Scholarlink

<p>To develop and validate a routine indicator-based multimarker panel for fungal infection (FI) diagnosis and prognosis prediction. A retrospective cohort of 134 patients (63 FI, 71 non-FI) was stratified randomly into a training set (<i>n</i> = 75) and validation set (<i>n</i> = 59). Differential indicators were screened via non-parametric tests and univariate logistic regression, with LASSO regression verifying core variables. Four machine learning algorithms (logistic regression, SVM, GNB, LightGBM) were compared via 5-fold cross-validation. The model with the best performance was validated in the validation set. Six core indicators (hsCRP, BNP, PT%, D-Dimer, IL-17, PCT) were selected. The logistic regression algorithm was finally selected for modeling because of its optimal efficacy in the training set. The logistic-based model in validation set showed better performance than others. The diagnostic AUC was 0.845, sensitivity was 80.00%, specificity was 72.41%. The prognostic AUC was 0.767, sensitivity was 72.41% and specificity was 90.00%, outperforming single indicators. The logistic regression-derived multimarker panel integrates inflammatory, coagulation, and cardiac biomarkers, offering a reliable, accessible tool for FI diagnosis and prognostic stratification, especially in resource-limited settings.</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

A multimarker panel for diagnosis and prognosis prediction of fungal infections

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月21日

查看全文
  • Yunyan Lou,
  • Fanghao Yu,
  • Yanting Zhao,
  • Jianxin He

摘要

To develop and validate a routine indicator-based multimarker panel for fungal infection (FI) diagnosis and prognosis prediction. A retrospective cohort of 134 patients (63 FI, 71 non-FI) was stratified randomly into a training set (n = 75) and validation set (n = 59). Differential indicators were screened via non-parametric tests and univariate logistic regression, with LASSO regression verifying core variables. Four machine learning algorithms (logistic regression, SVM, GNB, LightGBM) were compared via 5-fold cross-validation. The model with the best performance was validated in the validation set. Six core indicators (hsCRP, BNP, PT%, D-Dimer, IL-17, PCT) were selected. The logistic regression algorithm was finally selected for modeling because of its optimal efficacy in the training set. The logistic-based model in validation set showed better performance than others. The diagnostic AUC was 0.845, sensitivity was 80.00%, specificity was 72.41%. The prognostic AUC was 0.767, sensitivity was 72.41% and specificity was 90.00%, outperforming single indicators. The logistic regression-derived multimarker panel integrates inflammatory, coagulation, and cardiac biomarkers, offering a reliable, accessible tool for FI diagnosis and prognostic stratification, especially in resource-limited settings.