2,756 Pulmonary Ultrasound Images with Tumor Annotations

2,756 B-mode ultrasound images of peripheral pulmonary lesions (PPLs) are provided for segmentation and classification tasks. The dataset includes complete annotations and biopsy confirmations from expert sonographers. It was created by Qingbin Wang and last updated on May 11, 2026.

Use Cases

• Train segmentation models based on the provided lesion annotations.
• Develop classification models to distinguish between benign and malignant lesions based on biopsy-confirmed labels.
• Benchmark model performance using the provided five-fold stratified cross-validation splits.

Strengths

• 2,756 annotated medical images provide a substantial base for model training.
• Includes biopsy confirmations for ground-truth classification labels.
• Uses a stratified five-fold cross-validation strategy to ensure balanced class representation in training and testing sets.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific tasks.

Provenance

Source

figshare

Collection Method

Annotations and biopsy confirmations were provided by expert sonographers.

Freshness

Last updated 2026-05-11 20:06:41; freshness should be verified.