The present study strived to test the feasibility of the implementation of the Word2Vec AI model in the optimization of the acute management of MIS starting from the suspected diagnosis during patients’ triage. The model identified 15 text words highly predictive of MIS and it proved to be highly effective in the prompt and accurate diagnosis with a rate of false negatives of 0%. The model has been confirmed to be even more accurate than the conventional color code triage classification model, which is in use in all the Italian ED nowadays. More than 80% of strokes result from ischemic damage to the brain due to an acute reduction in the blood supply. The goal in the management of acute ischemic stroke is early arterial recanalization to limit the brain damage since the delay in starting the treatment is associated to worst physical e cognitive outcome, with a high level of disability and comorbidities [2, 15, 16]. Although faster triage, improvements in neuroimaging techniques, thrombolysis, and thrombectomy represent the major advances of MIS management, the overall outcome of patients affected by stroke is still largely dependent on a prompt and accurate diagnosis at admission at the ED [12, 17–22].
Based on our results, the presence of one of the 15 keywords identified by the proposed AI model is associated with a rapid diagnosis of stroke and the performance on the test set shows that stroke patients were successfully identified with a recall of 87.69% and an AUC of 97.2. Dysarthria and aphasia were the text words most importantly correlated with the stroke diagnosis. Noteworthy, the model was able to correctly associate with a suspected diagnosis of stroke also those misspelled text words that were accidentally recorded during the triage. “Disatria” instead of “disartria”, namely dysarthric speech, was an example. Because of a documented lower rate of false negatives, the Word2Vec AI model has proved to be more accurate than the color-code based classification model in-use in our department as many other hospitals in north Europe. The practical implication of this model in daily practice are non-negligible since it may contribute to the optimization of the acute management of patients affected by MIS. In a combined vision where the AI models are integrative rather than substitutive of the human resources, the availability of a computer alert generated by the AI algorithm may be of help to nurses and others to early recognize those patients suspected to be affected by ischemic stroke. Further AI algorithms like that reported in the present study may also be adopted for the hemorrhagic stroke.
One Hot Encoding and Word Embedding are two of the most popular concepts for vector representation in Natural Language Processing. Word2vec is an algorithm created in 2013 that uses a neural network model to identify words associated starting from a big matrix of data set and, once trained, it can select words with similar meaning from the words surrounding it. It represents each word identified by a list of numbers called vectors. The vectors are selected with a simple mathematical function and share a certain level of semantic similarity between the words associated with those vectors [23].
The choice of Word2vec embedding-based AI algorithm lets us work on a big volume of data in a simple way. This algorithm selected words with intrinsic meaning, starting with a numeric vector obtained from a dependent variable. From the numeric vector (whose length is about 300, established by our team) we process data with a statistic model that can interpret artificial neural networks obtained by using Word2vec algorithm.
Another algorithm that could be used because of the easiness of implementation is “One hot encoding”, working in a faster way than Word 2 embedding: every word has its own value in a vector, but in this process, it loses the semantic meaning of the word in a sentence. One hot encoding was one of the first techniques used in artificial intelligence models but with the birth of Word-embedding, it becomes obsolete, especially in scientific fields. Furthermore, by using a one-hot encoding algorithm the size of the embedding vector grows with the vocabulary so it could be difficult to elaborate those data because of the entity of the matrix of embedding obtained, so it doesn't work well in applications that require a large amount of data. Word2vec with its implementation could be a good middle ground even because the precision of word embedding depends on the volume of the dataset, so it works well on large datasets obtaining the best word embedding with the smallest matrix.
Other models of Artificial Intelligence include GloVe and FastText.
With Word2vec we train a neural network with a single hidden layer to predict a target word based on its context. With FastText each word is composed of character n-gram so it can help to generate better word embeddings for rare words or for out of vocabulary words; a big limit of this algorithm is that it takes longer to do the embedding and as the dataset grows, the memory required grows too, so in this way is similar to One Hot Encoding.
The gloVe is a word embedding technique similar to Word2vec, but it differs from it because it is a count-based model instead of the predictive model. In fact, GloVe focuses on words co-occurrences over the whole corpus, while Word2vec leverages co-occurrence within local context (neighboring words). Glove embeddings relate to the probability that two words appear together.
Word embedding techniques, with respect to count-based methods, on different language tasks such as semantic relatedness, synonym detection, concept categorization, and analogy. With Word2vec we observe large improvements in accuracy at much lower computational cost, i.e. it takes less than a day to learn high quality.
As reported, the need for continuous training of the AI model, by means of the increase of the data collected from other clinical studies, is a key aspect for the further improvement and optimization of the model itself [24, 25].
Limitations of the Study
The first limitation of the present study lies in the exclusion of hemorrhagic stroke or TIA, considering only MIS.
Furthermore, this word embedding-based AI model didn't explore the Vital Signs, which are extremely useful to detect the critical issues of the patient. Using Word2vec we obtained the classification of a word strongly associated with MIS in terms of clinical features, but this algorithm does not work on the definite diagnosis of the disease. With AI models it would be easy to create a warning signal with those “embedded words”, popping up on computers of triage's nurses, but the meaning of that “alert” must be evaluated according to the cases. For example, one of the words most associated with stroke diagnosis according to the model Word2vec is “disorientation” but only in few cases this clinical feature is observed in those patients. Another limitation of the algorithm is that the detection of true positive cases is not well balanced by the identification of true negative rates. It could overestimate the real impact of the disease in triage.
With Word2Vec the word embedding obtained by using the algorithm is “static”, which means that the model has no awareness of the framework in which the word is found. By using recurrent neural networks, the word embedding could become more dynamic and accurate: this new AI model is able to detect the hidden relationship between inputs as well as to provide a precise sequence prediction of words, giving a high accuracy to results.
Future perspectives could involve dynamic models of word embedding, more refined. In fact, while working on recurrent neural networks the word embedding will help us to obtain more precise results even on false-negative cases, taking all the vectors generated by the algorithm with new technology.