Scholarlink

Objective <p>To evaluate the predictive performance of multiple machine-learning algorithms in forecasting postoperative air-bone gap following tympanoplasty and to determine the relative importance of clinical prognostic variables.</p> Methods <p>This study was performed on patients who underwent tympanoplasty at a tertiary referral center. Preoperative factors were collected, including age, sex, smoking history, otorrhea, preoperative air-bone gap (ABG), contralateral ear status, diabetes mellitus (DM), and hypertension (HTN). The target variable was postoperative ABG prediction. The cohort consisted of four years of retrospective data followed by one year of prospective enrollment. Six supervised machine-learning algorithms were applied to predict postoperative hearing outcomes.</p> Results <p>A total of 439 cases were included in the analysis, with a mean age of 34.6 years. A development dataset of 366 tympanoplasty cases was analyzed, and an additional 73 cases were reserved for model validation. The Random Forest classifier achieved the highest predictive accuracy (93%) on the test dataset. Analysis of relative feature importance demonstrated that age, preoperative air-conduction threshold, and preoperative ABG were the primary determinants of postoperative hearing outcome.</p> Conclusion <p>Machine-learning algorithms showed promising accuracy in predicting tympanoplasty success. In particular, the Random Forest model demonstrated superior performance. Incorporating AI-driven decision-support tools in otologic surgery may facilitate patient optimization and improve surgical planning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Utility of artificial intelligence in prediction of hearing outcome after tympanoplasty

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月28日

查看全文
  • Zakaria Mesalam,
  • Abeer Twakol Khalil,
  • Mohamed Elkahwagi,
  • Waleed Moneir

摘要

Objective

To evaluate the predictive performance of multiple machine-learning algorithms in forecasting postoperative air-bone gap following tympanoplasty and to determine the relative importance of clinical prognostic variables.

Methods

This study was performed on patients who underwent tympanoplasty at a tertiary referral center. Preoperative factors were collected, including age, sex, smoking history, otorrhea, preoperative air-bone gap (ABG), contralateral ear status, diabetes mellitus (DM), and hypertension (HTN). The target variable was postoperative ABG prediction. The cohort consisted of four years of retrospective data followed by one year of prospective enrollment. Six supervised machine-learning algorithms were applied to predict postoperative hearing outcomes.

Results

A total of 439 cases were included in the analysis, with a mean age of 34.6 years. A development dataset of 366 tympanoplasty cases was analyzed, and an additional 73 cases were reserved for model validation. The Random Forest classifier achieved the highest predictive accuracy (93%) on the test dataset. Analysis of relative feature importance demonstrated that age, preoperative air-conduction threshold, and preoperative ABG were the primary determinants of postoperative hearing outcome.

Conclusion

Machine-learning algorithms showed promising accuracy in predicting tympanoplasty success. In particular, the Random Forest model demonstrated superior performance. Incorporating AI-driven decision-support tools in otologic surgery may facilitate patient optimization and improve surgical planning.