Synthetic Avatar Images for Lateral-Shift Estimation in Cervical Dystonia

Approximately 16,000 synthetic avatar images and 100 labeled synthetic avatars for training and validating automated assessment tools for cervical dystonia. The dataset was created by Roland Stenger to overcome the scarcity of clinical training examples for translational symptoms like lateral shift. It was last updated on May 31, 2026.

Use Cases

• Training deep learning models for head pose estimation based on synthetic avatar images.
• Validating automated scoring systems for cervical dystonia severity based on labeled synthetic data.
• Benchmarking image-based tools against expert clinical ratings based on the described validation study.

Strengths

• Contains ~16,000 synthetic images, a substantial volume for model training.
• Includes 100 labeled synthetic avatars for validation purposes.
• Designed specifically to address the scarcity of real clinical training examples for rare symptoms.

Limitations

• Dataset size is listed as 11.0 B, indicating a very limited scope.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data consists entirely of synthetic avatars, which may not fully capture the complexity of real patient images.

Provenance

Source

Roland Stenger via figshare.

Collection Method

Synthetic avatar generation for training a deep learning model.

Freshness

Last updated 2026-05-31 15:10:55; freshness should be verified.