Scholarlink

Background <p>This study employs LASSO regression and random forest algorithms to systematically investigate the factors influencing surgical difficulty in patients with advanced gastric cancer undergoing distal gastrectomy, and establishes a predictive model to assess surgical complexity.</p> Methods <p>This study enrolled patients with advanced gastric cancer who underwent distal gastrectomy at the First Affiliated Hospital of Nanchang University between January 1, 2018 and December 30, 2024. Cases were objectively classified into high-difficulty and routine-difficulty surgical groups based on threshold criteria of intraoperative blood loss (&gt; 75th percentile) or operative duration (&gt; 75th percentile). The analytical framework incorporated LASSO regression for high-dimensional data dimensionality reduction, coupled with random forest algorithms to evaluate variable importance and identify key predictive factors. Subsequently, univariate logistic regression preliminarily assessed variable associations, followed by multivariate logistic regression to determine independent determinants, ultimately establishing a visualized risk prediction model utilizing nomogram construction techniques.</p> Results <p>This study ultimately included 520 eligible cases of advanced gastric cancer undergoing radical gastrectomy. Through LASSO regression for feature dimensionality reduction and random forest-based variable importance ranking, seven core predictors were identified: BMI, prior abdominal surgery history, tumor size, and four additional clinical parameters. Univariate logistic regression preliminary screening and multivariate adjusted analysis confirmed all variables as independent risk factors for surgical difficulty (all <i>P</i> &lt; 0.05). The constructed predictive model demonstrated satisfactory discriminative ability with an AUC of 0.787, indicating clinically meaningful predictive performance.</p> Conclusion <p>Our study developed a predictive model using key clinical factors that effectively estimates surgical difficulty in advanced gastric cancer. This tool shows good accuracy (AUC 0.787) and could help surgeons plan operations better by identifying high-risk cases before surgery.</p> Graphical abstract <p></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

Development of a predictive model for surgical difficulty assessment in distal gastrectomy of advanced gastric carcinoma using LASSO-random forest integrated algorithm

  • 期刊论文

  • 发表日期：

    2025年7月21日

查看全文
  • Hongxin Yu,
  • Hualin Liao,
  • Houqiong Ju,
  • Yahang Liang,
  • Taiyuan Li,
  • Yuli Yuan

摘要

Background

This study employs LASSO regression and random forest algorithms to systematically investigate the factors influencing surgical difficulty in patients with advanced gastric cancer undergoing distal gastrectomy, and establishes a predictive model to assess surgical complexity.

Methods

This study enrolled patients with advanced gastric cancer who underwent distal gastrectomy at the First Affiliated Hospital of Nanchang University between January 1, 2018 and December 30, 2024. Cases were objectively classified into high-difficulty and routine-difficulty surgical groups based on threshold criteria of intraoperative blood loss (> 75th percentile) or operative duration (> 75th percentile). The analytical framework incorporated LASSO regression for high-dimensional data dimensionality reduction, coupled with random forest algorithms to evaluate variable importance and identify key predictive factors. Subsequently, univariate logistic regression preliminarily assessed variable associations, followed by multivariate logistic regression to determine independent determinants, ultimately establishing a visualized risk prediction model utilizing nomogram construction techniques.

Results

This study ultimately included 520 eligible cases of advanced gastric cancer undergoing radical gastrectomy. Through LASSO regression for feature dimensionality reduction and random forest-based variable importance ranking, seven core predictors were identified: BMI, prior abdominal surgery history, tumor size, and four additional clinical parameters. Univariate logistic regression preliminary screening and multivariate adjusted analysis confirmed all variables as independent risk factors for surgical difficulty (all P < 0.05). The constructed predictive model demonstrated satisfactory discriminative ability with an AUC of 0.787, indicating clinically meaningful predictive performance.

Conclusion

Our study developed a predictive model using key clinical factors that effectively estimates surgical difficulty in advanced gastric cancer. This tool shows good accuracy (AUC 0.787) and could help surgeons plan operations better by identifying high-risk cases before surgery.

Graphical abstract