Statistical Surgical Process Modeling: Analysis of Workflow and Performance of Emerging Technologies in Image-Guided Spine Surgery.

Abstract

Background: Surgical process modeling (SPM) presents a rigorous ontology and useful means of understanding complex surgical workflows. Early-stage understanding of the potential benefit of emerging technologies can be gained with such an approach- for example, in image-guided spine surgery.

Purpose: This work extends SPM ontology to a statistical framework for computational modeling and simulation, analyzing the effect of new technologies on quantitative outcomes. Two example use cases are detailed in spine surgery-target localization with long-length imaging and pedicle screw placement with intraoperative CT and navigation.

Methods: The SPM ontology of phase, step, and activity was mapped to a directed graph depiction of workflow in which activities were described by statistical distributions in outcomes, parameterized and validated via retrospective and/or prospective studies, and structured within an SPM table and object-oriented computational framework. The approach was applied to quantify the influence of two example technologies in spine surgery: (1) long-length radiographic imaging for target localization (compared to conventional fluoroscopic level-counting); and (2) intraoperative CT and 3D navigation for pedicle screw placement (compared to conventional fluoroscopic guidance). Distributions in cycle time, radiation dose, and geometric accuracy were investigated, with clear depiction of median and outlier performance provided by 3D visualization.

Results: Statistical SPMs yielded insight on complex workflows and the potential benefits of emerging technologies. For target localization, intraoperative long-length imaging reduced median cycle time from 15.5 min (for fluoroscopic level-counting) to 13.4 min and eliminated outliers associated with visibility of markers or human error. For pedicle screw placement, 3D navigation introduced an additional median ~ 15 min procedural overhead in planning and registration but reduced median procedure time from 229 min (for fluoroscopy guidance) to 180 min. 3D navigation also quantifiably improved the geometric accuracy of screw placement and reduced the frequency of pedicle screw breach (Grade C or worse).

Conclusions: Statistical SPMs provide a valuable methodology to evaluate and communicate complex interventional workflows and the potential benefit of emerging technologies. Two SPM use cases were developed in the context of spine surgery and provided a basis for investigating the potential performance of emerging technologies, such as intraoperative imaging, navigation, robotic assistance, and augmented reality systems. As a framework for virtual clinical trials, statistical SPMs can provide insight on outcome measures that are difficult to evaluate in the laboratory and expensive / time-consuming to measure in prospective clinical studies.