Ultrasound Radiomics Features for Breast Lesion Classification, 237 Patients

237 patients underwent preoperative breast ultrasound examinations at Quzhou People’s Hospital between April 2013 and April 2023. Radiomic features were extracted from ultrasound images to develop a machine learning model for distinguishing granulomatous lobular mastitis from breast cancer. The dataset includes 1,161 radiomic features per image, with a combined model achieving an AUC of 0.935 in the training cohort.

Use Cases

• Developing a diagnostic classifier for granulomatous lobular mastitis based on ultrasound radiomic features.
• Evaluating the performance of machine learning models like Extremely Randomized Trees and LightGBM for medical imaging tasks.
• Investigating the correlation between specific imaging biomarker features (e.g., lbp_3D_k_glszm_SmallAreaLowGrayLevelEmphasis) and disease prediction.
• Assessing clinical utility of a combined model integrating radiomics signatures with independent clinical predictors.
• Applying SHAP analysis for model explainability in a clinical diagnostic context.

Strengths

• Dataset includes 237 patient cases with pathological diagnoses.
• 1,161 radiomic features were extracted per ultrasound image, providing a detailed feature set.
• Model performance is quantified with AUC, accuracy, sensitivity, and specificity metrics, including confidence intervals.
• The study employs multiple feature selection methods (ICC, Pearson correlation, LASSO) and three machine learning algorithms.

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• Data may reflect geographic and temporal bias inherent to a single hospital (Quzhou People’s Hospital) over a 10-year period.
• The dataset is small (355 KB), indicating limited scope, likely containing processed feature data rather than raw images.

Provenance

Source

Quzhou People’s Hospital

Collection Method

Retrospective study of patients who underwent preoperative breast ultrasound and received pathological diagnoses.

Time Range

April 2013 to April 2023

Freshness

Last updated 2026-04-24 04:22:36; freshness should be verified.

Geography

Quzhou People’s Hospital (location unspecified)