Parkinson's Disease Voice Recordings and Clinical Benchmark Datasets

195 voice recordings with 22 acoustic features form one of three benchmark datasets used to evaluate a deep learning framework for disease prediction. The dataset, shared by Vijay U. Rathod on figshare, includes results from models like MLP achieving an AUC of 0.7538 for Parkinson's detection. It was last updated on 2026-05-08.

Use Cases

• Benchmarking disease classification models based on acoustic features from voice recordings.
• Evaluating model generalization across heterogeneous clinical datasets as described in the study.
• Training interpretable deep learning models for Parkinson's disease detection based on the 22 acoustic features.
• Studying the impact of dataset characteristics like sample size and class imbalance on model performance.

Strengths

• Includes performance metrics for specific models, such as an MLP achieving an AUC of 0.7538 with perfect specificity for Parkinson's disease.
• Framework evaluation uses three distinct benchmark datasets with specified sample sizes (e.g., 195, 303, 768) and feature counts.
• Dataset is small (5.5 KB) and in XLS format, facilitating quick inspection and download.

Limitations

• Row count for the primary Parkinson's dataset is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics for the 22 acoustic features must be inferred after download.
• Description metadata is limited; actual data quality and completeness require manual inspection.

Provenance

Source

figshare, author Vijay U. Rathod

Collection Method

Likely collected or compiled as part of a research study proposing a unified deep learning framework.

Freshness

Last updated 2026-05-08 17:34:32; freshness should be verified.