Scholarlink

<p>Bone and soft tissue tumors (BSTTs) are prone to metastasize to the lungs to form pulmonary nodules, which are quite different from other metastatic nodules or primary nodules, such as those from lung, breast, or colorectal cancers. Early detection of these specific lung nodules is crucial for timely intervention and treatment of BSTTs. To promote the development of BSTTs detection algorithm based on deep learning, this paper releases a Computed Tomography (CT) dataset containing 59 patients and 779 BSTTs metastatic pulmonary nodules with pixel-level annotations. We further publish benchmark lung nodule detection experiments on this dataset, achieving F1 score of 0.842, respectively, demonstrating its research potential. We release raw CT data, preprocessing codes for data conversion and coordinate extraction, to enable researchers to obtain precisely calibrated datasets for training deep learning models and promote the utilization of this clinical scientific data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advance Learning in Oncology - A CT Imaging Dataset of Lung Nodule Metastases from Bone and Soft Tissue Tumors

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月26日

查看全文
  • Shaobo Han,
  • Hao Wang,
  • Wenjian Sun,
  • Xiang Liu,
  • Wacili Da,
  • Yang Luo,
  • Li Min,
  • Chunbo Luo

摘要

Bone and soft tissue tumors (BSTTs) are prone to metastasize to the lungs to form pulmonary nodules, which are quite different from other metastatic nodules or primary nodules, such as those from lung, breast, or colorectal cancers. Early detection of these specific lung nodules is crucial for timely intervention and treatment of BSTTs. To promote the development of BSTTs detection algorithm based on deep learning, this paper releases a Computed Tomography (CT) dataset containing 59 patients and 779 BSTTs metastatic pulmonary nodules with pixel-level annotations. We further publish benchmark lung nodule detection experiments on this dataset, achieving F1 score of 0.842, respectively, demonstrating its research potential. We release raw CT data, preprocessing codes for data conversion and coordinate extraction, to enable researchers to obtain precisely calibrated datasets for training deep learning models and promote the utilization of this clinical scientific data.