Diagnosing Influenza Infection from Pharyngeal Images using Deep Learning: Machine Learning Approach (Preprint)

Background Influenza is a major global burden of disease, causing annual epidemics and occasionally, pandemics. Given that influenza primarily infects the upper respiratory system, influenza infection may be able to be diagnosed by applying deep learning to pharyngeal images. Objective We aimed to develop a deep learning model to diagnose influenza infection using the data on pharyngeal images and clinical information. Methods We recruited patients who visited clinics and hospitals due to influenza-like symptoms. In the training stage, we developed a diagnostic prediction artificial intelligence (AI) model based on deep learning to predict polymerase chain reaction (PCR)-confirmed influenza from pharyngeal images and clinical information. In the validation stage, we assessed the diagnostic performance of the AI model. In the additional analysis, we compared the diagnostic performance of the AI model with that of three physicians, and also interpreted the AI model using the importance heatmaps. Results A total of 7,831 patients were enrolled at 64 hospitals between Nov 1, 2019 and Jan 21, 2020 in the training stage, and 659 patients (including 196 patients with PCR-confirmed influenza) at 11 hospitals between Jan 25, 2020 and Mar 13, 2020 in the validation stage. The area under the receiver operating characteristic curve of the AI model was 0.90 (95% confidence interval, 0.87–0.93), and its sensitivity and specificity were 76% (70–82%) and 88% (85–91%), respectively, outperforming three physicians. In the importance heatmaps, the AI model often focused on follicles on the posterior pharyngeal wall. Conclusions We developed the first AI model that can accurately diagnose influenza from pharyngeal images, which has the potential to assist physicians make timely diagnosis. ### Competing Interest Statement SO is the CEO and HK is the CSO of Aillis, Inc. and they hold stock in the company. MF, MS, WT, MIk, and HK are employees of Aillis, Inc. YT and MIw received consultant fees from the company to supervise the study and draft the manuscript. ### Funding Statement This study was funded by the New Energy and Industrial Technology Development Organization (NEDO), Japan (research number 30STS713) and Aillis, Inc.. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Certified Review Board, Hattori Clinic Clinical Research Ethics Board, Haradoi Hospital, Social Medical Corporation Kobori Central Clinical Research Ethics Committee Medical corporation Takahashi clinic clinical trial examination committee I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable. Yes The data used in this study, mainly pharyngeal images, are licensed to Aillis, Inc.. The data have not been opened in public, and could be used for future projects for the development of medical devices and diagnostic technologies. Proposals and requests for data access should be directed to the corresponding author via email.