Pasit Jakkrawankul, C. Chunharas, Wasan Akarathanawat, P. Vorasayan, Sedthapong Chunamchai, Ploy N. Pratanwanich, P. Punyabukkana, E. Chuangsuwanich
{"title":"Risk Prediction of Cardioembolic Stroke using Clinical Data and Non-contrast CT","authors":"Pasit Jakkrawankul, C. Chunharas, Wasan Akarathanawat, P. Vorasayan, Sedthapong Chunamchai, Ploy N. Pratanwanich, P. Punyabukkana, E. Chuangsuwanich","doi":"10.1109/SSP53291.2023.10207950","DOIUrl":null,"url":null,"abstract":"Cardioembolic stroke is a dangerous subtype of ischemic stroke. Patients with this subtype need special treatments to prevent recurrent events that might be fatal. Thus, identifying underlying stroke categories between cardioembolic and non-cardioembolic subtypes is of great importance. We propose a multimodal machine learning model that takes into account basic clinical information and non-contrast computed tomography (CT) images to predict the risk of cardioembolic stroke. The clinical information is not only used to provide additional information for the classification model but also to guide the attention module to extract better image features. Our model achieves a score of 0.840 using the area under the receiver operating characteristic curve (ROC-AUC) metric. Besides the capability to classify the stroke subtypes, the method can provide a heatmap for large infarct localization, which is crucial for stroke diagnosis.","PeriodicalId":296346,"journal":{"name":"2023 IEEE Statistical Signal Processing Workshop (SSP)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 IEEE Statistical Signal Processing Workshop (SSP)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/SSP53291.2023.10207950","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Cardioembolic stroke is a dangerous subtype of ischemic stroke. Patients with this subtype need special treatments to prevent recurrent events that might be fatal. Thus, identifying underlying stroke categories between cardioembolic and non-cardioembolic subtypes is of great importance. We propose a multimodal machine learning model that takes into account basic clinical information and non-contrast computed tomography (CT) images to predict the risk of cardioembolic stroke. The clinical information is not only used to provide additional information for the classification model but also to guide the attention module to extract better image features. Our model achieves a score of 0.840 using the area under the receiver operating characteristic curve (ROC-AUC) metric. Besides the capability to classify the stroke subtypes, the method can provide a heatmap for large infarct localization, which is crucial for stroke diagnosis.