2012)

    work page 2012

  21. [21]

    Archives of neurology35(6), 364–367 (1978)

    Benton, A.L., Varney, N.R., Hamsher, K.d.: Visuospatial judgment: A clinical test. Archives of neurology35(6), 364–367 (1978)

    work page 1978

  22. [22]

    The Clinical Neuropsychologist12(1), 43–55 (1998)

    Benedict, R.H., Schretlen, D., Groninger, L., Brandt, J.: Hopkins verbal learning test–revised: Normative data and analysis of inter-form and test-retest reliability. The Clinical Neuropsychologist12(1), 43–55 (1998)

    work page 1998

  23. [23]

    Pearson, ??? (2008)

    Wechsler, D.: Wechsler Adult Intelligence Scale WAIS-IV Canadian. Pearson, ??? (2008)

    work page 2008

  24. [24]

    Occupational Medicine 65(9), 764–765 (2015)

    Hobson, J.: The montreal cognitive assessment (moca). Occupational Medicine 65(9), 764–765 (2015)

    work page 2015

  25. [25]

    Neurological Sciences35(9), 1405–1411 (2014)

    Zarino, B., Crespi, M., Launi, M., Casarotti, A.: A new standardization of semantic verbal fluency test. Neurological Sciences35(9), 1405–1411 (2014)

    work page 2014

  26. [26]

    Los Angeles, CA (2013)

    Smith, A.: Symbol digit modalities test. Los Angeles, CA (2013)

    work page 2013

  27. [27]

    Nature machine intelligence2(1), 56–67 (2020)

    Lundberg, S.M., Erion, G., Chen, H., DeGrave, A., Prutkin, J.M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., Lee, S.-I.: From local explanations to global under- standing with explainable ai for trees. Nature machine intelligence2(1), 56–67 (2020)

    work page 2020

  28. [28]

    Scientific Reports14(1), 20941 (2024)

    Dentamaro, V., Impedovo, D., Musti, L., Pirlo, G., Taurisano, P.: Enhancing early parkinson’s disease detection through multimodal deep learning and explainable ai: insights from the ppmi database. Scientific Reports14(1), 20941 (2024)

    work page 2024

  29. [29]

    Alexandria Engineering Journal107, 568–582 (2024) 16

    Priyadharshini, S., Ramkumar, K., Vairavasundaram, S., Narasimhan, K., Venkatesh, S., Amirtharajan, R., Kotecha, K.: A comprehensive framework for parkinson’s disease diagnosis using explainable artificial intelligence empow- ered machine learning techniques. Alexandria Engineering Journal107, 568–582 (2024) 16

    work page 2024