Research Question, objective and context
The topics of interest in this study are artificial intelligence and swallowing analysis. The guiding research question is: What are the AI-based applications for swallowing analysis described in the literature? The question encompasses the concept of AI and strategies applied to swallowing assessment. The objectives are to map the AI-based applications for swallowing analysis described in the literature, describe the forms of swallowing assessment found and point out possible gaps in the literature.
The pre-planned inclusion specifications are what the study contains as a basis for analyzing AI-based swallowing applications.
Studies that do not compare scientific applications in artificial intelligence with swallowing videofluoroscopy. In the study analysis process, other exclusion criteria can be defined.
This work is focused on studies that carried out swallowing assessments with AI-based applications and compared these results to swallowing videofluoroscopy exams in individuals over 18 years of age, regardless of the pathology.
Search and selection of studies
This review work will follow the five steps described by Ashley and O'Malley: (1) identify the research question, (2) identify relevant studies, (3) study selection, (4) data collection, (5) mapping , summary and description of results. To construct the research question, we used an adaptation of the PECO strategy (acronym for: population, exposure, control and development/result), where we used to construct the research question only “P” corresponding to adults, “E” corresponding to swallowing and desphagia and “O” to specific applications in artificial intelligence.
The health descriptors used to reference artificial intelligence applications will be: Deep Learning, Hierarchical Learning, Neural Networks, Computer, Model, Neural Network, Computational Neural Network, Perceptron, Connectionist Models, Support Vector Machine, Support Vector Network, Machine Learning, Transfer Learning, Artificial Intelligence, Computational Intelligence, Machine Intelligence, Computational Reasoning, Computer Vision System, Knowledge Acquisition (Computer), Supervised Machine Learning, Unsupervised Machine Learning, and those used for swallowing and dysphagia will be: Disorders Swallowing, Swallowing, swallowing, swallowing, swallowing, oropharynx, oropharyngeal, problem, disorder, impairment, difficult, Swallowing Disorders, Swallowing Disorders, Dysphagia, Oropharyngeal Dysphagia, Swallowing. Using the Boolean operators “AND” and “OR”, the search will be carried out in the Embase, Pubmed, Web Of Science and Scopus databases, without delimiting publication data. Additionally, a search was conducted on the gray literature using the Google Scholar search engine.
To identify articles potentially eligible for the study, titles and abstracts will be read and detailed and selected studies will be read in full by two independent reviewers (MT and RSR), when relevant, data will be extracted. For extra information, we can contact the authors of the included studies. Possible disagreements during the article selection phase can be resolved through discussion meetings. In case of non-agreement, a third reviewer will be consulted (FNH). To organize the screening and extraction of data, reference manager software will be used.
Extracting and mapping the data
Data removal will occur in the excel file. The data extracted from the works will be: Title, author, year of publication, periodical of publication, country, objective, method (application of artificial intelligence used, swallowing assessment method, sample characteristics, variables explored) and results. The studies will be classified according to the two main pillars of this review: the artificial intelligence applications used and the swallowing assessment methods.
Summary and report of results
The results of the selected studies will be summarized. To analyze the results, descriptive statistics will be performed. The data from the selected studies will be organized according to the characteristics of the subjects, underlying pathologies, application of artificial intelligence used and aspects evaluated in swallowing. The prism extension [15] will be used in this review. An article will be published with the results of this study in a peer-reviewed journal.