跳转至

文章背景与核心概要

人工耳蜗（CIs）通过电刺激恢复听力，而混合型人工耳蜗则采用电声联合刺激（EAS），将残余的低频声学听力与电刺激相结合。然而，植入后耳蜗内纤维化的形成会阻碍残余听力，并随着时间推移降低EAS的有效性。

为了帮助研究并减少啮齿动物模型中的纤维化负担，本文引入了一个新颖的、经过人工标注的光学相干断层扫描（OCT）图像数据集，该数据集取自慢性植入的豚鼠。作者评估了几种最先进的语义分割模型，并证明了一种名为 2D-OCT-UNET 的改进型UNET架构（在放大后的OCT输入分辨率下运行）能够实现卓越的性能，从而能够可靠地自动计算耳蜗纤维化负担。

迈向残余听力损失研究：新型耳蜗OCT数据集中的纤维化定量分析 (Towards Investigating Residual Hearing Loss: Quantification of Fibrosis in a Novel Cochlear OCT Dataset)

作者： Julia Dietlmeier, Benjamin Greenberg, Wenxuan He, Teresa Wilson, Rubing Xing, Jordan Hill, Adrienne Fettig, Madeline Otto, Teyhana Rounsavill, Lina A. J. Reiss, Jingang Yi, Noel E. O'Connor, George W.S. Burwood
发表于： IEEE Transactions on Biomedical Engineering, 72(7), pp. 2218–2228, July 2025
arXiv ID: arXiv:2608.21189 [cs.CV]
提交时间： 2026年8月21日

Authors: Julia Dietlmeier, Benjamin Greenberg, Wenxuan He, Teresa Wilson, Rubing Xing, Jordan Hill, Adrienne Fettig, Madeline Otto, Teyhana Rounsavill, Lina A. J. Reiss, Jingang Yi, Noel E. O'Connor, George W.S. Burwood
Published in: IEEE Transactions on Biomedical Engineering, 72(7), pp. 2218–2228, July 2025
arXiv ID: arXiv:2608.21189 [cs.CV]
Submitted: 21 August 2026

📋 摘要 (Summary)

人工耳蜗（CIs）通过电刺激恢复听力，而混合型人工耳蜗则采用电声联合刺激（EAS），将残余的低频声学听力与电刺激相结合。然而，植入后耳蜗内纤维化的形成会阻碍残余听力，并随着时间推移降低EAS的有效性。

为了帮助研究并减少啮齿动物模型中的纤维化负担，本文引入了一个新颖的、经过人工标注的光学相干断层扫描（OCT）图像数据集，该数据集取自慢性植入的豚鼠。作者评估了几种最先进的语义分割模型，并证明了一种名为 2D-OCT-UNET 的改进型UNET架构（在放大后的OCT输入分辨率下运行）能够实现卓越的性能，从而能够可靠地自动计算耳蜗纤维化负担。

Cochlear implants (CIs) restore hearing through electrical stimulation, while hybrid CIs utilize electroacoustic stimulation (EAS) to combine residual low-frequency acoustic hearing with electrical stimulation. However, the formation of intracochlear fibrosis in response to the implant can impede residual hearing and reduce the effectiveness of EAS over time.

To help study and minimize fibrotic burden in rodent models, this paper introduces a novel, manually annotated dataset of optical coherence tomography (OCT) images taken from chronically implanted guinea pigs. The authors evaluate several state-of-the-art semantic segmentation models and demonstrate that a modified UNET architecture—dubbed 2D-OCT-UNET—operating on upscaled OCT input resolution achieves superior performance, enabling reliable automated calculation of cochlear fibrotic burden.

📌 研究概览 (Research Overview)

  • 研究目的： 评估人工耳蜗植入引起的耳蜗内纤维化，以更好地理解并保护残余听力损失。
  • 研究方法： 生成并标注了一个来自慢性植入豚鼠的高分辨率光学相干断层扫描（OCT）图像新数据集。对比了各种最先进的语义分割模型。
  • 研究结果： 所提出的改进型UNET架构（2D-OCT-UNET）在放大后的OCT输入分辨率下运行，取得了最高的分割和识别性能。
  • 研究意义： 代表了计算机视觉技术首次成功应用于包含纤维化的植入耳蜗OCT数据集，为计算纤维化负担提供了一种可靠的自动化方法。
  • Objective: Assess intracochlear fibrosis resulting from cochlear implants to better understand and preserve residual hearing loss.
  • Methods: Generated and annotated a novel dataset of high-resolution optical coherence tomography (OCT) images from chronically implanted guinea pigs. Compared various state-of-the-art semantic segmentation models.
  • Results: The proposed modified UNET architecture (2D-OCT-UNET) running on upscaled OCT input resolution yielded the highest segmentation and identification performance.
  • Significance: Represents the first successful application of computer vision techniques to an OCT dataset of implanted cochleae containing fibrosis, providing a reliable automated approach for calculating fibrotic burden.