Scholarlink

Background <p>Secondary mitral regurgitation (SMR) is a common condition, with typical treatment options including conventional medical therapy and the latest transcatheter edge-to-edge repair. However, there is currently a lack of sufficient guidance on how to select the appropriate treatment approach for SMR patients. SMR is a common condition secondary to ischemic or non-ischemic heart failure, surgical or interventional therapy is preserved for those patients who showed no improvement after tailored guideline-directed medical therapy (GDMT) for heart failure.</p> Objective <p>Our goal is to develop an interpretable model to predict SMR improvement after medical treatment in the real-world, i.e. to find those patients who are candidates for surgical therapy. We employ the SHapley Additive exPlanations (SHAP) method to interpret the model and explore factors influencing improvement.</p> Methods <p>In this retrospective study, we extracted data on patients diagnosed with moderate or greater secondary mitral regurgitation. Predictive models were constructed, with the dataset randomly divided such that 80% was used for model training and 20% for model validation. We compared the predictive performance of the XGBoost model with three other machine learning models by analyzing the area under the curve (AUC). The XGBoost model was interpreted using the SHAP method, with variable importance assessed and the impact of key variables evaluated accordingly.</p> Results <p>The study ultimately included 1,572 eligible SMR patients. The XGBoost model demonstrated the highest predictive performance among the four models, with an AUC of 0.73. The SHAP method identified and ranked the top 20 predictive factors for SMR, with age being the most significant predictor.</p> Conclusion <p>We have developed an effective model to predict whether SMR patients will benefit from non-surgical treatment, which can help clinicians formulate better treatment plans and achieve optimal resource allocation. Additionally, the interpretable framework enhances the transparency of the model and helps doctors understand the reliability of the predictive model.</p> Clinical trial registration <p>Clinical trial number: not applicable.</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 predictive model for the treatment outcomes of patients with secondary mitral regurgitation based on machine learning and model interpretation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月22日

查看全文
  • Keyi Liu,
  • Ting Liu,
  • Yumen Lang,
  • Qing Zhang

摘要

Background

Secondary mitral regurgitation (SMR) is a common condition, with typical treatment options including conventional medical therapy and the latest transcatheter edge-to-edge repair. However, there is currently a lack of sufficient guidance on how to select the appropriate treatment approach for SMR patients. SMR is a common condition secondary to ischemic or non-ischemic heart failure, surgical or interventional therapy is preserved for those patients who showed no improvement after tailored guideline-directed medical therapy (GDMT) for heart failure.

Objective

Our goal is to develop an interpretable model to predict SMR improvement after medical treatment in the real-world, i.e. to find those patients who are candidates for surgical therapy. We employ the SHapley Additive exPlanations (SHAP) method to interpret the model and explore factors influencing improvement.

Methods

In this retrospective study, we extracted data on patients diagnosed with moderate or greater secondary mitral regurgitation. Predictive models were constructed, with the dataset randomly divided such that 80% was used for model training and 20% for model validation. We compared the predictive performance of the XGBoost model with three other machine learning models by analyzing the area under the curve (AUC). The XGBoost model was interpreted using the SHAP method, with variable importance assessed and the impact of key variables evaluated accordingly.

Results

The study ultimately included 1,572 eligible SMR patients. The XGBoost model demonstrated the highest predictive performance among the four models, with an AUC of 0.73. The SHAP method identified and ranked the top 20 predictive factors for SMR, with age being the most significant predictor.

Conclusion

We have developed an effective model to predict whether SMR patients will benefit from non-surgical treatment, which can help clinicians formulate better treatment plans and achieve optimal resource allocation. Additionally, the interpretable framework enhances the transparency of the model and helps doctors understand the reliability of the predictive model.

Clinical trial registration

Clinical trial number: not applicable.