Word2Vec Word Embedding-Based Artificial Intelligence Model in the Triage of Patients with Suspected Diagnosis of Major Ischemic Stroke: A Feasibility Study

Abstract

Abstract IntroductionThe possible benefits of using semantic language models in the early diagnosis of major ischemic stroke (MIS) based on artificial intelligence (AI) are still underestimated.The present study strives to assay the feasibility of the Word2Vec word embedding-based AI model in decreasing the risk of false-negatives during the triage of patients with suspected MIS in the emergency department (ED).MethodsThe main ICD-9 codes related to MIS were used for a 7-years retrospective data collection of patients managed at the ED with a suspected diagnosis of stroke. Data underwent “tokenization” and “lemmatization”. Word2Vec word embedding AI algorithm was used for text data vectorization and a batch strategy was adopted for model training.The false-negatives were rated from the incorrect attribution of the low-intensity code to patients later diagnosed with MIS. ResultsOut of 649 MIS, the Word2Vec AI algorithm allowed successfully identify 87.69% of them through the recognition of 15 top words, with an area under the curve of 97.2%. The rate of false-negatives related to the implementation of AI model was 0%.ConclusionsWord2Vec word embedding-based AI model is reliable and effective in decreasing the risk of false negatives of MIS during patients’ triage in the ED. Further studies on larger cohorts are required to validate the model.