Machine Learning-Based Model for Prediction of Hemorrhage Transformation in Acute Ischemic Stroke After Alteplase

 Why would your mentors and senior researchers allow you to do useless research like this rather than research that will prevent such hemorrhage transformation?

Machine Learning-Based Model for Prediction of Hemorrhage Transformation in Acute Ischemic Stroke After Alteplase

Yanan Xu¹, Xiaoli Li^2*, Di Wu², Zhengsheng Zhang² and Aizhong Jiang¹

• ¹Department of Neurology, ZhongDa Hospital Southeast University (JiangBei) (NanJing DaChang Hospital), Nanjing, China
• ²Department of Neurology, Affiliated ZhongDa Hospital, Southeast University, Nanjing, China

Hemorrhage transformation (HT) is the most dreaded complication of intravenous thrombolysis (IVT) in acute ischemic stroke (AIS). The prediction of HT after IVT is important in the treatment decision-making for AIS. We designed and compared different machine learning methods, capable of predicting HT in AIS after IVT. A total of 345 AIS patients who received intravenous alteplase between January 2016 and June 2021 were enrolled in this retrospective study. The demographic characteristics, clinical condition, biochemical data, and neuroimaging variables were included for analysis. HT was confirmed by head computed tomography (CT) or magnetic resonance imaging (MRI) within 48 h after IVT. Based on the neuroimaging results, all of the patients were divided into the non-HT group and the HT group. Then, the variables were applied in logistic regression (LR) and random forest (RF) algorithms to establish HT prediction models. To evaluate the accuracy of the machine learning models, the models were compared to several of the common scales used in clinics, including the multicenter stroke survey (MSS) score, safe implementation of treatments in stroke (SITS) score, and SEDAN score. The performance of these prediction models was evaluated using the receiver operating characteristic (ROC) curve (AUC). Forty-five patients had HT (13.0%) within 48 h after IVT. The ROC curve results showed that the AUCs of HT that were predicted by the RF model, LR model, MSS, SITS, and SEDAN scales after IVT were 0.795 (95% CI, 0.647–0.944), 0.703 (95% CI, 0.515–0.892), 0.657 (95% CI, 0.574–0.741), 0.660 (95% CI, 0.580–0.740) and 0.655 (95% CI, 0.571–0.739), respectively. The RF model performed better than the other models and scales. The top four most influential factors in the RF importance matrix plot were triglyceride, Lpa, the baseline NIHSS, and hemoglobin. The SHapley Additive exPlanation values made the RF prediction model clinically interpretable. In this study, an RF machine learning method was successfully established to predict HT in AIS patients after intravenous alteplase, which the sensitivity was 66.7%, and the specificity was 80.7%.

Introduction

Stroke is a major global public health issue with high morbidity and mortality. According to China's Ministry of Health, 77.8% of strokes are ischemic (1). Currently, poststroke outcomes are significantly improved by reperfusion therapy, such as intravenous thrombolysis and endovascular thrombectomy (2–6). Despite its efficacy, IVT also increases the risk of the development of hemorrage transformation (HT). The occurrence of HT significantly affects functional recovery and is independently associated with a higher mortality (7). It was shown in previous studies that a total of 10–48% of the AIS patients who received IVT developed HT (8) and the incidence of symptomatic intracranial hemorrhage (SICH) ranged from 2 to 7% (9). There is currently a lack of understanding of the benefits of IVT relative to its risks, which include HT, in both clinicians and patients (or their relatives). This results in patients not always receiving a timely intervention.

Previous reports have indicated that there are a number of clinical, laboratory, and radiographic factors associated with the risk of HT after alteplase. A systematic review and meta-analysis of 55 studies have reported on factors that are associated with HT, and these included age; the severity of the stroke; the baseline glucose; the use of antiplatelet drugs or statins; leukoaraiosis; early signs of infarction on head CT; the presence of atrial fibrillation, diabetes, previous ischemic heart disease or cerebral vascular diseases; and congestive heart failure (10). Lipoprotein (11), the neutrophil-to-lymphocyte ratio (NLR) (12) and the serum homocysteine level (13) were also reported as factors associated with HT after IVT. Numerous scoring systems have been devised to assess the risk for HT in AIS patients after IVT, and these include the Hemorrhage After Thrombolysis (HAT) score (14), multicenter stroke survey (MSS) score (15), safe implementation of treatments in stroke (SITS) score (16), SEDAN score (17) and GRASPS score (18). However, existing HT models cannot address the full complexity of all the factors that are involved. Each of the methods has limitations and disadvantages (14–18), which implies a certain degree of inaccuracy. The risk of HT after IVT in AIS patients cannot be avoided to the greatest extent when the stroke patients are only assessed based on the existing assessment system. Therefore, it is an urgent priority to find more reliable and effective methods to develop an early and timely prediction for HT after IVT in AIS patients.

Machine learning (ML) is a branch of artificial intelligence that can establish ideal models for classification, prediction, and estimation by allowing computers to “learn” from large, noisy or complex input and output datasets (19). In recent years, machine learning has been applied in medical research. For example, machine learning has been applied for the prediction of the outcomes in acute stroke patients (20) and endovascular treatment outcome in AIS (21). In the present study, we developed machine learning-based models that included a logistic regression (LR) model and a random forest (RF) model to predict HT. We then compared the predictability to several scales that are currently used in the clinic, including the MSS score, SITS score and SEDAN score.

 More at link.

Prediction-Driven Decision Support for Patients With Mild Stroke: A Model Based on Machine Learning Algorithms

But if you don't tell us the protocols used to recover from a mild stroke. Useless. 

Prediction-Driven Decision Support for Patients With Mild Stroke: A Model Based on Machine Learning Algorithms

Xinping Lin^1,2†, Shiteng Lin^1,2†, XiaoLi Cui^3†, Daizun Zou^1,2, FuPing Jiang⁴, JunShan Zhou⁵, NiHong Chen⁵, Zhihong Zhao^6*, Juan Zhang^3* and Jianjun Zou^2,7*

• ¹School of Basic Medicine and Clinical Pharmacy, China Pharmaceutical University, Nanjing, China
• ²Department of Clinical Pharmacology, Nanjing First Hospital, Nanjing Medical University, Nanjing, China
• ³Department of Neurology, Nanjing Yuhua Hospital, Yuhua Branch of Nanjing First Hospital, Nanjing Medical University, Nanjing, China
• ⁴Department of Geriatrics, Nanjing First Hospital, Nanjing Medical University, Nanjing, China
• ⁵Department of Neurology, Nanjing First Hospital, Nanjing Medical University, Nanjing, China
• ⁶Department of Neurology, The First Affiliated Hospital (People's Hospital of Hunan Province), Hunan Normal University, Changsha, China
• ⁷Department of Clinical Pharmacology, Nanjing First Hospital, China Pharmaceutical University, Nanjing, China

Background and Purpose: Treatment for mild stroke remains an open question.(Well then, solve that problem by creating protocols. We aim to develop a decision support tool based on machine learning (ML) algorithms, called DAMS (Disability After Mild Stroke), to identify mild stroke patients who would be at high risk of post-stroke disability (PSD) if they only received medical therapy and, more importantly, to aid neurologists in making individual clinical decisions in emergency contexts.

Methods: Ischemic stroke patients were prospectively recorded in the National Advanced Stroke Center of Nanjing First Hospital (China) between July 2016 and September 2020. The exclusion criteria were patients who received thrombolytic therapy, age <18 years, lack of 3-month modified Rankin Scale (mRS), disabled before the index stroke, with an admission National Institute of Health stroke scale (NIHSS) > 5. The primary outcome was PSD, corresponding to 3-month mRS ≥ 2. We developed five ML models and assessed the area under curve (AUC) of receiver operating characteristic, calibration curve, and decision curve analysis. The optimal ML model was selected to be DAMS. In addition, SHapley Additive exPlanations (SHAP) approach was introduced to rank the feature importance. Finally, rapid-DAMS (R-DAMS) was constructed for a more urgent situation based on DAMS.

Results: A total of 1,905 mild stroke patients were enrolled in this study, and patients with PSD accounted for 23.4% (447). There was no difference in AUCs between the five models (ranged from 0.691 to 0.823). Although there was similar discriminative performance between ML models, the support vector machine model exhibited higher net benefit and better calibration (Brier score, 0.159, calibration slope, 0.935, calibration intercept, 0.035). Therefore, this model was selected for DAMS. In addition, SHAP approach showed that the most crucial feature was NIHSS on admission. Finally, R-DAMS was constructed and there was similar discriminative performance between R-DAMS and DAMS, but the former performed worse on calibration.

Conclusions: DAMS and R-DAMS, as prediction-driven decision support tools, were designed to aid clinical decision-making for mild stroke patients in emergency contexts. In addition, even within a narrow range of baseline scores, NIHSS on admission is the strongest feature that contributed to the prediction.

Introduction

Around half of patients with ischemic stroke have mild neurological symptoms (1), usually with the expectation that such patients will come back to their pre-stroke activities regardless of the treatment. However, over one-third of mild stroke patients present with some degree of post-stroke disability (PSD) (2–4), which may be the result of inadequate acute treatments, early stroke recurrence, serious complications, or other reasons (1, 5). For the acute treatment of mild stroke patients, the guidelines from the American Heart Association/American Stroke Association (AHA/ASA) (6) distinguish disabling from non-disabling stroke and recommend intravenous (IV) alteplase only for the former. Nonetheless, the more certain, but not definitive, concept of “disabling stroke” is subjective and requires interpretation by individual neurologists. On the other hand, there is a trade-off between the benefits of IV alteplase and the risk of symptomatic intracranial hemorrhage (sICH). Therefore, decisions on how to treat mild stroke patients should be made on an individual basis.

3-month modified Rankin Scale (mRS), a valuable instrument for testing therapeutic interventions (7, 8), was used to assess the levels of PSD (5, 8). For mild stroke patients who only received medical therapy but had PSD, such therapy is not enough. Therefore, mild stroke patients who would be at high risk of PSD if they only received medical therapy should be early identified in emergency contexts, and some aggressive treatments, such as IV alteplase or close monitoring preventing worsening, should be taken in time. Unexpectedly, neurologists' overall accuracy for identifying those patients was staggeringly low (16.9%) (9). Each day that such a problem continues to exist means that uncounted mild stroke patients are being left with preventable disability.

However, none of the previously published risk models which were developed to predict the function outcome after stroke are fit to solve this problem. For example, the Totaled Health Risks in Vascular Events (THRIVE) score and the Houston Intra-Arterial Therapy (HIAT) score assign 0 points for National Institute of Health stroke scale (NIHSS) ≤ 5, losing the predictive power of NIHSS in mild stroke patients (10, 11). NIHSS on admission has been proven to be a strong predictor of PSD (5). Thus, despite convenient clinical applicability, these models cannot accurately identify mild stroke patients at high risk of PSD. Such models remain inadequate.

With the increased clinical data gathered for each patient, modern medical decision-making demands accurate, novel, and prediction-driven decision support. Machine learning (ML) algorithm, as a burgeoning statistical approach, is well-suited for that mission. Numerous studies with a considerable number of patients have shown great potential for ML approaches to predict recurrence (12), swallowing recovery (13), or aphasia (14) in patients with stroke. However, a model based on ML algorithms, focusing on the more debatable area of treating MS, has not yet been established.

Here, our goal was to develop and validate a prediction-driven decision support tool based on ML algorithms, called DAMS (Disability After Mild Stroke), to early identify mild stroke patients who would be at high risk of PSD if they only received medical therapy, and more importantly, to assist neurologists to make individual clinical decisions for mild stroke patients.