The development of postoperative hearing prediction system and the analysis of temporal bone imaging by artificial intelligence

2022 Fiscal Year Final Research Report

• Project/Area Number: 19K18787
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 56050:Otorhinolaryngology-related
• Research Institution: The University of Tokyo
• Principal Investigator: Koyama Hajime 東京大学, 医学部附属病院, 助教 (80825167)
• Project Period (FY): 2019-04-01 – 2023-03-31
• Keywords: 耳科学 / 耳科手術 / 人工知能 / 機械学習

Outline of Final Research Achievements

We applied machine learning techniques in artificial intelligence to four clinical questions: (1) postoperative hearing outcomes in chronic otitis media, (2) postoperative hearing outcomes after cochlear implantation, (3) vestibular dysfunction after pediatric cochlear implantation, and (4) mapping conditions after pediatric cochlear implantation, with the aim of predicting postoperative complications, postoperative outcomes, and setting conditions, as well as identifying factors that influence them. The results of the study were as follows. The results showed that these machine learning predictions were useful for all otologic procedures, with preoperative air-bone gaps for tympanoplasty and preoperative speech outcome with hearing aids for cochlear implant surgery being the most important predictive factors. Vestibular function after cochlear implant surgery was also found to be potentially impaired over time.

Free Research Field

耳科学

Academic Significance and Societal Importance of the Research Achievements

耳科手術において、個々の患者にとって最も有益な情報とは、手術の一般的な成功割合や合併症の発生率ではなく、患者自身の改善の程度や合併症の発生率である。個別化医療が提唱されて久しいが、外科手術における個別化医療のためには、患者個々の状態に応じた手術治療成績が必要であり、機械学習による予測はそのための貴重な情報となる。本研究により、患者個人個人の状態に応じた最適の手術の提示、およびその手術によって得られる聴力やリスクなどを、高い精度を持って個別に提供できることがわかった。