Akademik Veri Yönetim Sistemi
Indian Journal of Otolaryngology and Head and Neck Surgery, 2026 (ESCI, Scopus)
The aim of this study is to systematically evaluate the diagnostic accuracy of ‘machine learning’ (ML) and ‘deep learning’ (DL) algorithms that classify tinnitus patients by analysing their electroencephalography (EEG) data, and to map the objective neurophysiological biomarkers that the models identify. This study was designed and conducted in accordance with the PRISMA 2020 guidelines. The PubMed, Web of Science (WoS), IEEE Xplore and ScienceDirect databases were searched for studies published between January 2010 and January 2026. Following the review, 24 studies with confirmed methodological adequacy were included in the analysis. The quality of the studies was assessed using the QUADAS-2 tool. A meta-analysis demonstrated that EEG data has great potential for diagnosing tinnitus using artificial intelligence algorithms. Analysis of the included studies revealed an accuracy rate of 86.2%, and the area under the receiver operating characteristic curve (AUC) was found to be 0.878. Furthermore, models built using deep learning (DL) architectures were found to be more accurate (89.0%) than those built using traditional machine learning models (83.7%). Meta-regression findings showed that different results could be obtained depending on the level at which the EEG data was input into the model. It was found that ‘sensor-level’ data improved model accuracy, while ‘source-level’ data enhanced recall performance. Converting EEG data obtained from patients with subjective tinnitus into objective data using artificial intelligence algorithms could open up new, personalized diagnostic and treatment options for tinnitus—a condition for which treatments have long been difficult to generalize.