Keywords
augmented reality (AR) - calvarial biopsy - hands-free - neuronavigation - guidance
Introduction
Calvarial lesions are uncommonly encountered, compromising approximately 0.8% of bone
neoplasms.[1] These intraosseous lesions of the calvarium are often slow, progressive processes
that are either asymptomatic and found incidentally or have localized symptoms such
as pain or a palpable mass. In addition, a variety of diseases can present as calvarial
lesions, such as arachnoid, dermoid, or epidermoid cysts, fibromas, intraosseous hemangiomas,
low-grade meningiomas, osteoid osteomas, and metastases, among others.[2] Most often, these lesions have a wide differential with variable imaging features,[3] and therefore, tissue sampling may be the best option for diagnosis,[3] understanding pathology, and to guide treatment.[2]
However, biopsies of calvarial lesions can be uniquely challenging due to its proximity
to critical structures including the cortex and neurovascular structures near the
skull base.[3] Current methods of biopsy include magnetic resonance (MR)-guided, computed tomography
(CT)-guided biopsy, and CT-fluoroscopy, of which the last two have an increased level
of radiation exposure.[3]
[4] Augmented reality (AR) offers a potential alternative to CT guidance that reduces
radiation exposure and provides hands-free intraoperative guidance through complex
and challenging surgical approaches.[5]
[6]
[7]
[8] The purpose of this study was to introduce and evaluate the feasibility, safety,
and diagnostic yield of a novel method of preoperative AR guidance to perform a calvarial
biopsy.[5]
[6]
[7]
[8]
Clinical Presentation
Presentation
The patient is an 86-year-old female with significant past medical history of coronary
heart disease. She presented to the hospital after two transient episodes of right
upper extremity incoordination and paresthesia. Upon arrival, the patient underwent
head CT without contrast ([Fig. 1A]) and MR imaging (MRI) with and without contrast ([Fig. 1B, C]). These demonstrated a left parietal lytic skull lesion with extracranial extension.
A chest CT also demonstrated a left upper lobe spiculated mass ([Fig. 1D]). At that time, the patient reported no known history of malignancy and no prior
seizures.
Fig. 1
Preoperative imaging. (A) Preoperative CT, (B) preoperative T1 MRI, (C) preoperative T2 MRI. White arrows point to lesion. (D) Preoperative CT of the thorax. Blue arrow demonstrates the left upper lobe spiculated
mass. CT, computed tomography; MRI, magnetic resonance imaging.
Preoperative Imaging
The left parietal intraosseous calvarial lesion measured 4.5 cm (anterior posterior) × 2.2 cm
(cranial caudal) × 2.9 cm (right to left), with extension into the overlying scalp
and underlying pachymeninges. There was also mass effect and vasogenic edema within
the underlying left parietal and posterior frontal lobe. There were no enhancing intraparenchymal
lesions. Differential from imaging studies included metastasis, primary osseous lesion,
or meningioma. The chest CT found a 3.2 × 2.2 × 2.8 cm left upper lobe spiculated
pulmonary mass concerning for neoplasm and thus raised suspicion that the calvarial
lesion was indeed a metastasis.
Operative Technique
The patient was placed supine on the operating table and underwent general anesthesia.
The head was placed in a Mayfield horseshoe head holder. Pinning was not required.
Using Surgical AR-based (Medivis, New York, New York, United States) navigation, a
trajectory was planned centered on the lesion. Surgical AR was registered using point-to-point
registration reliant on four anatomic fiducials (lateral canthi, tip of the nose,
tragus, and nasion). A ground truth, a bi-faced adhesive tag that measures 2 cm × 1 cm
with a QR code on each side that the AR system tracks, was placed on the patient's
forehead, which linked to the registered holographic overlay. Visualization was achieved
with a head-mounted display (HMD), the HoloLens2 (Microsoft, Redmond, Washington,
United States). The center of the lesion was marked according to the designated preplanned
trajectory ([Fig. 2]). A small incision was made over this mark. A combination of Leksell and Kerrison
rongeurs was used to remove a small portion of the overlying skull. Multiple pieces
of the lytic skull lesion were biopsied. Diagnostic yield was appropriate on initial
sampling. Intra-operative pathology was consistent with metastatic carcinoma. Total
surgical time was 35 minutes from incision to closure, including intraoperative pathology
analysis.
Fig. 2
Intraoperative AR workflow. (A) Planning the registration points on the holographic AR rendering. (B) Planning the trajectory for the calvarial biopsy. (C) Intraoperative overlaid hologram with the biopsy trajectory planned, which is outlined
in orange. AR, augmented reality.
Postoperative Course
Postoperative CT and MRI imaging showed interval postoperative changes related to
the biopsy in the center of the lesion as planned ([Fig. 3A, B]). It is important to note that the planned incision and postoperative analysis matched
perfectly. Given the pathology, the patient underwent subsequent resection and cranioplasty
a few days later. A small subjacent mixed collection and pneumocephalus were seen
next to the surgical bed, with persistent left frontal and parietal vasogenic edema,
similar in distribution to the preoperative scan. While the immunohistochemical profile
of this metastatic adenocarcinoma is not entirely specific to a single organ, it was
most consistent with lung as the primary site of origin, supported by the presence
of a lung lesion on chest CT. As there was concern for residual disease, the patient
opted for gamma knife radiosurgery treatment 10 days after resection. The resection
cavity bed was targeted, including both dural and bony edges. Unfortunately, follow-up
MRI imaging on 4/2/2023 found recurrent disease leading to the patient requiring whole
brain radiation therapy. The patient passed from disease progression 5 months after
initial diagnosis.
Fig. 3
Postoperative imaging. (A) Postoperative CT scan and (B) postoperative T2 MRI demonstrate interval postoperative changes related to the biopsy
in the center of the lesion as planned. White arrows point to the site of entry. CT,
computed tomography; MRI, magnetic resonance imaging.
Discussion
Surgical navigation systems have continuously improved since their introduction in
the 1990s.[7]
[8]
[9] These systems have traditionally allowed surgeons to track instrumentation relative
to a patient's anatomy based on co-registration, with a dynamic reference frame and
preoperative imaging.[10] However, computer-based surgical planning and stereotactic navigation are notably
limited by the time needed for registration and the requirement for a separate viewing
screen outside of the surgical field.[11] In addition, the need for framed fixation, which computer-based surgical planning
and stereotactic navigation require, has their own complication and risk profile,
which is well documented in the literature.[12] Guidance systems that force the operator to shift their attention from the surgical
field may interfere with attention and efficient workflow.[13] In addition, any contact with the reference array will lead to inaccurate navigation.
Reregistration intra-operatively is often difficult and in most cases, not possible
unless the patient is re-draped, which increases risk of infection.
The implementation of AR in surgical navigation has allowed image overlay of computer-generated
patient information directly onto the surgical field[14] or in a HMD device.[7]
[8]
[15] This, in turn, has allowed for better visualization of anatomical structures and
surgical targets in the operating room (OR), as well as the attention of the operating
surgeon is not shifted away from the field. Additionally, AR with the use of anatomic
fiducial markers requires less setup time compared with conventional neuronavigation
methods, which may reduce the financial burden to patients and the hospital system
from overall decreased OR utilization.[11] This efficiency, more recently, has been further improved by the optical “snap-to”
registration, which uses artificial intelligence, to map and match the face to the
preoperative imaging.[11] The main limitations of AR navigation, such as latency and a lack of real-time feedback,
are not limited to the use of AR and are intrinsic to intraoperative navigation systems
that rely on preoperative imaging. The use of dynamic AR systems that adjust preoperative
imaging based on real-time data either from ultrasound or computer vision analysis
of cortical architecture with point cloud renderings constructed from the HMD sensor
inputs can adapt to parenchymal deformation. These are future areas for advancement
to address these shortcomings.[16]
The use of frameless AR-guidance in a patient with calvarial metastatic adenocarcinoma
of the lung is described here. The novelty of the Surgical AR system in conjunction
with an HMD for biopsy is that it obviates the need for pinning the patient. In fact,
with the use of a ground truth as an environmental anchor, the hologram moves with
the patient's head and stays accurately registered, as demonstrated here ([Video 1]). The accuracy of a similar hands-free AR-based navigation system (VisAR, Novarad,
American Fork, Utah, United States) has previously been reported and studied in gelatin-based
models.[6] This report adds to the current evidence of literature that the use of AR for cranial
biopsies is accurate and safe, while also minimizing cognitive load and improving
efficiency in the OR through workflow simplification. Since pinning with a Leksell
G-frame or Mayfield C-clamp is unnecessary, the biopsy can then be done under light
sedation with local anesthetic. This has immense impact, especially in the older or
sicker population, where general anesthesia may not be tolerated with the patient's
risk profile.
Video 1 The surgeon quickly registers the patient by mapping out key points on the anatomic
image to the hologram. The hologram produced by the AR system can be seen with anatomic
landmarks as well as the designated planned trajectory. As the patient's head is manipulated
for a more preferred or comfortable positioning, the hologram is appropriately changed
to be in line with the patient's head based on the ground-truth fiducial. This is
demonstrated in a real-time manner.
For more complicated intra-operative applications, AR navigation for cranial surgery
has been investigated previously in the setting of tumor resection.[17] This has afforded better visualization of deep-seated lesions surrounded by complex
anatomical structures. Roethe et al performed a prospective randomized control trial
comparing the use of a HMD-based AR navigation system to conventional neuronavigation
for intracranial tumor resection.[18]
[19] The study demonstrated that the integrated continuous display allows for decreased
work-flow interruption and reduced intraoperative cognitive load to the surgeon. Skyrman
et al developed an AR navigation system that can achieve biopsy at submillimeter accuracies
in a skull phantom.[10]
[20] This was performed within a hybrid OR comprised of a robotic C-arm with intraoperative
cone-beam CT and integrated video tracking. To the best of our knowledge, AR navigation
has not been implemented for biopsies of the calvarium, and as such, this is the first
reported case of successful calvarial biopsy utilizing AR navigation.
Despite the successful localization and collection of biopsy samples, there are several
limitations in the current study. First, this was completed on a single patient. While
the purpose of the article was to understand the utility of AR in a biopsy case, a
larger cohort of patients needs to be evaluated to fully understand the safety, efficacy,
and functionality of the use of AR in biopsies. Finally, this case was particularly
straight-forward as the biopsy in question was more superficial. In future cases,
it would be helpful to have a broader range of depth of biopsies to see how AR helps
in those cases.
Conclusion
The frameless AR navigation system in this study successfully allowed accurate location,
visualization, and biopsy of a calvarial lesion. More so, this was completed without
obscuring the surgical field or requiring time-consuming setup or registration. This
report adds to the current body of literature that AR can be used successfully in
calvarial biopsies.
