RTS-Net: Thyroid Ultrasound Nodule Segmentation Data for Model Training

RTS-Net is a deep learning model for thyroid nodule segmentation in ultrasound images, achieving an 81.66% F1-score on the TN3K dataset. The model integrates dual-path attention and graph convolution, and was trained on TN3K, DDTI, and a large-scale clinical dataset. The dataset, authored by Xiaojie Sun, was last updated on April 13, 2026.

Use Cases

• Training segmentation models based on thyroid ultrasound images mentioned in the description
• Benchmarking new segmentation architectures against reported performance metrics like F1-score and IoU
• Studying the impact of attention mechanisms and graph convolutions on medical image feature representation
• Investigating model generalization across different clinical datasets (TN3K, DDTI)

Strengths

• Model performance is quantitatively reported, with an 81.66% F1-score on the TN3K dataset
• The dataset was used to train a model that outperformed state-of-the-art methods including UNet and DeepLabv3+
• The model employs specific architectural components like dual-path attention and graph convolution, as detailed in the description

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• The description focuses on model methodology; the exact content and format of the underlying data files are not specified

Provenance

Source

figshare, author Xiaojie Sun

Collection Method

Model training and evaluation data for the proposed RTS-Net segmentation network.

Time Range

null

Freshness

Last updated 2026-04-13 04:35:26; freshness should be verified

Geography

null