Image-based artificial intelligence for otitis media effusion diagnosis: a systematic review and meta-analysis

Abstract

Background: Otitis media with effusion (OME) is one of the most common ear conditions encountered in primary healthcare, particularly in facilities without supporting audiometry or tympanometry. OME is often misdiagnosed or delayed in treatment because it requires specialised skills to interpret otoscopic findings. Artificial Intelligence (AI) can help to simplify the diagnosis of OME so that appropriate management can be provided immediately. Purpose: To determine the performance of AI based on otoscope results in the diagnosis of OME. Method: A systematic review search based on the PRISMA flowchart was conducted in three databases (PubMed, Cochrane Library, ScienceDirect) until July 2024. Data extraction was performed to assess the sensitivity and specificity of AI for the diagnosis of OME. The risk of bias was assessed using QUADAS-2. All data were analyzed using Review Manager 5.4 and MetaDTA 2.0. Result: Twelve studies with 20,452 otoscope results were included in this review. The meta-analysis showed that AI based on otoscope results had a high sensitivity of 0.93 (95% CI; 0.87- 0.96), and a high specificity of 0.96 (95% CI; 0.94-0.98). In pediatric patients, this otoscope-based AI was also shown to diagnose OME with high sensitivity and specificity of 0.93 (95% CI; 0.64-0.99); 0.94 (95% CI; 0.86-0.97). Conclusion: AI demonstrates high sensitivity and specificity performance for the diagnosis of OME in both adult and pediatric patients. AI shows promising potential to help identify OME in primary care settings before referral for further management, and to avoid unnecessary antibiotic use. Further research is needed.