Super-Resolution Ultrasound Radiomics for Predicting Nondiagnostic Thyroid Nodules

437 patient records from a single-center retrospective study evaluating a GAN-based super-resolution radiomics pipeline for predicting nondiagnostic thyroid nodule cytology. The dataset includes development and validation cohorts and was published by Shaozheng He on figshare under a CC-BY-4.0 license. The file was last updated in May 2026.

Use Cases

• Training classifiers for predicting nondiagnostic fine-needle aspiration outcomes based on radiomic features.
• Comparing the performance of super-resolution versus normal-resolution ultrasound image analysis.
• Conducting qualitative error analysis on false-positive and false-negative cases in diagnostic models.
• Evaluating post hoc probability calibration techniques like Platt scaling on clinical prediction models.

Strengths

• Includes data from 437 patients with a split of 338 for development and 99 for independent validation.
• Describes a multistep feature selection pipeline using Mann-Whitney U test, Spearman correlation, RFE, and LASSO regression.
• Reports specific model performance metrics, such as an SR-RF model achieving an AUC of 0.808 in training and 0.7335 in validation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is from a single-center retrospective study, which may limit generalizability.

Provenance

Source

figshare, author Shaozheng He.

Collection Method

Retrospective collection of 2D B-mode ultrasound images from patients with thyroid nodules.

Freshness

Last updated 2026-05-01 04:18:41; freshness should be verified.