Scholarlink

Objectives <p>About 20% of renal masses are proved to be benign after surgery. We aimed to develop a strategy based on radiomics to differentially diagnose renal masses.Methods: In this retrospective study, 726 renal masses (54 benign and 672 malignant) were collected from two hospitals in China and two public databases. Two diagnostic models, kidney mass diagnosis model (KMDM) and kidney cancer subtypes diagnosis model (KCSDM), were developed for differential diagnosis of renal cell carcinoma and their subtypes. Each model was divided into three units (radiomics model, clinical model, and integrated model) and evaluated in the training set and two independent validation sets.</p> Methods <p>In this retrospective study, 726 renal masses (54 benign and 672 malignant) were collected from two hospitals in China and two public databases. Two diagnostic models, kidney mass diagnosis model (KMDM) and kidney cancer subtypes diagnosis model (KCSDM), were developed for differential diagnosis of renal cell carcinoma and their subtypes. Each model was divided into three units (radiomics model, clinical model, and integrated model) and evaluated in the training set and two independent validation sets.</p> Results <p>A total of 39 radiomic variables were used for the KMDM, and 78 variables for the KCSDM. The radiomics model performed better than the clinical model in both KMDM and KCSDM. The integrated model of KMDM showed the best performance (AUC of 0.937 in training set, 0.942 in both two validation sets), when compared to the other two models and two experts. A cutoff value of risk of malignant index (RMI=1.661) was designated to distinguish the benign and malignant renal masses with a sensitivity of 0.929 and a specificity of 0.833 in validation sets. Similar results were found in the integrated model of KCSDM.</p> Conclusion <p>The integrated model can noninvasively distinguish benign from malignant renal masses and renal cell carcinoma subtypes with satisfactory sensitivity and specificity.</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

Noninvasive renal mass diagnostic models based on radiomic features and clinical data: a multi-center study

  • 期刊论文

  • 发表日期：

    2026年2月7日

查看全文
  • Aiming Xu,
  • Yufan Luo,
  • Jiajun Xing,
  • Yiheng Ding,
  • Shihang Pu,
  • Shijie Qian,
  • Minda Li,
  • Shifeng Su,
  • Yingying Gu,
  • Qianwei Xing,
  • Sipeng Shen,
  • Zengjun Wang

摘要

Objectives

About 20% of renal masses are proved to be benign after surgery. We aimed to develop a strategy based on radiomics to differentially diagnose renal masses.Methods: In this retrospective study, 726 renal masses (54 benign and 672 malignant) were collected from two hospitals in China and two public databases. Two diagnostic models, kidney mass diagnosis model (KMDM) and kidney cancer subtypes diagnosis model (KCSDM), were developed for differential diagnosis of renal cell carcinoma and their subtypes. Each model was divided into three units (radiomics model, clinical model, and integrated model) and evaluated in the training set and two independent validation sets.

Methods

In this retrospective study, 726 renal masses (54 benign and 672 malignant) were collected from two hospitals in China and two public databases. Two diagnostic models, kidney mass diagnosis model (KMDM) and kidney cancer subtypes diagnosis model (KCSDM), were developed for differential diagnosis of renal cell carcinoma and their subtypes. Each model was divided into three units (radiomics model, clinical model, and integrated model) and evaluated in the training set and two independent validation sets.

Results

A total of 39 radiomic variables were used for the KMDM, and 78 variables for the KCSDM. The radiomics model performed better than the clinical model in both KMDM and KCSDM. The integrated model of KMDM showed the best performance (AUC of 0.937 in training set, 0.942 in both two validation sets), when compared to the other two models and two experts. A cutoff value of risk of malignant index (RMI=1.661) was designated to distinguish the benign and malignant renal masses with a sensitivity of 0.929 and a specificity of 0.833 in validation sets. Similar results were found in the integrated model of KCSDM.

Conclusion

The integrated model can noninvasively distinguish benign from malignant renal masses and renal cell carcinoma subtypes with satisfactory sensitivity and specificity.