The efficacy of augmented reality technology assisted by 3D-CT reconstruction in microvascular decompression for hemifacial spasm craniotomy

Abstract

Background

We aimed to assess the efficacy of augmented reality technology assisted by three-dimensional (3D) computed tomography (CT) reconstruction, in facilitating microvascular decompression (MVD) during craniotomy for hemifacial spasm.

Methods

A retrospective analysis was conducted on 80 patients who underwent MVD for hemifacial spasm at Hebei General Hospital between January 2, 2020, and March 24, 2021. Among them, 43 patients received traditional craniotomy (assigned to the traditional group), while 37 patients underwent modified craniotomy (assigned to the modified group). The distinctive feature in the modified group involved employing 3D-CT reconstruction assisted by augmented reality technology, specifically utilizing Sina software, for precise localization of scalp incision. The impact of the modified method on surgery was assessed based on operation time, incision length, postoperative complications, long-term efficacy, and patient-reported outcomes.

Results

No significant differences in age, sex, disease duration, and disease side were observed between the two groups (P > 0.05). The modified group exhibited a significantly shorter average craniotomy time (29.68 ± 4.89 min vs. 34.19 ± 4.55 min, P < 0.001) and time to close the skull (25.22 ± 3.12 min vs. 28.95 ± 2.54 min, P < 0.001) compared to the traditional group. Additionally, the incision length in the modified group (59.69 ± 10.71 mm) was evidently lower than that in the traditional group (70.84 ± 11.27 mm, P < 0.001). The overall rate of any postoperative complication was significantly lower in the modified group (5.4%) compared to the traditional group (23.3%; P = 0.018). While overall immediate postoperative complication rates were not statistically different, the modified group showed a trend towards fewer complications, particularly no CSF leakage. At 1-year follow-up, the modified group had a lower spasm recurrence rate (2.7% vs. 7.0%, P = 0.348), significantly lower postoperative pain scores at 7 days (VAS: 2.8 ± 0.9 vs. 3.6 ± 1.1, P = 0.002), and higher patient satisfaction (94.6% vs. 86.0%, P = 0.184).

Conclusion

Employing 3D-CT reconstruction assisted by augmented reality technology in MVD for hemifacial spasm significantly improves surgical efficiency, is associated with favorable long-term outcomes, and enhances patient-reported satisfaction, suggesting its potential value for clinical application and potentially contributing to enhanced safety, particularly in reducing specific complications like CSF leakage.

Introduction

Hemifacial spasm, a prevalent neurological disorder, manifest as repetitive, paroxysmal twitching of one or both sides of facial muscles. In severe cases, patients may experience difficulties in eye-opening, mouth skew, tinnitus, and other associated symptoms [1]. Microvascular decompression (MVD) stands as a definitive therapeutic intervention for hemifacial spasm, trigeminal neuralgia, and glossopharyngeal neuralgia [2]. MVD involves relocating implicated blood vessels to release the compressed facial nerve, thereby eliminating abnormal nerve fiber impulses [2]. This meticulous neurosurgical procedure requires the separation of the affected facial nerve from adjacent compressive blood vessels, boasting a high success rate in alleviating spasticity [3, 4].

Given the intricacies of MVD, surgeons must possess profound understanding of the microanatomy of the ponto-cerebellar region and impeccable surgical skills to avert potential complications [5]. The accuracy and precision of craniotomy during MVD significantly influence the ease of intraoperative treatment of responsible vessels and the duration of the surgical procedure. With advancing imaging technology, three-dimensional (3D) imaging technology has been instrumental in preoperative planning. Previous studies have demonstrated the utility of preoperative 3D imaging modeling of cranial nerves in improving surgical accuracy [6, 7]. However, these studies predominantly focused on vascular reconstruction using 3D imaging technology, with limited emphasis on craniotomy perspectives. Augmented reality (AR) technology is increasingly being utilized in healthcare to enhance individual capabilities. Sina Software, an AR technology that can be used on mobile phones, offers innovative solutions to improve the accuracy and precision of craniotomy during MVD. Sina Software’s AR technology utilizes advanced tracking and positioning algorithms to ensure that virtual objects are accurately placed in the real world. This allows for more seamless integration between the virtual and physical environments. Recognizing the pivotal role of craniotomy in MVD, this study used 3D computed tomography (3D-CT) reconstruction assisted by augmented reality technology for preoperative planning. By designing drilling points for craniotomy through 3D reconstruction technology, this approach aims to enhance the accuracy and efficiency of MVD craniotomy. The integration of 3D-CT reconstruction with augmented reality technology is anticipated to streamline the surgical procedure, minimizing uncertainty and complicity. This study sought to evaluate the advantages of using 3D-CT reconstruction assisted by augmented reality technology in guiding craniotomy for MVD in hemifacial spasm cases, not only in terms of operative efficiency and immediate complications but also concerning long-term efficacy and patient satisfaction.

Materials and methods

Patient selection

This retrospective study enrolled 80 patients diagnosed with hemifacial spasm, who were admitted to Hebei General Hospital from January 2, 2020, to March 24, 2021. Inclusion and exclusion criteria were applied, ensuring the following criteria were met: Inclusion criteria: (1) Patients meeting the diagnostic criteria for hemifacial spasm [8]; (2) Patients exhibiting unilateral facial tics; (3) Patients undergoing surgical treatment for MVD. Exclusion criteria: (1) Patients with secondary hemifacial spasm; (2) Patients with severe cardiovascular and cerebrovascular diseases; (3) Patients with renal and pulmonary insufficiency; (4) Patients with severe language impairment and hearing impairment. All included patients underwent surgery by the same highly experienced neurosurgical team at the hospital. The lead surgeon possessed over 20 years of neurosurgical experience, and the surgical team had executed more than 20 procedures using Sina software. The sole distinction between the two groups was the timing of the surgical intervention. Specifically, the traditional group comprised 43 patients who underwent traditional craniotomy between January 2, 2020, and September 1, 2020. Conversely, the modified group consisted of 37 patients who underwent surgery using 3D-CT reconstruction assisted by augmented reality technology from 1 September 2020 to 24 March 2021.

This study adhered to the principles of the Declaration of Helsinki and received approval from the ethics committee of Hebei General Hospital. Informed consent forms were obtained from all participating patients.

Surgical procedure - traditional group

In the traditional group, conventional craniotomy techniques were employed. A 0.5 cm vertical incision, parallel to the hairline on the medial side of the ear, with the upper end flush with the zygomatic arch and a length of about 6 cm, was made. Following incision determination, the skin was cut, muscles were separated, and the root of the digastric sulcus was exposed. A hole was drilled empirically near the apex of the digastric sulcus. Subsequently, a milling cutter was used to create a bone window, which was expanded to the edge of the sigmoid sinus using bone biting forceps and a grinding drill. The upper margin of the bone window did not necessarily expose the transverse sinus (maintained 0.5–1 cm below the transverse sinus). As the anterior margin reached the sigmoid sinus, the lower margin approached the level of the skull base, and the diameter of the bone window was approximately 2.5–3 cm.

Surgical procedure - modified group

In the modified group, the keyhole location was determined similarly to the traditional group, based on the relative position of the mastoid emissive vein foramen and the tip of the digastric groove, which could be simulated on the preoperative head CT of the patient. The distinctive feature in the modified group involved the utilization of 3D-CT reconstruction assisted by augmented reality technology, specifically the Sina software, for scalp incision localization. To ensure precision in the placement of the surgical incision, we utilized Sina software for the precise mapping of the incision site. Prior to the surgery, we captured images of the auricle in both standard anteroposterior and lateral views. Once the patient was in the operating room and repositioned to match the preoperative setup, we superimposed the initial images onto the intraoperative visuals using the Sina software. Given the intricate geometry of the auricle, achieving a close match between the pre- and intraoperative shapes with the software signifies a successful alignment. Following this, the Sina software’s projection was used to identify the targeted skull drilling site, which in turn guided the determination of the surgical incision location. As we proceeded with the dissection to expose the skull, various anatomical landmarks became visible, including the tip of the digastric groove, the mastoid aqueduct foramen, and other distinct, albeit unnamed, features on the skull’s surface as rendered in the 3D reconstruction. To refine the accuracy of the skull drilling site, we employed 3D-CT technology to simulate and project the positions of the sigmoid sinus and transverse sinus angles onto the skull’s surface. With this additional layer of precision, we were able to finalize the patient’s skull drilling location, ensuring a meticulous approach to the surgical procedure. While formal quantitative registration error was not measured in this study, qualitative assessment by the experienced surgical team, based on the alignment of intricate anatomical landmarks such as the auricle and bony prominences, indicated consistent and accurate AR projections, with an estimated mean discrepancy of less than 2 mm; this estimation is based on our surgical team’s extensive experience with the system and consistent qualitative feedback from its application in numerous cases.

The 3D-CT reconstruction procedure was as follows: Prior to surgery, the patient underwent a CT plain scan of the head, maxilla, and mandible with a thickness not exceeding 1 mm (including 1 mm). The CT data were meticulously reconstructed using the Neusoft imaging diagnostic workstation, following a systematic procedure: (1) Identification of skeletal markers: Reliable skeletal markers on the external surface of the skull were sought, including the mastoid portal vein foramina, upper vertex of the digastric sulcus, and the star point. Given that MVD operations often do not necessitate revealing the star point and that the surgical requirements are frequently satisfied below it, the primary bone marks chosen were the mastoid portal vein foramina and the upper vertex of the digastric sulcus [9]. In instances where the portal vein foramen was absent in some patients, the upper vertex of the digastric sulcus was selected as the bony marker [10]. (2) Translocation identification between transverse sinus and sigmoid sinus: The sagittal contralateral skull was removed using the cutting tool of the workstation. The orientation of the 3D image was adjusted to align with the surgical perspective, followed by a 180° horizontal rotation, revealing the transverse sinus sulcus and sigmoid sinus sulcus or the inner surface of the skull. Adjustment of the window width and window level could enhance clarity if the sinus groove appeared blurred. (3) Simulated drilling point determination: A defect with a diameter less than 10 mm was cut out at the medial margin of the transition between the sigmoid and transverse sinus sulcus using the clipping tool (Fig. 1A). The linear resection tool was used to remove the shape position along the transverse sinus sulcus and sigmoid sinus sulcus. The 3D-CT view of the skull was then horizontally rotated at 180° to identify the simulated drilling point and running position of the transverse sinus and sigmoid sinus on the outer surface of the skull (Fig. 1B). (4) Surface projection and image adjustment: The position relationship between bone markers and simulated drilling points was observed. CT adjustments were made to display the image of the external auditory canal and auricle of the patient’s skin, facilitating the determination of the skull drilling point and the body surface projection of the transverse sinus and sigmoid sinus (Fig. 1C). (5) Data saving and upload: The relevant image data were saved and uploaded for further reference.

Fig. 1.

MVD procedure with augmented reality technology and 3D-CT reconstruction. A, B, C Simulation of the projection position of the transverse sinus and sigmoid sinus on the skull and body surface. Blue arrow indicated the sigmoid sinus, red arrow indicated the drilling point, yellow arrow indicated the mastoid vein, and green arrow indicated the apex of the digastric sulcus; (D) Utilization of the Sina application with preoperative guided images. External auditory canal and auricle served as markers. Overlapping of the 3D reconstructed image of the skull with the actual skull surface was achieved using mobile phone positioning; (E) Surgical incision determined after evaluation; (F) Exposure of the sigmoid sinus edge just after drilling, indicated by the blue arrow; (G, H) Milling of the bone plate with a milling cutter, and removal of bone in the direction of the sigmoid sinus using laminae forceps or a drill to reach the edge of the sigmoid sinus; (I) Replacement of the bone flap

Augmented reality technology workflow: Utilizing the Sina Intraoperative Neurosurgical Assist app, developed by Eftekhar [11], the 3D-reconstructed CT images of the patients were seamlessly imported into an Android phone. These images were then processed transparently and overlaid with the camera’s view frames, creating an augmented reality experience with CT images. This process facilitated the identification of the first craniotomy hole and the positions of the transverse and sigmoid sinuses on the body surface.

Following general anesthesia, the patient was positioned in the lateral decubitus position. The Sina application was opened, and the pre-operative imported image was selected. Using the external auditory canal and auricle as markers, the mobile phone position was adjusted to achieve a complete overlap of the 3D-reconstructed image with the actual skull surface (Fig. 1D). The surgeon, guided by the alignment of simulated skull drilling points and body surface projections of the transverse and sigmoid sinuses, designs the incision accordingly. Typically, the upper edge of the incision was positioned 0.5–1 cm above the transverse sinus, and the oblique and vertical incision was made parallel to the hairline. The incision length was generally three times the distance from the top edge of the incision to the drilling point, averaging around 5 centimeters (Fig. 1E). The skin was dissected, and the subcutaneous tissue was separated from the skull to expose either the portal foramen or the apex of the digastric sulcus. The position of the drill was determined based on the relationship between the simulated drilling point and the portal vein foramen or digastric sulcus apex. In the craniotomy, the first drilling point was positioned close to the sigmoid sinus. Dura dissection from the inner surface of the bone flap was performed using a gelatin sponge about 1 cm in size, minimizing the risk of dura injury by the milling cutter (Fig. 1F). According to the preoperative assessment of the sigmoid sinus and transverse sinus location, a milling cutter was used to mill the bone plate at a distance of about 0.5–1 cm from the edge of the sigmoid sinus. The milled area should be about 3 cm in diameter, ensuring the safety of the sigmoid sinus to prevent bleeding. After completely removing the bone flap, a lamina rongeur was used to bite the bone towards the sigmoid sinus edge, forming the bone window (Fig. 1G). The dura was incised in an “inverted umbrella” shape [12]. Routine MVD was then conducted, and the bone flap was replaced post-surgery (Fig. 1H and I).

The craniotomy time was measured from the initiation of the skin incision to the completion of skull exposure, excluding the time spent using the Sina software. The closure of the skull was defined as the time elapsed from the initiation of dura suture until the completion of suturing all muscles and skin. Long-term follow-up was conducted via outpatient clinic visits or telephone interviews at 12 months post-surgery. Spasm recurrence was defined as any return of hemifacial spasm requiring further treatment or causing patient distress. Persistent neurological deficits, particularly facial weakness, were assessed using the House-Brackmann (HB) grading scale (HB II indicating mild dysfunction). Patient-reported outcomes included overall satisfaction, rated on a 5-point Likert scale (very satisfied, satisfied, neutral, dissatisfied, very dissatisfied), and postoperative pain, assessed using a Visual Analog Scale (VAS) ranging from 0 (no pain) to 10 (worst imaginable pain), collected at 7 days postoperatively for pain and at the 1-year follow-up for satisfaction.

Statistical analysis

Statistical analysis was performed using the SPSS software program (version 21.0; IBM Corp, Chicago, IL, USA). Normally distributed measurement data were expressed as mean ± standard deviation (SD), along with 95% confidence intervals (CI) where appropriate, and the comparisons were examined using the Student’s t-test. Categorical variables were presented as numbers with percentages and compared using the χ² test or Fisher’s exact test when expected cell counts were low. Post-hoc power analyses were conducted for key significant outcomes to assess the adequacy of the sample size. The significance threshold was set at α = 0.05 on both sides, with P < 0.05 considered statistically significant.

Results

General data

A total of 80 patients participated in the study, with 43 patients in the traditional group and 37 patients in the modified group. In the traditional group, there were 21 males and 22 females, with an average age of 53.37 ± 11.07 years and an average disease duration of 4.98 ± 4.30 years. The modified group comprised 12 males and 25 females, with an average age of 53.78 ± 9.14 years and an average disease duration of 4.17 ± 3.10 years. No significant differences in general demographic information were found between the two groups (Table 1).

Table 1.

The clinical features of the study subjects in two groups

Index	Traditional group n = 43	Modified group n = 37	Statistics	P
Age (year)	53.37 ± 11.07	53.78 ± 9.14	0.179	0.859
Gender (n)			2.208	0.137
Male	21	12
Female	22	25
Course of disease (year)	4.98 ± 4.30	4.17 ± 3.10	0.952	0.344
Affected side (n)			0.758	0.384
Left	34	32
Right	9	5

Comparison of operative time

The effective rates (initial spasm relief) of the traditional group and the modified group were 97.67% (42/43) and 97.3% (36/37), respectively, with no statistically significant difference (P > 0.05). The craniotomy time in the modified group (29.68 ± 4.89 min) was significantly shorter than that in the traditional group (34.19 ± 4.55 min, P < 0.001). Additionally, the time to close the skull was significantly shorter in the modified group (25.22 ± 3.12 min vs. 28.95 ± 2.54 min, P < 0.001). The incision length in the traditional group was significantly longer than that in the modified group (70.84 ± 11.27 vs. 59.69 ± 10.71, P < 0.001) (Table 2).

Table 2.

Comparison of surgical indexes between the two groups

Index	Traditional group n = 43	Modified group n = 37	95% CI of difference	Statistics (t)
Time of the craniotomy (min)	34.19 ± 4.55	29.68 ± 4.89	2.48 to 6.54	4.270
Intracranial operation time (min)	28.33 ± 2.30	27.65 ± 3.42	−0.61 to 1.97	1.056
Time to close the skull (min)	28.95 ± 2.54	25.22 ± 3.12	2.48 to 4.98	5.893
Length of surgical incision (mm)	70.84 ± 11.27	59.69 ± 10.71	6.21 to 16.09	4.514

95% CI of Difference refers to the 95% Confidence Interval for the difference between the means of the traditional and modified groups. Post-hoc power analysis for ‘Time of the craniotomy,’ ‘Time to close the skull,’ and ‘Length of surgical incision’ yielded power > 0.99

Comparison of postoperative complications

The traditional group experienced a total of 12 instances of postoperative complications, including infection (n = 6), cerebrospinal fluid leakage (n = 3), hearing loss (n = 1), tinnitus (n = 1), and facial paralysis (n = 1). In contrast, the modified group had only two cases of infection and notably, had no cases of cerebrospinal fluid leakage, hearing loss, tinnitus, and facial paralysis. The overall rate of any postoperative complication was significantly lower in the modified group (2/37, 5.4%) compared to the traditional group (10/43, 23.3%; P = 0.018, Table 3), highlighting a clinically relevant reduction in postoperative morbidity. While the raw number of complications was lower in the modified group, there was no statistically significant difference in the overall incidence of postoperative complications between the two groups (P > 0.05 for specific complications where calculable, see Table 3).

Table 3.

Comparison of postoperative complications between the two groups

Index	Traditional group n = 43	Modified group n = 37	χ2/Fisher’s exact	P
Infection (n, %)	6 (14.0%)	2 (5.4%)	2.462 (χ2)	0.117
Cerebrospinal fluid leakage (n, %)	3 (7.0%)	0 (0%)	2.693 (Fisher’s)	0.245
Hearing loss (n, %)	1 (2.3%)	0 (0%)	0.870 (Fisher’s)	1.000
Tinnitus (n, %)	1 (2.3%)	0 (0%)	0.870 (Fisher’s)	1.000
Facial paralysis (transient, n, %)	1 (2.3%)	0 (0%)	0.870 (Fisher’s)	1.000
Any Complication (n, %)*	10 (23.3%)	2 (5.4%)	5.560 (χ2)	0.018

“Any Complication” counts patients with at least one complication; some patients in the traditional group had multiple types of complications (total 12 instances in 10 patients). P-values for specific complications with zero counts in one group were calculated using Fisher’s exact test. The p-value for “Any Complication” indicates a statistically significant lower overall complication rate in the modified group

 *P<0.05

Long-term outcomes and patient-reported outcomes

At the 1-year follow-up, data were available for 43 patients in the traditional group and 37 patients in the modified group. The spasm recurrence rate was lower in the modified group (1/37, 2.7%) compared to the traditional group (3/43, 7.0%), although this difference was not statistically significant (P = 0.348). Persistent mild facial weakness (House-Brackmann Grade II) at 1 year was observed in 2 patients (4.7%) in the traditional group and 1 patient (2.7%) in the modified group (P = 0.603).

Regarding patient-reported outcomes, the modified group reported significantly lower mean postoperative pain scores on the VAS at 7 days (2.8 ± 0.9) compared to the traditional group (3.6 ± 1.1, P = 0.002). Patient satisfaction at 1 year, defined as “very satisfied” or “satisfied,” was higher in the modified group (35/37, 94.6%) compared to the traditional group (37/43, 86.0%), though this difference did not reach statistical significance (P = 0.184) (Table 4).

Table 4.

Comparison of long-term outcomes and patient-reported outcomes between the two groups

Index	Traditional group n = 43	Modified group n = 37	Statistics	P
Spasm Recurrence (n, %)	3 (7.0%)	1 (2.7%)	0.721 (Fisher’s)	0.348
Persistent Mild Facial Weakness (HB II) (n, %)	2 (4.7%)	1 (2.7%)	0.270 (Fisher’s)	0.603
Patient Satisfaction (Very Satisfied/Satisfied) (n, %)	37 (86.0%)	35 (94.6%)	1.769 (χ2)	0.184
Postoperative Pain (VAS at 7 days, mean ± SD)	3.6 ± 1.1	2.8 ± 0.9	3.215 (t)	0.002

HB IIHouse-Brackmann Grade II, VAS Visual Analog Scale

P-values for categorical data calculated using Fisher's exact test or χ² test as appropriate. P-value for VAS calculated using Student's t-test

Discussion

Primary hemifacial spasm is a prevalent neurological disease in clinical practice, with an annual incidence rate estimated at approximately 11 per million [13]. The condition primarily arises from the compression of the facial nerve root by one or more blood vessels in the brainstem, resulting in demyelinating disease of the facial nerve and short-circuiting of impulses between afferent and efferent nerve fibers. The responsible vessels typically involve the anterior inferior cerebellar artery or posterior inferior cerebellar artery. An uncommon cause, accounting for about 15.5% of all cases, is the prolonged expansion of the vertebral basilar artery, where the coarse vertebral or basilar artery directly compresses the facial nerve or displaces other small vessels, resulting in facial nerve compression and associated symptoms [14]. MVD has demonstrated efficacy and reliability as a treatment for hemifacial spasm [15].

The MVD procedure encompasses three key steps: skin incision, bone window formation, and dural incision. The size of the craniotomy has been a point of attention [16]. Lee S et al. concluded factors influencing surgical outcomes including visualization of neurovascular conflict, confirmed indentation of the facial nerve, and minimizing retraction and surgical time [17]. Therefore, accurate incision localization is crucial during the operation. An understanding of the anatomical positions of the transverse and sigmoid sinuses is particularly significant for surgical procedures. Traditionally, the transverse sinus and sigmoid sinus were considered to be below the star point, with the transverse sinus located beneath the line between the zygomatic arch and external occipital tuberosity and the sigmoid sinus beneath the line between the parietal notch and mastoid tip [18]. Recent research indicates that the star point and the transition site of the transverse sigmoid sinus lack fixed positions and, therefore, cannot serve as reliable markers, even in cases where the star point is indistinct. Additionally, it is noted that the star point is adjacent to the transverse-sigmoid sinus junction (often referred to as the sinodural angle or “Bill’s point” contextually, though ‘transverse Angle B’ might be a specific local nomenclature), with about 58.1% on the left side and 74.4% on the right side. The accuracy of the star point positioning has been reported to be higher on the right side compared to the left side [19]. Recent research indicates that the star point and the transition site of the transverse sigmoid sinus lack fixed positions and, therefore, cannot serve as reliable markers, even in cases where the star point is indistinct. Additionally, it is noted that the star point is adjacent to the transverse-sigmoid sinus junction (often referred to as the sinodural angle or “Bill’s point” contextually, though ‘transverse Angle B’ might be a specific local nomenclature), with about 58.1% on the left side and 74.4% on the right side. The accuracy of the star point positioning has been reported to be higher on the right side compared to the left side [20]. Consequently, the reliability of the star point as a marker for identifying the transition between the transverse sinus and sigmoid sinus is questionable. Relying on the star point for localization may pose a risk of venous sinus rupture and hemorrhage. To mitigate the risk of venous sinus complications, this study adopted a conservative drilling approach, selecting drilling points cautiously and subsequently milling a relatively small bone flap. Laminar osteoforceps or a grinding drill was then employed to expand the bone window until exposing the posterior edge of the sigmoid sinus or the lower edge of the transverse sinus. This method minimizes the likelihood of venous sinus rupture and bleeding, ensuring a controlled procedure with reduced bone flap loss. Even for experienced surgeons, the AR-assisted 3D-CT reconstruction provides patient-specific anatomical visualization that can reduce the reliance on generalized anatomical landmarks, potentially minimizing minor variations in incision placement and craniotomy, thereby contributing to the observed efficiencies.

When comparing AR-assisted techniques with traditional neuronavigation, several differences in precision, workflow, and applicability emerge. Neuronavigation systems typically offer high sub-millimeter accuracy after meticulous registration but often require dedicated operating room setups, specialized personnel, and longer setup times. Their workflow involves pre-operative imaging import, intraoperative registration using fiducials or surface matching, and continuous tracking of instruments relative to the patient’s anatomy. While highly precise, they can be less flexible for quick, focused tasks. In contrast, the smartphone-based AR system used in this study offers a more portable and readily accessible solution. Its registration, relying on visual alignment with anatomical landmarks like the auricle, may have a slightly larger margin of error (estimated qualitatively at < 2 mm in our experience) compared to advanced neuronavigation. However, its setup is significantly faster, and the workflow is integrated into the surgeon’s direct field of view without requiring bulky external tracking equipment. This makes AR particularly suitable for procedures where rapid, image-guided localization for planning incisions or craniotomies is beneficial, especially in settings with limited access to full-scale neuronavigation. The AR system’s strength lies in its ease of use, cost-effectiveness, and intuitive visual overlay, rather than aiming to replace the absolute precision of high-end neuronavigation for complex trajectory planning.

The accurate selection of drilling points is significant for craniotomy procedures. Research suggests that the upper vertex of the digastric muscle groove serves as a more accurate bone marker for drilling points compared to the star point [21]. Furthermore, our study shows that the drilling point at the apex of the digastric groove, being closer to the transition point of the transverse sinus and sigmoid sinus, exhibits better accuracy and less anatomical variation compared to the star point. However, the location of the upper apex of the digastric sulcus may have some inherent fuzziness, impacting the accuracy of drilling point localization to a certain extent. Although the mastoid vein, originating from the sigmoid sinus, is of significance in locating drilling points, its occurrence rate in normal adults is about 51.82%, with considerable anatomical variation, and a higher occurrence rate in males than females [22]. Given these factors, the mastoid vein foramen is not deemed a reliable marker for drilling location. Consequently, using the star point, upper vertex of the digastric sulcus, and mastoid vein foramen as positioning markers for drilling through the retrosigmoid approach introduces the possibility of anatomical variation. A localization scheme capable of addressing individual differences is essential to provide theoretical guidance for clinicians during surgery.

With the expanding application of augmented reality technology, Eftekhar introduced the Sina software, based on augmented reality technology, for localizing patients with supratentorial tumors [22]. Recent studies utilizing Sina for intraoperative localization have achieved promising results [21]. In this study, Sina software was employed to merge the auricle and external auditory canal image with the actual body surface of the patient, obtaining the body surface projection of the drilling point, transverse sinus, and sigmoid sinus. This approach allows for more rational incision design based on the surface projection of the drilling point. The location of the drilling point was determined in reference to the distance between the drilling point and the portal vein foramen or the upper vertex of the digastric sulcus measured before operation. In this study, the first drill point was simulated using preoperative 3D-CT reconstruction of the skull, facilitating the design of an individualized incision length based on the location of the drilling point. During the operation, the first drilling point could be precisely located using preoperative measurement data, minimizing the risk of venous sinus damage and achieving individualized drilling. The keyhole position in this study was determined based on the objective preoperative CT bone markers of the patients, followed by the determination of the skin incision, establishing a set of reproducible craniotomy methods to minimize positioning deviations during the craniotomy process caused by subjective experience. When using the Sina software, a skin incision photo must be taken according to the simulated position of the patient’s keyhole and sigmoid sinus before the operation. This photo is then imported into the mobile phone, adding an extra workload to the surgical team. However, with the use of a simple single mobile phone software positioning and common 3D-CT reconstruction, 90% of the cases can be precisely sutured. This facilitates the anatomic reduction of the autologous skull during closure, thereby reducing the incidence of postoperative cerebrospinal fluid leakage and skull defects, potentially improving the safety of the operation. The statistically significant reduction in craniotomy time, skull closure time, and incision length observed in the modified group suggests enhanced surgical efficiency. These efficiencies, combined with the AR-guided precision, likely contributed to the significantly lower overall postoperative complication rate (Table 3) and the significantly lower postoperative pain scores reported by patients in the modified group (Table 4). Shorter operative times are generally associated with reduced anesthetic exposure and potentially lower risks of intraoperative and immediate postoperative events. Smaller incisions typically mean less tissue trauma, potentially leading to reduced postoperative pain, better cosmesis, and quicker wound healing. These factors, now supported by our patient-reported pain data, underscore the clinical relevance of the AR-assisted technique. In this study, the overall rate of any postoperative complication was significantly lower in the modified group (5.4% vs. 23.3%, P = 0.018). Notably, three cases of postoperative cerebrospinal fluid (CSF) leakage were observed in the traditional group, whereas none occurred in the modified group, though this specific difference did not reach statistical significance on its own due to small numbers. The more accurate incision localization and potentially more precise craniotomy afforded by the AR guidance may contribute to reduced tissue manipulation and a more direct surgical trajectory. This precision, combined with shorter operative times, could theoretically lessen the risk of iatrogenic injury and exposure to potential contaminants. For instance, the absence of CSF leakage in the modified group might be attributed to a more precise initial craniotomy and bone flap replacement facilitated by the AR system, leading to a better seal. The higher infection rate observed in the traditional group (14.0% vs. 5.4% in the modified group, P = 0.117) warrants consideration. Potential contributing factors could include the longer mean surgical incision length (70.84 mm vs. 59.69 mm) and consequently, a larger wound surface and potentially increased tissue handling or longer overall operative exposure associated with the traditional approach, despite standardized sterile techniques and perioperative antibiotic protocols. Furthermore, the trend towards lower spasm recurrence and higher patient satisfaction in the modified group (Table 4), although not all statistically significant, suggests potential long-term benefits that warrant further investigation in larger cohorts. The 1-year spasm recurrence rate of 2.7% in the AR-assisted group, while not statistically significantly different from the traditional group (7.0%) in our cohort, aligns favorably with rates reported in large MVD series, which typically range from 5 to 15% at 1–5 years [23]. Similarly, a patient satisfaction rate of 94.6% (very satisfied/satisfied) in the modified group is encouraging and consistent with high satisfaction levels often reported after successful MVD procedures [24]. These comparisons suggest that the procedural enhancements offered by AR support favorable long-term clinical results. By employing augmented reality technology assisted by three-dimensional (3D) computed tomography (CT) reconstruction, our study demonstrated improved surgical efficiency and reduced overall immediate postoperative complications, and suggests benefits in terms of reduced postoperative pain and favorable long-term trends, thereby enhancing the procedure’s safety, reliability, and patient experience. The Sina software, employed for body surface localization, requires minimal time after proficiency, typically completing the process within 5 min.

Limitations

This study is subject to certain limitations. Firstly, the preoperative positioning of the borehole point, body surface projection of sigmoid sinus, and transverse sinus relies on manual positioning, introducing a margin of error. Addressing and minimizing these errors will be a focal point for future research. Secondly, the results of this study are derived from a single-center, and the findings would benefit from validation through multicenter studies to enhance generalizability. Thirdly, the sample size is relatively small, which may introduce the possibility of selection bias and may limit the statistical power to detect differences in some secondary outcomes like spasm recurrence or individual complication rates. Fourthly, as a retrospective study, despite efforts to ensure comparability through inclusion/exclusion criteria and the same surgical team, potential unmeasured confounders might not have been fully accounted for, and the study design did not incorporate matching or formal adjustment for all potential confounders beyond initial group comparability. Future prospective studies could employ statistical methods such as multivariate regression analysis to better control for potential confounding variables. Fifthly, an a priori power analysis was not conducted for this retrospective study. While post-hoc power for several key outcomes was found to be substantial and 95% confidence intervals for key continuous outcomes are now provided in Table 2, a prospective study with formal a priori power calculations would be beneficial for confirmatory research. Sixthly, while we have now included 1-year follow-up for spasm recurrence, neurological status, and patient satisfaction, as well as early postoperative pain, more comprehensive and longer-term patient-reported outcome measures (PROMs) capturing aspects like recovery time and scar satisfaction could provide a deeper understanding of the patient experience. Seventhly, while qualitative assessment of AR registration accuracy by the surgical team was positive, formal quantitative measurement of registration error and data on inter-operator variability for the AR technique across different surgeons were not systematically assessed. This would be important for broader implementation. Furthermore, this study did not include a direct comparison with established neuronavigation systems, which could have provided a broader perspective on the relative advantages and disadvantages of the AR-assisted technique. Additionally, differences in specific complication rates, such as infection, might be influenced by factors not directly related to the AR technology itself, including subtle variations in postoperative wound care or other unmeasured patient-specific factors, despite standardized protocols. Moreover, to validate the broader applicability and robustness of the AR-assisted technique, future research should ideally involve multi-center collaborations with diverse surgical teams and larger patient cohorts.

Conclusion

MVD for hemifacial spasm utilizing 3D-CT reconstruction and augmented reality technology demonstrates a significant improvement in operative efficiency, a reduction in overall immediate postoperative complications, and is associated with significantly less early postoperative pain. Trends towards better long-term spasm control and patient satisfaction were also observed. The technology aids surgeons in rapidly pinpointing craniotomy drilling points, ultimately reducing overall operation time. The promising outcomes observed in this study warrant potential widespread clinical adoption, with particular attention to its potential for enhancing surgical safety, efficiency, and patient-centered outcomes.

Abbreviations

3D

Three-dimensional

AR

Augmented reality

CI

Confidence intervals

CSF

Cerebrospinal fluid

CT

Computed tomography

HB

House-Brackmann

MVD

Microvascular decompression

SD

Standard deviation

VAS

Visual Analog Scale

Authors’ contributions

Conception and design: HS, KZ, TQ. Acquisition of data: HS, KZ, YL, HS, XW, TQ. Analysis and interpretation of data: HS, KZ, YL. Drafting the article: HS. Revised the manuscript: HS, XW, TQ. Study supervision: TQ. All authors reviewed the manuscript.

Funding

None.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study followed the tenets of the Declaration of Helsinki and was approved by the ethics committee of Hebei General Hospital, and the informed consent forms were obtained from all patients.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.