[论文解读] From Preoperative CT to Postmastoidectomy Mesh Construction: Mastoidectomy Shape Prediction for Cochlear Implant Surgery
论文提出一种混合自监督和弱监督框架,通过术前 CT 预测乳突切除形状,平均 Dice 分数为 0.72,并推进 CI 手术的 3D 术后乳突表面重建。
Cochlear Implant (CI) surgery treats severe hearing loss by inserting an electrode array into the cochlea to stimulate the auditory nerve. An important step in this procedure is mastoidectomy, which removes part of the mastoid region of the temporal bone to provide surgical access. Accurate mastoidectomy shape prediction from preoperative imaging improves pre-surgical planning, reduces risks, and enhances surgical outcomes. Despite its importance, there are limited deep-learning-based studies regarding this topic due to the challenges of acquiring ground-truth labels. We address this gap by investigating self-supervised and weakly-supervised learning models to predict the mastoidectomy region without human annotations. We propose a hybrid self-supervised and weakly-supervised learning framework to predict the mastoidectomy region directly from preoperative CT scans, where the mastoid remains intact. Our hybrid method achieves a mean Dice score of 0.72 when predicting the complex and boundary-less mastoidectomy shape, surpassing state-of-the-art approaches and demonstrating strong performance. The method provides groundwork for constructing 3D postmastoidectomy surfaces directly from the corresponding preoperative CT scans. To our knowledge, this is the first work that integrating self-supervised and weakly-supervised learning for mastoidectomy shape prediction, offering a robust and efficient solution for CI surgical planning while leveraging 3D T-distribution loss in weakly-supervised medical imaging.
研究动机与目标
- 为了帮助术前规划、机器人辅助和术中导航,在耳蜗植入(CI)手术中实现对乳突切除形状的准确预测。
- 开发一个通过自监督学习利用术后 CT 数据来消除人工标注的框架。
- 在弱监督设置中通过新颖的 3D T-Distribution 损失提高对嘈杂标签的鲁棒性。
提出的方法
- 两阶段框架,结合来自术前和术后 CT 对的自监督学习,以预测乳突切除区域。
- 基于 SegMamba 的网络,使用预训练的 SAM-Med3D 编码器,在术前 CT 上产生乳突切除区域的反概率图。
- 利用自监督输出推导的弱标签进行弱监督精炼,并通过 3D T-Distribution 损失对不规则的乳突切除几何进行鲁棒建模。
- 引入基于 Student-t 分布的 3D T-Distribution 损失,以提高在 3D 医学图像分割中对嘈杂标签的鲁棒性。
- 在与标准 3D 分割模型的对比评估中,显示在测试数据上具有更优的 Dice 和更低的 HD95。
实验结果
研究问题
- RQ1能否在没有人工标注的情况下,仅凭术前 CT 直接准确预测乳突切除形状?
- RQ2混合自监督和弱监督的方法是否能提高对 3D 乳突切除分割中嘈杂标签的鲁棒性?
- RQ3提出的 3D T-Distribution 损失是否优于传统损失在弱监督的 3D 医学影像任务中的表现?
- RQ4预测的术后乳突表面如何支持下游导航和可视化,以用于 CI 手术?
主要发现
- 从术前 CT 预测复杂的乳突切除形状的平均 Dice 为 0.72。
- 与 UNet、Transformer、SwinUNETR 基线模型相比,所提出的方法在多项指标上达到最新的性能。
- 3D T-Distribution 损失在对嘈杂/弱标签的鲁棒性方面优于 CE、BCE、Focal、MSE 和 MAE 损失,在消融测试中表现良好。
- 消融结果显示基于 Mamba 的 SegMamba 架构在自监督学习中优于 U-Net 和 Transformer 基线。
- 表 2 显示所提出的方法在 Dice(0.721)和 HD95(16.159)上达到竞争方法中的最高值和最低值。
- 该框架使能够从术前 CT 重建术后乳突表面,以提升术中导航。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。