Scholarlink

Background <p>Guideline-recommended algorithms (GL-algorithm) often results in indeterminate left ventricular filling pressure (LVFP). Despite high accuracy, machine learning (ML) methods lack interpretability, which necessitates the development of explainable ML models for clinical use.</p> Objective <p>To develop an explainable ML model for estimating LVFP, providing patient-level interpretation using gold-standard right heart catheterization (RHC) data.</p> Methods <p>We retrospectively enrolled 956 patients who underwent echocardiography and RHC at three hospitals within a median of 3&#xa0;days. Two extreme gradient boosting models were trained using data from two hospitals (n = 621) to estimate elevated pulmonary artery wedge pressure (PAWP ≥ 18&#xa0;mmHg) as a surrogate for elevated LVFP. Model 1 used variables from GL-algorithm, while Model 2 used variables selected based on Shapley additive explanations (SHAP) values. Models’ area under the receiver-operating characteristic curve (AUROC) for elevated LVFP were compared using external test data from the other hospital (n = 335).</p> Results <p>Overall, 31.0% had elevated PAWP, and 42.7% were classified as indeterminate LVFP by GL-algorithm, whereas the ML models classified all patients. AUROCs of Model 1 (0.82, 95% CI 0.73–0.92) and Model 2 (0.83, 95% CI 0.75–0.91) in classifiable cases by GL-algorithm significantly outperformed that of GL-algorithm (0.72, 95% CI 0.60–0.83, p = 0.020 and 0.016, respectively), Model 2 performed equally well for indeterminate cases. SHAP force plots visualized each variable’s contribution to the ML model’s decision for each patient.</p> Conclusions <p>Explainable ML outperformed GL-algorithm in estimating LVFP, providing a user-friendly tool for clinicians with patient-level interpretability.</p> Graphical abstract <p>Machine learning models improve estimation of elevated PAWP: Using XGBoost, ML models to estimate PAWP elevation from the echocardiographic parameters were developed. Compared with conventional algorithm, the ML models predicted PAWP elevation more accurately. By interpreting the impact of each parameter on the prediction using SHAP, the importance of each parameter for prediction could be understood for each patient and for the entire population. <i>AUROC</i> area under the receiver-operating characteristic curve, <i>CI</i> confidence interval, <i>LAVI</i> left atrial volume index, <i>ML</i> machine learning, <i>PAWP</i> pulmonary artery wedge pressure, <i>RHC</i> right heart catheterization, <i>SHAP</i> shapley additive explanations, <i>TRV</i> tricuspid regurgitation pressure gradient, <i>TTE</i> transthoracic echocardiography, <i>XGBoost</i> extreme gradient boosting</p> <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

Explainable machine learning for estimation of elevated left ventricular filling pressure: a multicenter validation

  • 期刊论文

  • 发表日期：

    2026年5月29日

查看全文
  • Yutaka Nakamura,
  • Nobuyuki Kagiyama,
  • Sirish Shrestha,
  • Yuka Akama,
  • Shusuke Kito,
  • Sachiko Takamatsu,
  • Misako Toki,
  • Takeshi Kitai,
  • Taiji Okada,
  • Azusa Murata,
  • Tomohiro Kaneko,
  • Yuya Matsue,
  • Sakiko Miyazaki,
  • Shinya Okazaki,
  • Tohru Minamino

摘要

Background

Guideline-recommended algorithms (GL-algorithm) often results in indeterminate left ventricular filling pressure (LVFP). Despite high accuracy, machine learning (ML) methods lack interpretability, which necessitates the development of explainable ML models for clinical use.

Objective

To develop an explainable ML model for estimating LVFP, providing patient-level interpretation using gold-standard right heart catheterization (RHC) data.

Methods

We retrospectively enrolled 956 patients who underwent echocardiography and RHC at three hospitals within a median of 3 days. Two extreme gradient boosting models were trained using data from two hospitals (n = 621) to estimate elevated pulmonary artery wedge pressure (PAWP ≥ 18 mmHg) as a surrogate for elevated LVFP. Model 1 used variables from GL-algorithm, while Model 2 used variables selected based on Shapley additive explanations (SHAP) values. Models’ area under the receiver-operating characteristic curve (AUROC) for elevated LVFP were compared using external test data from the other hospital (n = 335).

Results

Overall, 31.0% had elevated PAWP, and 42.7% were classified as indeterminate LVFP by GL-algorithm, whereas the ML models classified all patients. AUROCs of Model 1 (0.82, 95% CI 0.73–0.92) and Model 2 (0.83, 95% CI 0.75–0.91) in classifiable cases by GL-algorithm significantly outperformed that of GL-algorithm (0.72, 95% CI 0.60–0.83, p = 0.020 and 0.016, respectively), Model 2 performed equally well for indeterminate cases. SHAP force plots visualized each variable’s contribution to the ML model’s decision for each patient.

Conclusions

Explainable ML outperformed GL-algorithm in estimating LVFP, providing a user-friendly tool for clinicians with patient-level interpretability.

Graphical abstract

Machine learning models improve estimation of elevated PAWP: Using XGBoost, ML models to estimate PAWP elevation from the echocardiographic parameters were developed. Compared with conventional algorithm, the ML models predicted PAWP elevation more accurately. By interpreting the impact of each parameter on the prediction using SHAP, the importance of each parameter for prediction could be understood for each patient and for the entire population. AUROC area under the receiver-operating characteristic curve, CI confidence interval, LAVI left atrial volume index, ML machine learning, PAWP pulmonary artery wedge pressure, RHC right heart catheterization, SHAP shapley additive explanations, TRV tricuspid regurgitation pressure gradient, TTE transthoracic echocardiography, XGBoost extreme gradient boosting