Scholarlink

<p>In this study, we present an innovative approach in medicine, leveraging the potential of artificial intelligence (AI) for the diagnosis of tympanic membrane perforations. Due to its central role in sound transmission, the tympanic membrane is a key element in clinical assessment and therapeutic decision-making. Our research focuses on the automated detection of tympanic membrane perforations using advanced AI algorithms, capable of accurately identifying whether a perforation is marginal or not. This study evaluates the performance of a convolutional neural network (CNN) optimized by a genetic algorithm (GA) for the automated classification of tympanic membrane perforations. A dataset of 6000 otoendoscopic images, collected between January 2021 and March 2025 at the University Hospital Center of Oujda, Morocco, was used to train and validate the model. The images were classified into four categories: normal, non-marginal punctate perforation, subtotal perforation, and marginal perforation. The proposed GA-CNN model achieved 96.2% accuracy, 95.7% correctness, 95.9% recall, and an F1-score of 95.8%, surpassing standard architectures such as VGG19, ResNet50, InceptionV3, and EfficientNetV2-B3. These results demonstrate the potential of the proposed CNN as a clinical decision support tool for the objective and reproducible assessment of tympanic perforations.</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

Artificial Intelligence at the Service of the ENT: A New Perspective on Tympanic Membrane Perforations

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Issam Berrajaa,
  • Achraf Berrajaa,
  • Achraf Amine Sbai,
  • Imane Demnati,
  • Drissia Benfadil,
  • Azeddine Lachkar

摘要

In this study, we present an innovative approach in medicine, leveraging the potential of artificial intelligence (AI) for the diagnosis of tympanic membrane perforations. Due to its central role in sound transmission, the tympanic membrane is a key element in clinical assessment and therapeutic decision-making. Our research focuses on the automated detection of tympanic membrane perforations using advanced AI algorithms, capable of accurately identifying whether a perforation is marginal or not. This study evaluates the performance of a convolutional neural network (CNN) optimized by a genetic algorithm (GA) for the automated classification of tympanic membrane perforations. A dataset of 6000 otoendoscopic images, collected between January 2021 and March 2025 at the University Hospital Center of Oujda, Morocco, was used to train and validate the model. The images were classified into four categories: normal, non-marginal punctate perforation, subtotal perforation, and marginal perforation. The proposed GA-CNN model achieved 96.2% accuracy, 95.7% correctness, 95.9% recall, and an F1-score of 95.8%, surpassing standard architectures such as VGG19, ResNet50, InceptionV3, and EfficientNetV2-B3. These results demonstrate the potential of the proposed CNN as a clinical decision support tool for the objective and reproducible assessment of tympanic perforations.