Scholarlink

<p>This study developed a machine learning model for predicting the presence of tertiary lymphoid structures (TLSs) and treatment response to neoadjuvant therapy (NAT) in triple-negative breast cancer (TNBC). This multicenter study retrospectively included 697 patients, including the training cohort (<i>n</i> = 137), the TLS validation cohort (<i>n</i> = 63) and the NAT response validation cohorts (<i>n</i> = 560). Five machine learning models were developed to predict the presence of TLSs, and the XGBoost model, which exhibited the best performance, was selected as the radiomics-based TLS (rTLS) predictive model. The rTLS predictive model demonstrated robust predictive performance, including across various patient subgroups. Prognostic analysis showed that the rTLS predictive score was significantly correlated with disease-free survival (DFS) in TNBC receiving NAT, and was identified as a strong independent prognostic factor. Pathomic features further explained the pathological heterogeneity of TNBC with different responses to NAT. Overall, the rTLS predictive model, which accurately predicted the presence of TLSs and treatment response to NAT in TNBC, held promise for future clinical application in formulating personalized strategies for TNBC, ultimately improving prognosis, aiding in individualized patient treatment.</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

Machine learning model for predicting tertiary lymphoid structures and treatment response in triple-negative breast cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月1日

查看全文
  • Yidan Lin,
  • Yushuai Yu,
  • Qing Wang,
  • Kaiyan Huang,
  • Shukai Guo,
  • Jie Zhang,
  • Yihui He,
  • Xin Yu,
  • Jiwen Zhang,
  • Fan Meng,
  • Shicong Tang,
  • Junhui Yuan,
  • Chuangui Song

摘要

This study developed a machine learning model for predicting the presence of tertiary lymphoid structures (TLSs) and treatment response to neoadjuvant therapy (NAT) in triple-negative breast cancer (TNBC). This multicenter study retrospectively included 697 patients, including the training cohort (n = 137), the TLS validation cohort (n = 63) and the NAT response validation cohorts (n = 560). Five machine learning models were developed to predict the presence of TLSs, and the XGBoost model, which exhibited the best performance, was selected as the radiomics-based TLS (rTLS) predictive model. The rTLS predictive model demonstrated robust predictive performance, including across various patient subgroups. Prognostic analysis showed that the rTLS predictive score was significantly correlated with disease-free survival (DFS) in TNBC receiving NAT, and was identified as a strong independent prognostic factor. Pathomic features further explained the pathological heterogeneity of TNBC with different responses to NAT. Overall, the rTLS predictive model, which accurately predicted the presence of TLSs and treatment response to NAT in TNBC, held promise for future clinical application in formulating personalized strategies for TNBC, ultimately improving prognosis, aiding in individualized patient treatment.