Picture archiving and communication systems (PACS) allow for easy storage and viewing of medical imaging information. Traditional PACS viewers present images in x-ray, computed tomography (CT), and magnetic resonance imaging (MRI) data on a 2-dimensional (2D) workstation screen to be examined by a surgical team in preparation for a complex procedure [1,2]. While these systems have been shown to be accurate and easy to use for the analysis of medical images [3], they are also limited by their requirement of a desktop computer, laptop, or smartphone screen [4]. Dias et al [5] report that 2 of the most common problems of traditional PACS are the mismatch between the 2D viewing screen and the real world and the accompanying lack of flexibility and efficiency of use.

Augmented reality (AR) and virtual reality (VR) technologies have the potential to address these shortcomings. AR and VR alike allow for the realistic and interactive digital representation of objects in a 3D space. As such, both technologies are already successfully deployed across a diverse set of applications, including terrestrial navigation [6], architectural modeling [7], automotive engineering [8], and education [9]. The same properties could be applied to present a realistic overlay of medical devices and tools on patients’ anatomy in 3D space on a portable, shared visualization method.

Whereas VR presents an entirely digital representation of objects and their environment, AR allows for the overlay of digital holograms on a live real-world scene. In addition, many VR systems require a dedicated physical play space to allow for the experience of the completely immersive digital experience [10]. These characteristics make AR a more likely candidate for the development of interactive tools assisting the dynamic clinical workflow.

The potential of AR systems to assist in clinical tasks has been extensively reviewed by Uppot et al [11]. Possible use cases include supplementing radiology training; communicating with colleagues, referring clinicians, and patients; and aiding in interventional radiology procedures. Additional uses for AR in medicine include providing simulations for advanced life support training [12], visualizing patient anatomy including tumors [13], and guiding assistants during robotic surgery [14]. The increased spatial understanding of anatomy with AR has been shown to positively impact surgical care during laparoscopic surgery for visualizing hidden patient anatomy [15], resection of neurological tumors without causing new neurological deficit [16], and breast tumor resection by maximizing breast conservation [17]. Multiple other non–patient outcome benefits have been proposed, including overall operating room efficiency [18,19], and more specifically—reduced operating room time, increased surgical precision, and reduced radiation exposure [20].

In order to create an AR model suitable for presurgical planning, the medical image from a CT or MRI scan must first be segmented using a DICOM viewer to visualize only the object or organ of interest. The resulting image is passed onto an image processing software that renders the object’s volumes and surfaces into a 3D scalar field model. This model can later be loaded in a dedicated AR software designed for projecting the image onto an AR or mixed reality headset display. Similar technologies have evaluated the use of AR systems for the visualization of MRI data [21]. However, the focus of this study is the validation of CT-derived holograms. Although the visualization of CT-derived holograms has been assessed, measurement systems for these CT-derived holograms are rarely evaluated or utilized.

As AR becomes more widely used in presurgical planning, it is crucial to know that these systems meet the gold standard for medical image measurement. This study aims to validate the feasibility, safety, and efficacy of a novel ARPPS, compared to a standard-of-care PACS viewer, in order to support its use in the presurgical visualization and measurement of CT-derived imaging of patient anatomy and surgical tools.

Lin’s concordance correlation coefficient showed almost perfect concordance of the standard PACS viewer and the ARPPS (Figure 4, Table 2). Additionally, no significant difference in interrater reliability was observed for the circumference and angle tool measurements for both the PACS and ARPPS separately (Figure 4, Table 2).

The Shapiro-Wilk tests failed to reject the null hypothesis of normality (Table 3). Bland-Altman plots evaluating the circumference tool showed an average bias of 0.08% with a 95% CI –4.20% to 4.36%. Bland-Altman plots evaluating the angle tool showed an average bias of –1.84% with a 95% CI –6.17% to 2.14%. The bias and confidence intervals of each of the 3 measures for the circumference and angle tools are reported in Table 3. The Bland-Altman plots of each of the measurements, as well as the combined measurements are shown for the circumference tool (Figure 5 a-d) and angle tool (Figure 5 e-h).

The variability of the percent error of each of the measurements using the ARPPS as compared to using the standard PACS are visualized in individual box plots in Figure 6.