Scholarlink

Objective <p>To investigate the feasibility of multimodal imaging genomics models in predicting major genetic alterations and assessing prognostic outcomes of patients with lung cancer following various treatments.</p> Methods <p>Computer tomography (CT) images of 533 patients diagnosed with lung adenocarcinoma were retrospectively analyzed together with their pretreatment demographic and clinical data. The patients received standard therapeutic interventions and were pathologically diagnosed with either anaplastic lymphoma kinase (ALK) fusions or epidermal growth factor receptor (EGFR) mutations. Radiomic features were extracted from the CT images, and appropriate feature engineering were applied to address the challenges related to high-dimensional data and class imbalance. Separate predictive models incorporating clinical data, radiomic data, and a combination of both (multimodal) were constructed and validated for predicting patient outcomes. The performance of the model was assessed based on patients' Progression-Free Survival (PFS) across different treatment regimens.</p> Results <p>The multimodal model comprising various radiomics features&#xa0;and clinical data showed good prediction performance, with an area under the curve (AUC) of 0.72 for EGFR mutation status and 0.75 for ALK fusion status. It was observed that patients receiving targeted treatment had better PFS in the first-line setting (14.5 vs 10.7/7.1&#xa0;months, <i>P</i> &lt; 0.001), but this benefit was not observed in the second-line setting (7.3 vs 4.7&#xa0;months, <i>P</i> = 0.43).</p> Conclusions <p>Multimodal data-based predictive models can accurately predict the prognosis of lung adenocarcinoma patients based on genetic alterations. Targeted therapies significantly increase PFS in non-small cell lung cancer (NSCLC) patients compared to conventional treatment alternatives in the first-line setting.</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

CT-based radiomics analysis for predicting EGFR and ALK mutational status in non‑small cell lung cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月3日

查看全文
  • Ruimin He,
  • Tengxiang Li,
  • Li Sun,
  • Peng Lu,
  • Dan Sun,
  • Rongrong Zhou,
  • Dangchi Li,
  • Xiaohua Yang,
  • Zijian Zhang

摘要

Objective

To investigate the feasibility of multimodal imaging genomics models in predicting major genetic alterations and assessing prognostic outcomes of patients with lung cancer following various treatments.

Methods

Computer tomography (CT) images of 533 patients diagnosed with lung adenocarcinoma were retrospectively analyzed together with their pretreatment demographic and clinical data. The patients received standard therapeutic interventions and were pathologically diagnosed with either anaplastic lymphoma kinase (ALK) fusions or epidermal growth factor receptor (EGFR) mutations. Radiomic features were extracted from the CT images, and appropriate feature engineering were applied to address the challenges related to high-dimensional data and class imbalance. Separate predictive models incorporating clinical data, radiomic data, and a combination of both (multimodal) were constructed and validated for predicting patient outcomes. The performance of the model was assessed based on patients' Progression-Free Survival (PFS) across different treatment regimens.

Results

The multimodal model comprising various radiomics features and clinical data showed good prediction performance, with an area under the curve (AUC) of 0.72 for EGFR mutation status and 0.75 for ALK fusion status. It was observed that patients receiving targeted treatment had better PFS in the first-line setting (14.5 vs 10.7/7.1 months, P < 0.001), but this benefit was not observed in the second-line setting (7.3 vs 4.7 months, P = 0.43).

Conclusions

Multimodal data-based predictive models can accurately predict the prognosis of lung adenocarcinoma patients based on genetic alterations. Targeted therapies significantly increase PFS in non-small cell lung cancer (NSCLC) patients compared to conventional treatment alternatives in the first-line setting.