Developing a Virtual Endoscopic Surgery Planning System to Optimize Surgical Outcomes

Abstract

Endoscopic nasal sinus surgery [1] has been the procedure of choice for the treatment of nasal obstruction and sinus disease. Nasal endoscopy provides landmark visualization and navigation for small instruments to operate within the nasal airway and sinuses. However, outcomes of these surgeries are highly variable, from short-term favorable outcomes of 60%‒90% to long-term outcomes as low as 30% [2, 3], potentially due to fact that planning and predicting functional outcomes (e.g., airflow) of endoscopic sinus surgeries can be difficult or deceptive based solely on visualizing computerized tomography or endoscopic images.

Virtual surgery planning (VSP) has been widely used in many other surgical fields that allow the surgeon to virtually plan the surgery to optimize surgical outcomes. Since previous endoscopic sinus surgery simulators [4-10] mostly serve as surgical skill-training tools, we thus assembled an interdisciplinary team of engineers, computational scientists, and clinicians to develop an endoscopic sinus VSP system for experienced surgeons rather than a training platform—that utilizes state of the art computer graphics and 3D modeling to provide a prospective, preoperative guide that can optimize and improve surgical outcomes.

The prototype, running on desktop workstation with Intel Xeon E5-1650 v3 processor and NVIDIA Quadro P5000 graphics card, comprises three synergistic components—a volume renderer to provide 3D endoscopic visualization based on patient's clinical computerized tomography (CT) imaging; a haptic endoscopy tool to navigate and perform the virtual surgery; and the computational fluid dynamic (CFD) system to predict the nasal airflow outcome. The rendering system uses CUDA for direct volume rendering and OpenGL (version 4.0) for graphics rendering. A haptic device (PHANTOM, SensAble Technologies, Inc.) that can apply forces in three degrees of freedoms, provides interaction between virtual endoscopic tools and the 3D CT-based volumetric model during virtual surgery. The combination of 3D visual rendering and the haptic feedback allows the surgeon to remove the virtual obstructive tissue similar to in endoscopic surgery.

The developed VSP can load any patient's CT in DICOM format in real-time from various sources, for example, cone beam CTs or conventional spiral CTs, and provide highly detailed 3D visual display of the nasal sinus airway (see Figure 1, Video S1, and Supporting Information Method). The ability to directly load versatile clinical CT images without pre-processing is the key to allow its wide utilization in clinical settings. Pre-calculated airflow resistance, wall shear stress, pressure drop are displayed on the anatomy to identify potential sites of obstruction. In Figure 1, a microdebrider was selected, with available endoscopic tool selections to be expanded in future software versions. The size of the microdebrider can be virtually adjusted. To ease operation, the surgical tool and camera are combined into a single haptic device, so that the surgeon can operate with one hand and toggle between different tools, such as the endoscope and the microdebrider, using buttons accessible by the user's index finger. While the camera view is updated as the virtual endoscope navigates the nasal airway, the angle and distance of the camera to the tool tip can be adjusted. CT image guidance is also supported so that the surgeon can visualize the real-time tool tip location on three orthogonal CT slices, mimicking the operating room (OR) experience. Virtual surgeries can be saved intermittently (“mini save” button) or undone (“restore volume” button) to allow for easy surgical planning. By restoring and saving multiple versions of different surgical approaches, different outcomes on a single patient can be compared and the process is reiterated until an optimal result is reached.

As proof-of-concept, we performed a series of isolated or combined procedures on one patient, who had nasal valve obstruction, septal body hypertrophy, and a middle concha bullosa, and confirmed unilateral olfactory loss only to the left side (phenyl ethyl alcohol [PEA] threshold: left = 5, right = 9.3, normative range: ≥8). Before virtual surgery, this patient's computed area-averaged PEA absorption flux to olfactory fossa (OF) on the left side was below the normal range (Figure 2e) that was established from 22 healthy controls [11], confirming the likely obstruction-driven olfactory loss, whereas the patient's right pre-surgery olfactory PEA flux was within the normal range. Based on VSP, an isolated medial partial middle turbinectomy (PMT) demonstrated the best outcome, better than traditionally performed lateral PMT, while septal body reduction worsened air/odor flow to OF.

This proof of concept trial demonstrates the potential usefulness of VSP in preoperative planning based on objective benchmarks and could be a valuable tool for optimizing future surgical outcomes. In contrast to previous systems which emphasize training [4-10], the current VSP emphasizes on surgical planning. The endoscope environment and haptic feedback only serve as an intuitive interface for experienced surgeons to remove obstructive tissue and plan for surgery in a familiar and more effective way than say on CT scan or using a keyboard and mouse. The integration of an intuitive interface, ability to easily load patient image data, with clinically relevant outcome data (CFD) are significant advancements over previous attempts of sinus surgical planning simulators [12]. Such a well-implemented planning tool is likely to have a higher transfer rate to the clinical practice with broader clinical impact than those only for surgical training purposes. Finally, the optimized surgical plans can potentially be translated into a CT scan (Figure 1) and loaded into image guidance systems in ORs to guide the surgeon during surgery, which may complete the full circle of personalized medicine: from planning, optimizing, and to the OR.

Conceptualization, methodology, project administration, supervision, and funding acquisition: Kai Zhao, Gregory J. Wiet, and Don Stredney. Investigation: Bradley Hittle, Guillermo Maza, Ahmad Odeh, Brenda Shen, Bradley A. Otto, Don Stredney, Gregory J. Wiet, and Kai Zhao. Visualization: Guillermo Maza, Bradley Hittle, Ahmad Odeh, Brenda Shen, and Kai Zhao. Writing—original draft: Kai Zhao. Writing—review and editing: Kai Zhao, Ahmad Odeh, Guillermo Maza, Bradley Hittle, and Gregory J. Wiet.

The authors declare they have no conflicts of interest.