Multimodal Data-assisted Primary Screening for Allergic Rhinitis Based on Voice Recognition and Face Recognition

Collect facial images and voice and audio of patients with rhinitis in the department of otolaryngology, and collect the examination results of patients with rhinitis who have received electronic fiber nasopharyngoscopy. Skin prick test to a standard panel of aeroallergens or by using the ImmunoCAP Phadiatop test for detecting immunoglobulin E antibodies against various common inhalant allergenswere detected, and a prediction model for the type of rhinitis was finally established.

The incidence of allergic rhinitis is high and the progression of the disease is serious, but public awareness of the disease is limited. Mistaking allergic rhinitis for the common cold or other respiratory illnesses and purchasing non-specific medications for its treatment not only delays proper diagnosis and treatment, but may also lead to further aggravation of the disease and complications. Such omission, misdiagnosis and mistreatment of allergic rhinitis not only affects the management and control of the disease, but may also result in unnecessary wastage of healthcare resources and increased treatment costs. In this study, the investigators propose to capture face photographs and audio files of rhinitis patients coming to the otolaryngology clinic using a work cell phone to determine whether the patients are allergic or non-allergic rhinitis by using an allergy detection test. The face photos, audio files and basic clinical information were multimodally fused to construct a prediction model, and the effectiveness of the model was evaluated. Ultimately, it is expected that the predictive model can simply identify and screen for allergic rhinitis, improve public awareness and understanding of allergic rhinitis, and take proper treatment measures.