From Preoperative CT to Postmastoidectomy Mesh Construction:1Mastoidectomy Shape Prediction for Cochlear Implant Surgery

TL;DR

This work tackles the challenge of predicting the mastoidectomy shape from preoperative CT to support cochlear implant surgery, addressing the lack of ground-truth labels with a hybrid self-supervised and weakly-supervised framework. A SegMamba-based network leverages postoperative CTs for self-supervision and introduces a $3D$ T-Distribution loss for robust weak-label learning, achieving a mean Dice of $0.72$ and state-of-the-art performance compared with standard architectures. The approach uses a large 3D CT dataset (751 image pairs) and novel losses such as $L_{msssim ext{_}cscc}$, $L_{smooth}$, and $L_{TD}$, enabling reconstruction of a 3D post-mastoidectomy surface directly from preoperative scans. This framework has potential to improve preoperative planning, robotic assistance, and intraoperative navigation for CI procedures, though further validation across diverse centers and enhancement of reconstructed mesh textures are needed for clinical deployment.

Abstract

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.

Figures (10)

• Figure 1: Example CT scans and Mastoidectomy Region. Top row shows preoperative CT scans and bottom row illustrates the corresponding highly noisy postoperative CT scans. The highlighted area in red is an example of the mastoidectomy shape region.
• Figure 2: Framework overview. We compare the predicted mastoidectomy region applied to the preoperative CT $\rho \otimes \delta$ against the postoperative CT $\omega$ using similarity-based loss function $L_{msssim\_cscc}$. A smoothing term $L_{smooth}$ is also applied to $\delta$ to further reduce noise output by network $f_{\theta}$.
• Figure 3: SegMamba-based Neural Network with pretrained SAM-Med3D Image Encoder
• Figure 4: Framework Overview. Snowflake indicate frozen and non-trainable parameters, fire represents trainable parameters, and "SIM" compares the similarity between two objects.
• Figure 5: Mamba-based Architecture. The model takes preoperative CT as input and outputs the predicted mastoidectomy shape.