Segmentation of cortical bone, trabecular bone, and medullary pores from micro-CT images using 2D and 3D deep learning models

Computed tomography (CT) enables rapid imaging of large-scale studies of bone, but those datasets typically require manual segmentation, which is time-consuming and prone to error. Convolutional neural networks (CNNs) offer an automated solution, achieving superior performance on image data. In this methodology-focused paper, we used CNNs to train segmentation models from scratch on 2D and 3D patches from micro-CT scans of otter long bones. These new models, collectively called BONe (Bone One-shot Network), aimed to be fast and accurate, and we expected enhanced results from 3D training due to better spatial context. Contrary to expectations, 2D models performed slightly better than 3D models in labeling details such as thin trabecular bone. Although lacking in some detail, 3D models appeared to generalize better and predict smoother internal surfaces than 2D models. However, the massive computational costs of 3D models limit their scalability and practicality, leading us to recommend 2D models for bone segmentation. BONe models showed potential for broader applications with variation in performance across species and scan quality. Notably, BONe models demonstrated promising results on skull segmentation, suggesting their potential utility beyond long bones with further refinement and fine-tuning.

Unveiling the vulnerability of the human abducens nerve: insights from comparative cranial base anatomy in mammals and primates

The topographic anatomy of the abducens nerve has been the subject of research for more than 150 years. Although its vulnerability was initially attributed to its length, this hypothesis has largely lost prominence. Instead, attention has shifted toward its intricate anatomical relations along the cranial base. Contrary to the extensive anatomical and neurosurgical literature on abducens nerve anatomy in humans, its complex anatomy in other species has received less emphasis. The main question addressed here is why the human abducens nerve is predisposed to injury. Specifically, we aim to perform a comparative analysis of the basicranial pathway of the abducens nerve in mammals and primates. Our hypothesis links its vulnerability to cranial base flexion, particularly around the sphenooccipital synchondrosis. We examined the abducens nerve pathway in various mammals, including primates, humans (N = 40; 60% males; 40% females), and human fetuses (N = 5; 60% males; 40% females). The findings are presented at both the macroscopic and histological levels. To associate our findings with basicranial flexion, we measured the cranial base angles in the species included in this study and compared them to data in the available literature. Our findings show that the primitive state of the abducens nerve pathway follows a nearly flat (unflexed) cranial base from the pontomedullary sulcus to the superior orbital fissure. Only the gulfar segment, where the nerve passes through Dorello's canal, demonstrates some degree of variation. We present evidence indicating that the derived state of the abducens pathway, which is most pronounced in humans from an early stage of development, is characterized by following the significantly more flexed basicranium. Overall, the present study elucidates the evolutionary basis for the vulnerability of the abducens nerve, especially within its gulfar and cavernous segments, which are situated at the main synchondroses between the anterior, middle, and posterior cranial fossae-a unique anatomical relation exclusive to the abducens nerve. The principal differences between the pathways of this nerve and those of other cranial nerves are discussed. The findings suggest that the highly flexed human cranial base plays a pivotal role in the intricate anatomical relations and resulting vulnerability of the abducens nerve.

In-house three-dimensional printing for surgical planning: learning curve from a case series of temporomandibular joint and related disorders

Three-dimensional (3D) printed models can improve the understanding of the structural anatomic changes in cases of temporomandibular joint ankylosis and pseudoankylosis leading to closed jaw locking. Their use in pre-surgical planning and intraoperative guidance has been reported, contributing to the predictability and success of these surgery procedures, which can be quite complex, especially in small animal patients. The use and production of 3D tools and models remain challenging and are so far limited to institutions with high (economical and human) resources. This study aims to propose simplified workflows using open-source software to facilitate an in-house 3D printing process. To illustrate this, three cases of temporomandibular joint ankylosis and one of pseudoankylosis were reviewed, where in-house 3D printed models were used for client communication and surgical management. The 3D models were segmented from computed tomography and printed via stereolithography. They were used to support discussion with clients (n = 4), to allow surgeons to pre-surgical plan and practice (n = 4) and for intraoperative guidance during surgery (n = 2). Surgical cutting guides were produced in one case to improve precision and define more accurately osteotomy lines. It is essential to consider the initial time and financial investment required for establishing an in-house 3D printing production, particularly when there is a need to produce biocompatible tools, such as surgical cutting guides. However, efficient and streamlined workflows encourage the integration of this technology, by accelerating the printing process and reducing the steep learning curves, while open-source software enhances accessibility to these resources.

Accuracy of bone segmentation and surface generation strategies analyzed by using synthetic CT volumes

Different kinds of bone measurements are commonly derived from computed-tomography (CT) volumes to answer a multitude of questions in biology and related fields. The underlying steps of bone segmentation and, optionally, polygon surface generation are crucial to keep the measurement error small. In this study, the performance of different, easily accessible segmentation techniques (global thresholding, automatic local thresholding, weighted random walk, neural network, and watershed) and surface generation approaches (different algorithms combined with varying degrees of simplification) was analyzed and recommendations for minimizing inaccuracies were derived. The different approaches were applied to synthetic CT volumes for which the correct segmentation and surface geometry were known. The most accurate segmentations of the synthetic volumes were achieved by setting a case-specific window to the gray value histogram and subsequently applying automatic local thresholding with appropriately chosen thresholding method and radius. Surfaces generated by the Amira® module Generate Lego Surface in combination with careful surface simplification were the most accurate. Surfaces with sub-voxel accuracy were obtained even for synthetic CT volumes with low contrast-to-noise ratios. Segmentation trials with real CT volumes supported the findings. Very accurate segmentations and surfaces can be derived from CT volumes by using readily accessible software packages. The presented results and derived recommendations will help to reduce the measurement error in future studies. Furthermore, the demonstrated strategies for assessing segmentation and surface qualities can be adopted to quantify the performance of new segmentation approaches in future studies.