Scholarlink

Background <p>Spinal cord injury (SCI) is a severe complication after thoracoabdominal aortic repair (TAAR), substantially increasing postoperative mortality and disability. This study aimed to develop an improved machine learning model using comprehensive perioperative data to predict SCI risk.</p> Methods <p>270 patients were included in the study, and 66 potentially meaningful variables were selected from 84 original variables as input data to drive machine learning. An optimized machine learning algorithm (PSO-FLXGBoost) was applied to stratify and rank predictors, and Shapley additive explanations (SHAP) were used to interpret feature contributions.</p> Results <p>The PSO-FLXGBoost model achieved strong discrimination with an AUC of 0.895. SHAP analysis highlighted four key predictors: intraoperative hemoglobin &lt; 70&#xa0;g/L, preoperative D-dimer &gt; 6&#xa0;µg/mL, platelet count &gt; 250 × 10^9/L, and operation time &gt; 500&#xa0;min, all of which significantly increased SCI risk.</p> Conclusions <p>The optimized PSO-FLXGBoost model reliably predicts SCI after TAAR, emphasizing intraoperative hemoglobin, coagulation status, and operative duration as critical risk factors. These insights may guide tailored perioperative strategies to improve patient outcomes.</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

Prediction of spinal cord injury after thoracoabdominal aortic repair by machine learning method

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月9日

查看全文
  • Jian Song,
  • Kai Xu,
  • Jingyu Wang,
  • Tianxia Gu,
  • Wenyuan Lu,
  • Yumeng Ji,
  • Yunpeng Ling,
  • Juntao Qiu,
  • Cuntao Yu

摘要

Background

Spinal cord injury (SCI) is a severe complication after thoracoabdominal aortic repair (TAAR), substantially increasing postoperative mortality and disability. This study aimed to develop an improved machine learning model using comprehensive perioperative data to predict SCI risk.

Methods

270 patients were included in the study, and 66 potentially meaningful variables were selected from 84 original variables as input data to drive machine learning. An optimized machine learning algorithm (PSO-FLXGBoost) was applied to stratify and rank predictors, and Shapley additive explanations (SHAP) were used to interpret feature contributions.

Results

The PSO-FLXGBoost model achieved strong discrimination with an AUC of 0.895. SHAP analysis highlighted four key predictors: intraoperative hemoglobin < 70 g/L, preoperative D-dimer > 6 µg/mL, platelet count > 250 × 10^9/L, and operation time > 500 min, all of which significantly increased SCI risk.

Conclusions

The optimized PSO-FLXGBoost model reliably predicts SCI after TAAR, emphasizing intraoperative hemoglobin, coagulation status, and operative duration as critical risk factors. These insights may guide tailored perioperative strategies to improve patient outcomes.