2025 Proffered Presentations
S307: TRANSNASAL ORBITAL FOREIGN BODY REMOVAL: AN APPLICATION OF ARTIFICIAL INTELLIGENCE BASED SEMI-AUTOMATIC SEGMENTATION FOR PREOPERATIVE MODELING
Margaret Mitchell, MD, MSHPEd1; Ryan Bartholomew1; Angela Zhu2; Barak Ringel1; Benjamin Bleier1; Noel Ayoub1; 1Mass Eye and Ear/Harvard Medical School; 2Warren Alpert Medical School
Introduction: Artificial intelligence-based segmentation generates three-dimensional models of orbital structures, including the extraocular muscles, optic nerve, and ophthalmic artery, and even pathology such as tumors. These models may also be valuable in specific clinical scenarios, such as transnasal orbital foreign body removal.
Objective: We utilized semi-automatic segmentation to construct a preoperative model of orbital contents with an embedded foreign body to guide planning for surgical removal.
Methods: A 60-year-old patient with an embedded orbital foreign body after a workplace accident was identified. The patient’s preoperative computed tomography imaging was obtained (Figure 1), and 3D slicer, an open source software for segmentation of biomedical images, was utilized to segment the foreign body with key anatomical structures (Figure 2). The surgical team utilized this model pre- and intraoperatively to guide removal.
Results: The 3D model revealed the foreign body’s precise position relative to key structures: abutting the skull base medially, tracking between the medial and inferior rectus, and lying inferior to the optic nerve and ophthalmic artery laterally (Figure 2). Crucially, the model indicated that any counterclockwise rotation of the foreign body in a coronal plane could potentially displace or sever the optic nerve and ophthalmic artery.
The patient underwent left-sided functional endoscopic sinus surgery. The orbit was then decompressed, and as the lamina was removed, the superior and medial-most aspect of the foreign body was visualized (Figure 3). This object, a 2cm sharp razor blade, then was carefully grasped and pulled medially and inferiorly for removal through the nasal cavity, ensuring no counterclockwise rotation to avoid damaging the surrounding structures.
The patient had an excellent outcome: prior to surgery, vision was limited to shadows, but after surgery, vision improved to 20/60.
Discussion: This case demonstrates the utility of three-dimensional modeling generated from artificial intelligence-based segmentation: the surgical team planned a certain approach, removing the foreign body in a specific vector, to avoid damage to key orbital structures and specifically inadvertent optic nerve section given its sharp edges.
Conclusion: This is a case of a patient with an excellent outcome after transnasal removal of large intraorbital foreign body; these preoperative models utilized may provide valuable insight when planning challenging orbital procedures.
Figures
Figure 1: Preoperative computed tomography imaging
Caption: Coronal cross-section showing the left intraorbital foreign body spanning the lamina.
Figure 2: 3-dimensional reconstruction of left orbit
Caption: Contents of the left orbit, as viewed from posteriorly with bony landmarks removed (S=superior, M=medial, L=lateral, I=inferior). Medial and inferior rectus are in brown, globe in yellow, and foreign body in purple. The optic nerve is in blue and ophthalmic artery in red.
Figure 3: Intraoperative image
Caption: Left nasal cavity. The foreign body is seen emanating from the orbit through the lamina (L) and into the posterior ethmoid cavity. MT denotes middle turbinate.