Downloads & Free Reading Options - Results

A scoping review that aims to answer the question: How has machine learning, deep learning, and artificial intelligence been applied to office laryngoscopy imaging?

Title: Applications of Artificial Intelligence and Machine Learning to Office Laryngoscopy: A Scoping Review Authors: Peter Yao Alexander German Katerina Andreadis Anaïs Rameau Contact: Peter Yao [email protected] Weill Cornell Medical College Review Questions: How has machine learning, deep learning, and artificial intelligence been applied to office laryngoscopy imaging? Searches: We searched the databases of MEDLINE, EMBASE, COCHRANE, Web of Science, and IEEE Explore. We searched using a combination of database-specific subject headings and text words for the main concepts of artificial intelligence and laryngoscopy, laryngeal structures or laryngeal pathology. The search strategy was customized for each database. Types of study to be included: There are no restrictions on the methodology, the type of model developed, or publication date. Studies not written in English, were not peer reviewed, or were review papers were excluded. Reviews were excluded from the analysis, but we screened the reference lists of relevant reviews to identify potential eligible studies. Condition or domain being studied: Artificial intelligence models applied to laryngoscopic videos or images. Participants/population: Patients undergoing office laryngoscopies. Intervention: Machine learning and artificial intelligence algorithms applied to video laryngoscopy or laryngoscopy images. Comparator/control: Not applicable. Primary Outcome: Identify, appraise, and synthesize the applications of artificial intelligence to laryngoscopy including by not limited to identification, detection, and segmentation of pathology, informative frame filtering, and anatomical segmentation. Additional outcome: This review will also identify existing gaps in the literature, and help formulate the best approach for applying artificial intelligence to laryngoscopic imaging to guide existing and future researchers. Data extraction: Screening: Literature search results were imported to Covidence, an Internet-based systematic review data management software that facilitates collaboration among reviewers during the study selection process. Articles were screened using a two-step process. In the first step, articles were screened by title and abstract by teams of two reviewers working independently. In the second step, articles that passed the title-abstract screen were screened by full-text by teams of two reviewers working independently. A third reviewer reviewed articles and resolved disagreements through consensus when the original two reviewers were not able to reach one. We will calculate the adjusted kappa statistic to measure interrater agreement for eligibility screening. Data Collection We divided the items within the data collection form into four blocks: (1) study information including publication year, author information, funding or sponsorship information, type of study, journal name and PICO elements; (2) database information including data source and sample size; (3) patient demographic information including gender, age, race, and disease diagnosis; (4) ML methodological information including ML model name, type, task, data classes, class split, how ground-truth labels are determined, objective function, and model performance. Data Synthesis: A systematic narrative synthesis will be provided with information presented in the text and tables to summarize and explain the characteristics and findings of the included studies. Analysis of subgroups or subsets: We will analyze studies in groups based on their objectives. For example, studies focused on diagnosis will be analyzed separately from studies focused on anatomical segmentation.