Fully automated segmentation in temporal bone CT with neural network: a preliminary assessment study

Abstract

Background Segmentation of important structures in temporal bone CT is the basis of image-guided otologic surgery. Manual segmentation of temporal bone CT is time- consuming and laborious. We assessed the feasibility and generalization ability of a proposed deep learning model for automated segmentation of critical structures in temporal bone CT scans. Methods Thirty-nine temporal bone CT volumes including 58 ears were divided into normal (n = 20) and abnormal groups (n = 38). Ossicular chain disruption (n = 10), facial nerve covering vestibular window (n = 10), and Mondini dysplasia (n = 18) were included in abnormal group. All facial nerves, auditory ossicles, and labyrinths of the normal group were manually segmented. For the abnormal group, aberrant structures were manually segmented. Temporal bone CT data were imported into the network in unmarked form. The Dice coefficient (DC) and average symmetric surface distance (ASSD) were used to evaluate the accuracy of automatic segmentation. Results In the normal group, the mean values of DC and ASSD were respectively 0.703, and 0.250 mm for the facial nerve; 0.910, and 0.081 mm for the labyrinth; and 0.855, and 0.107 mm for the ossicles. In the abnormal group, the mean values of DC and ASSD were respectively 0.506, and 1.049 mm for the malformed facial nerve; 0.775, and 0.298 mm for the deformed labyrinth; and 0.698, and 1.385 mm for the aberrant ossicles. Conclusions The proposed model has good generalization ability, which highlights the promise of this approach for otologist education, disease diagnosis, and preoperative planning for image-guided otology surgery.

Application value of a deep learning method based on a 3D V-Net convolutional neural network in the recognition and segmentation of the auditory ossicles

“… Li et al (2020) adopted a 3D Deep Supervised Densely (DSD) algorithm to obtain the DSCs of the malleus and the incus, which were both 0.82; however, the stapes was not segmented. Ke et al applied a 3D convolutional neural network to successfully realize the automatic segmentation of the labyrinth, the auditory ossicles, and the facial nerve in both conventional and abnormal temporal bone CTs and achieved excellent results ( Ke et al, 2020 ; Ding et al, 2021 ; Wang et al, 2021 ). Other scholars applied multi-view fusion and deep learning algorithms to design an accurate segmentation of the malleus and the incus and further improved the segmentation accuracy of the stapes with an active contour-loss constraint method ( Zhu et al, 2021 ).…”

Pipeline for Automated Processing of Clinical Cone-Beam Computed Tomography for Patient-Specific Temporal Bone Simulation: Validation and Clinical Feasibility

“…Future efforts should be made to further improve the automated segmentation of the facial nerve and chorda, and one approach might be the implementation of vessel tracking techniques (19). Other approaches to temporal bone segmentation such as statistical shape modeling (18) and neural networks (20) have been reported to also have good accuracy. Direct comparison of metrics with other approaches to automated segmentation is beyond the scope of this article and would require access to these algorithms and training data sets.…”

A Sleeve-Based, Micromotion Avoiding, Retractable and Tear-Opening (SMART) Insertion Tool for Cochlear Implantation

“…Tool placement based purely on the surgeon's anatomical expertise may therefore be possible, but this hypothesis remains to be tested. Further guidance could be derived from preoperative computed tomography images, possibly with a visualization through automatic segmentation of anatomical structures [31], augmented reality overlays [32] or customized mastoid templates [10] to indicate optimal tool placement.…”