Interpretable prediction of stroke prognosis: SHAP for SVM and nomogram for logistic regression

 ABSOLUTELY FUCKING USELESS! Predictions/'assessments' DO NOTHING to get survivors recovered!

Interpretable prediction of stroke prognosis: SHAP for SVM and nomogram for logistic regression

Kun Guo^1,2Bo Zhu¹Lei Zha¹Yuan Shao²Zhiqin Liu¹Naibing Gu^1*Kongbo Chen^2*

• ¹Xi’an Central Hospital, Xi’an, China
• ²Tongchuan Mining Bureau Central Hospital, Tongchuan, China

Background: Ischemic Stroke (IS) stands as a leading cause of mortality and disability globally, with an anticipated increase in IS-related fatalities by 2030. Despite therapeutic advancements, many patients still lack effective interventions, underscoring the need for improved prognostic assessment tools.(Wow, are you off base with your idea of needing assessments.) Machine Learning (ML) models have emerged as promising tools for predicting stroke prognosis, surpassing traditional methods in accuracy and speed.

Objective: The aim of this study was to develop and validate ML algorithms for predicting the 6-month prognosis of patients with Acute Cerebral Infarction, using clinical data from two medical centers in China, and to assess the feasibility of implementing Explainable ML in clinical settings.

Methods: A retrospective observational cohort study was conducted involving 398 patients diagnosed with Acute Cerebral Infarction from January 2023 to February 2024. The dataset included demographic information, medical histories, clinical evaluations, and laboratory results. Six ML models were constructed: Logistic Regression, Naive Bayes, Support Vector Machine (SVM), Random Forest, XGBoost, and AdaBoost. Model performance was evaluated using the Area Under the Receiver Operating Characteristic curve (AUC), sensitivity, specificity, predictive values, and F1 score, with five-fold cross-validation to ensure robustness.

Results: The training set, identified key variables associated with stroke prognosis, including hypertension, diabetes, and smoking history. The SVM model demonstrated exceptional performance, with an AUC of 0.9453 on the training set and 0.9213 on the validation set. A Nomogram based on Logistic Regression was developed for visualizing prognostic risk, incorporating factors such as the National Institutes of Health Stroke Scale (NIHSS) score, Barthel Index (BI), Watanabe Drinking Test (KWST) score, Platelet Distribution Width (PDW), and others. Our models showed high predictive accuracy and stability across both datasets.

Conclusion: This study presents a robust ML approach for predicting stroke prognosis, with the SVM model and Nomogram providing valuable tools for clinical decision-making. By incorporating advanced ML techniques, we enhance the precision of prognostic assessments and offer a theoretical and practical framework for clinical application.

1 Introduction

Ischemic stroke (IS) is a leading cause of mortality and disability worldwide, with a stark increase in global IS-related deaths reaching 3.29 million in 2023, and projections estimate a rise to 4.9 million by 2030 (1). The rapid aging and industrialization of societies, along with the spread of unhealthy lifestyle and dietary habits, have made IS the primary cause of death and disability among adults in China (2, 3). Despite advancements in the management, treatment, and prevention of IS, many patients still lack effective interventions. The prognosis of stroke is a complex process, influenced by a multitude of factors (4, 5). Timely determination of prognosis is crucial for physicians to adjust intervention strategies, prevent recurrence, ascertain adverse outcomes, and provide precision treatment plans (6). However, traditional statistical methods have limitations in terms of prognostic accuracy. The advent of machine learning has shown immense promise in handling large datasets and identifying complex patterns, offering a new horizon in the assessment of stroke prognosis.

Machine learning (ML) models have garnered attention for their prowess in handling vast datasets and discerning complex patterns (7). In the realm of prognosis assessment, these models swiftly identify independent predictive factors associated with adverse outcomes (8). The precision of these models is enhanced through a rigorous evaluation that encompasses a spectrum of metrics, including accuracy, recall, F1 score, Area Under the Receiver Operating Characteristic curve (AUC), and Shapley Additive explanations (SHAP) values. This approach not only surpasses traditional predictive methods in terms of speed but also demonstrates increasing accuracy in practical clinical applications (9).

For instance, a study optimized Principal Component Analysis (PCA) and integrated models such as Random Forest, Decision Trees, and K-Nearest Neighbors (KNN) to achieve a remarkable 98.6% accuracy rate in stroke prediction (10). Another investigation harnessed machine learning to develop a risk stratification model based on data from patients with acute ischemic stroke (AIS), which exhibited excellent predictive power as assessed by AUC values (11). Furthermore, research employing diverse machine learning algorithms to predict 90-day outcomes in stroke patients identified the Random Forest model as the ultimate predictor, with the highest AUC value.

While the performance of machine learning models is contingent upon the quality and completeness of the input data, and challenges regarding dataset representativeness and model generalizability persist, the prospects for their application in stroke prognosis assessment remain promising (12). The aim of this study is to harness clinical data to predict the 6-month prognosis of patients with cerebral infarction using machine learning algorithms and to evaluate the feasibility of explainable machine learning in clinical practice, thereby providing theoretical and practical support for its clinical application.

More at link.