Machine Learning Approaches to Influential ROI Selection in Parkinson’s Disease: A Comparative Analysis of LASSO, Recursive Feature Elimination, and Random Forest
| Author | Keyvan Olazadeh | en |
| Author | Nasrin Borumandnia | en |
| Author | Hamid Alavi Majd | en |
| Orcid | Keyvan Olazadeh [0000-0003-2828-5805] | en |
| Orcid | Nasrin Borumandnia [0000-0002-7583-8313] | en |
| Orcid | Hamid Alavi Majd [0000-0001-7772-2923] | en |
| Issued Date | 2025-10-31 | en |
| Abstract | Background: Identifying key brain regions implicated in Parkinson’s disease (PD) can enhance both diagnostic accuracy and our understanding of disease mechanisms. Objectives: This study aims to compare three machine learning methods — least absolute shrinkage and selection operator (LASSO), random forest (RF), and recursive feature elimination (RFE) — for selecting influential regions of interest (ROIs) from functional magnetic resonance imaging (fMRI) data to distinguish PD patients from healthy controls. Methods: This retrospective analysis used fMRI data from 15 patients with PD and 15 matched healthy controls, sourced from an open-access database. Three machine learning approaches were applied to identify significant ROIs associated with PD. The selected ROIs were subsequently evaluated using logistic regression models, assessing classification performance through area under the curve (AUC), sensitivity, and specificity. A comparative analysis of model performance was conducted using DeLong’s test. Results: The LASSO identified 9 ROIs, RF selected 10, and RFE identified 4 key ROIs. Logistic regression models constructed with these ROIs yielded AUC values of 0.96, 0.94, and 0.88 for LASSO, RF, and RFE, respectively. Both sensitivity and specificity were highest for LASSO (0.92 for both). DeLong’s test revealed statistically significant differences among the methods (P < 0.001), with LASSO outperforming RF and RFE. Conclusions: This study demonstrates that LASSO, RFE, and RF machine learning techniques are promising for identifying key brain regions, showing preliminary alignment with clinical observations. Focusing on patients with PD, it highlights regions associated with executive function, memory, motor skills, and sensory processing. Early detection of abnormal connectivity in these areas may potentially inform exploratory preventive strategies for PD. | en |
| DOI | https://doi.org/10.5812/ans-165741 | en |
| Keyword | Parkinson Disease | en |
| Keyword | Functional Magnetic Resonance Imaging (fMRI) | en |
| Keyword | Machine Learning (ML) | en |
| Keyword | Brain Mapping | en |
| Keyword | Logistic Models | en |
| Keyword | Cognition | en |
| Publisher | Brieflands | en |
| Title | Machine Learning Approaches to Influential ROI Selection in Parkinson’s Disease: A Comparative Analysis of LASSO, Recursive Feature Elimination, and Random Forest | en |
| Type | Research Article | en |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- ans-12-4-165741-publish-pdf.pdf
- Size:
- 387.18 KB
- Format:
- Adobe Portable Document Format
- Description:
- Article/s PDF