RACF: A Multimodal Deep Learning Framework for Parkinson’s Disease Diagnosis Using SNP and MRI Data
Open accessThe DOI matches its Crossref record; 44 of the 44 references it cites were found in a registry or an index.
Abstract
The clinical diagnosis of Parkinson’s disease (PD) primarily relies on clinician-administered observational assessment tools, such as the Unified Parkinson’s Disease Rating Scale (UPDRS). However, these approaches are significantly influenced by subjectivity and exhibit insufficient sensitivity for early-stage symptom detection. The introduction of deep learning techniques has opened new avenues for the early diagnosis of PD. In contrast to traditional methods, deep learning models are capable of processing large-scale, high-dimensional, and complex datasets to automatically learn latent feature relationships, making them particularly suitable for scenarios involving multimodal data fusion. The multimodal diagnosis of PD is confronted with two enduring challenges: (1) the dependence on pre-existing knowledge of established genetic risk loci, and (2) the low efficiency and limited interpretability in handling interactions among cross-modal features. To address these challenges, this study introduces an innovative multimodal deep learning framework with two primary contributions: (1) a Genome-Wide Association Study (GWAS)-Transformer architecture that autonomously selects single nucleotide polymorphism (SNP) features through GWAS and utilizes a multi-head attention mechanism to model potential associations between non-risk loci, thereby eliminating the reliance on known susceptibility genes; (2) a Residual Attention Contrastive Fusion (RACF) module that tackles the heterogeneity of cross-modal features by dynamically allocating attention weights and applying contrastive loss constraints. Evaluation results on the Parkinson’s Progression Markers Initiative (PPMI) dataset demonstrate that our model achieves a classification accuracy of 91.2% and an AUC of 0.94, and predicts nine potential novel risk loci. This work presents a novel paradigm for the discovery of new risk loci based on deep learning and offers valuable insights from a multi-omics perspective for advancing PD research.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538
XGBoost: A Scalable Tree Boosting System
Chen, T., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939785 · Cited by 54,569
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. (2008) · Movement Disorders · doi:10.1002/mds.22340 · Cited by 7,536
Hearst, M. A., Dumais, S. T., Osuna, E., Platt, J., & Scholkopf, B. (1998) · IEEE Intelligent Systems and their Applications · doi:10.1109/5254.708428 · Cited by 7,243
Poewe, W., Seppi, K., Tanner, C. M., Halliday, G. M., Brundin, P., Volkmann, J., Schrag, A.-E., & Lang, A. E. (2017) · Nature Reviews Disease Primers · doi:10.1038/nrdp.2017.13 · Cited by 5,415
AGEING AND PARKINSON'S DISEASE: SUBSTANTIA NIGRA REGIONAL SELECTIVITY
FEARNLEY, J. M., & LEES, A. J. (1991) · Brain · doi:10.1093/brain/114.5.2283 · Cited by 3,529
Epidemiology of Parkinson’s disease
Tysnes, O.-B., & Storstein, A. (2017) · Journal of Neural Transmission · doi:10.1007/s00702-017-1686-y · Cited by 2,414
Challenges in the diagnosis of Parkinson's disease
Tolosa, E., Garrido, A., Scholz, S. W., & Poewe, W. (2021) · The Lancet Neurology · doi:10.1016/s1474-4422(21)00030-2 · Cited by 1,874
The cognitive functions of the caudate nucleus
Grahn, J. A., Parkinson, J. A., & Owen, A. M. (2008) · Progress in Neurobiology · doi:10.1016/j.pneurobio.2008.09.004 · Cited by 940
Multilayer perceptron and neural networks
Popescu, M.-C., Bălaş, V. E., Perescu-Popescu, L., & Mastorakis, N. E. (2009) · doi:10.5555/1639537.1639542 · Cited by 586
Cited by
Integration of Genetic and Imaging Data for Alzheimer's Disease Diagnosis and Interpretation
Wang, Y., Wang, Q., Zhou, M., Liang, J., You, L., Asken, B. M., Zhou, X., & Song, Q. (2025) · Advanced Science · doi:10.1002/advs.202507629 · Cited by 8
Dehghanghanatkaman, A. (2025) · Scientific Reports · doi:10.1038/s41598-025-31898-y · Cited by 5
Zhang, J., Zhao, C., Zeng, L., Huang, H., Ding, Y., & Chen, W. (2025) · AI Sensors · doi:10.3390/aisens1010006 · Cited by 3
Zhang, Q., Liu, Z., Song, Z., Song, S., Li, X., Li, Z., & Zuo, M. (2025) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2025.110646 · Cited by 2
Dino, H. I., & Hassan, M. M. (2025) · Neuroscience & Biobehavioral Reviews · doi:10.1016/j.neubiorev.2025.106399 · Cited by 3
Yue, H., Kuang, H., He, M., Li, J., Wang, Y., & Wang, J. (2026) · Interdisciplinary Sciences Computational Life Sciences · doi:10.1007/s12539-026-00829-4 · Cited by 1
Design of a deep learning prediction model for Alzheimer's and Parkinson's Disease using MRI images
Velu, K., & Jaisankar, N. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2026.1777236 · Cited by 1
Reddy, C. K. K., Fatima, Q., Anisha, P. R., Dewan, B., & Zubair, R. (2026) · Discover Computing · doi:10.1007/s10791-026-10045-9 · Cited by 1
Mabasha, S., Gupta, S., & Babu, K. K. (2026) · International Journal of Computers and Applications · doi:10.1080/1206212x.2026.2690510 · Cited by 1
“Although attention visualization, Grad-CAM, and SHAP-based methods have improved interpretability, they provide only limited information about the intricate decision-making processes involved in multimodal integration analysis [25,26].”
InfluentialCamastra, C., Pelagi, A., Quattrone, A., & Sarica, A. (2026) · Brain Sciences · doi:10.3390/brainsci16040405 · Cited by 1
Peixoto, E., de Sousa, F. C., Chen, J., Camacho, D., Benjamin, S. R., & de Albuquerque, V. H. C. (2026) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2026.111468 · Cited by 1
Vishalini, P., & Anuradha, T. (2026) · International Conference on Communication, Computing and Electronics Systems (ICCCES) · doi:10.1109/iccces62661.2026.11437067
“Many studies do not fully document modality-wise preprocessing pipelines, fusion strategies, data splits, or uncertainty estimation (e.g., [23], [27], [28], [33]).”
InfluentialHybrid Machine Learning Based Model for Early Parkinson's Disease Detection
Sotelo, R., Perez, C., & Santisteban, J. (2026) · International Conference on Artificial Intelligence, Computer, Data Sciences and Applications (ACDSA) · doi:10.1109/acdsa67686.2026.11467613
“Moreover, a few studies proposed the late fusion approach, which leverages the explanatory capacity of tabular models and the representational strength of CNNs, ensuring response times under two seconds—suitable for clinical environments [12].”
Multimodal Feature Fusion in Mixed Reality for Detection of Neurodegenerative Disorders
Hemmerling, D., Stępień, J., Krzywdziak, J., Dudek, M., Szecowka, W., Śliwa, A., Wodziński, M., Danioł, M., Igraś-Cybulska, M., & Wójcik-Pędziwiatr, M. (2026) · IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW) · doi:10.1109/vrw70859.2026.00150
“These patterns align with prior evidence that wearable inertial and kinematic signals are highly informative for PD motor manifestations [21, 6, 27, 16, 29, 1, 22, 20, 18, 5] and that multimodal models outperform unimodal counterparts by capturing complementary physiology [8, 24, 15, 26].”
Bakaraniya Parul, V., & Padole, M. (2026) · Journal of Trends in Computer Science and Smart Technology · doi:10.36548/jtcsst.2026.2.010
“RACF by She et al. [27] is a multimodal DL model that provides a method for integrating single-nucleotide polymorphism (SNP) data and structural MRI to diagnose Parkinson’s disease, utilizing dynamic cross -modal fusion through attention and contrastive learning.”
Early Diagnosis of Parkinson’s Disease using Handwriting Pattern Analysis and Deep Learning
Vashisth, A., Bakhshizada, R., Hossain, S., & Kaur, G. (2026) · International conference Computer Networks and Inventive Communication Technologies · doi:10.1109/iccnct68477.2026.11590406
“Moreover, the authors tested deep learning models on speech and handwriting data to detect early PD, and the multimodal results are effective [6].”
Dino, H. I., & Hassan, M. M. (2026) · Signal Image and Video Processing · doi:10.1007/s11760-026-05575-0
“Despite a modest cohort of 241 subjects, LiteCrossFu-sion achieves 96.77% ACC, 0.9861 AUC, and 0.976 F1 — competitive with or exceeding Cao and Long [18] (91.2%), WhileKarthigeyanandRani[22]reported99 %with ∼ 12M parameters and Dentamaro et al. [17] achieved 96.6% with 14.4M parameters, our…”
InfluentialNachukuru, M., & Sushama, C. (2026) · Annals of Movement Disorders · doi:10.4103/aomd.aomd_5_26
Sun, C., Xu, Y., Xu, J., Wang, H., Yang, L., 刘腾飞, Li, W., Chen, L., Wu, X. L., & Zhu, W. (2026) · Information Fusion · doi:10.1016/j.inffus.2026.104818
Hani, M., Betrouni, N., Mahmoudi, S., & Benjelloun, M. (2026) · AI Computer Science and Robotics Technology · doi:10.5772/acrt.20250083
Parkinson's disease pathogenesis: From molecular insights to novel therapies
Ramli, M. D. C., Abidin, Z. Z., Okunsai, B., Umbar, G. S. A., & Mauta, T. E. (2026) · Elsevier eBooks · doi:10.1016/b978-0-443-27702-3.00027-3
Rajendra, K. R., Gao, J., Devihosur, R. G., & Kathrotiya, M. (2026) · Informatics in Medicine Unlocked · doi:10.1016/j.imu.2026.101824
Metadata
- Authors
- Cao, J., & Long, X.
- Affiliations
- Chinese Academy of Sciences (CN)
- Corresponding author
- Long, Xiaojing
- ORCID
- Cao, Jiangbo
- Venue
- Applied Sciences
- Publisher
- MDPI AG
- Published
- 2025-04-19
- Volume
- 15
- Issue
- 8
- Pages
- 4513
- ISSN
- 2076-3417 (ISSN-L)
- OpenAlex
- W4409606449
- Cited by
- 22
- References
- 44
- Citations by year
- 2026: 17 · 2025: 5
- FWCI
- 13.46
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Applied Sciences (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Key R&D Program of China (2021ZD0200401, 2023YFC2410504, JCYJ20241202124922031, 2024B1212010010)
- Shenzhen Science and Technology Program (2021ZD0200401, 2023YFC2410504, JCYJ20241202124922031, 2024B1212010010)
- Guangdong Provincial Key Laboratory of Multimodality Non-Invasive Brain–Computer Interfaces (2021ZD0200401, 2023YFC2410504, JCYJ20241202124922031, 2024B1212010010)
- Topics
- Health Sciences › Medicine › Clinical Neurology › Parkinson's Disease Mechanisms and Treatments
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
- Keywords
- Parkinson's disease, early diagnosis, multimodal fusion, single nucleotide polymorphisms, magnetic resonance imaging, multi-head attention, contrastive learning, residual attention, multi-omics, PPMI
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.