Preoperative Widefield Swept-Source Optical Coherence Tomography Versus Intraoperative Findings in Detecting Posterior Vitreous Detachment

Abstract

Purpose: To assess the accuracy of swept-source optical coherence tomography (SS-OCT) in detecting complete posterior vitreous detachment (PVD) in comparison with intraoperative findings.

Methods: The retrospective study included 145 eyes of 145 consecutive patients who underwent surgery for epiretinal membranes or macular holes. Within a week prior to surgery, PVD status was evaluated by SS-OCT with a depth of field of 3 mm and a capture window of 16 × 8 mm. Complete PVD was identified when the hyaloid condensation was visible clearly on any B-scan or when the vitreous cortex reflectivity was not visible on all 33 B-scans. Sensitivity, specificity, positive predictive value, and negative predictive value of SS-OCT for detection of complete PVD were then compared with those evaluated during a triamcinolone acetonide-assisted vitrectomy.

Results: Of the 101 eyes diagnosed as complete PVD by SS-OCT preoperatively, 97 eyes were found to have complete PVD and four eyes were found to have attached vitreous intraoperatively. Of the 44 eyes categorized as attached vitreous by SS-OCT preoperatively, 43 eyes were graded as attached vitreous and one eye was graded as complete PVD during surgery. The sensitivity of SS-OCT for detecting complete PVD was 99.0% and the specificity was 91.5%. The positive predictive value and the negative predictive value were 96.0% and 97.7%, respectively.

Conclusions: Widefield (16 × 8 mm) SS-OCT showed high accuracy for the diagnosis of complete PVD in patients with epiretinal membranes or macular holes.

Translational Relevance: Widefield SS-OCT has great potential to evaluate PVD status preoperatively and explore the mechanisms of vitreoretinal diseases.

Introduction

A posterior vitreous detachment (PVD) is a physiologic process of the separation of the posterior vitreous cortex from the internal limiting membrane of the retina, which results from vitreous gel liquefaction and weakening of vitreoretinal adhesion.¹ The incidence of PVD increases with age, and a PVD typically occurs between the ages of 45 and 65 years in the general population.² Although PVD is considered a physiological process that occurs naturally with age, the progression of PVD plays a pivotal role as a pathogenic factor in various vitreoretinal disorders, including epiretinal membrane,³ macular hole,⁴ vitreomacular traction,⁵ retinal tears,⁶ and retinal detachments.⁷ Moreover, the absence of preoperative PVD may increase the risk of iatrogenic retinal tears during vitrectomy.⁸ Thus, an accurate evaluation of the preoperative PVD status is important. 

Slit-lamp biomicroscopy or indirect ophthalmoscopy is an efficient approach to diagnose complete PVD by the presence of a Weiss ring. However, it may be difficult to ascertain the vitreous status in the absence of a Weiss ring.⁹ B-scan ultrasonography (US) is known to be useful in assessing PVD status in a real-time kinetic mode, regardless of media opacity or inadequate pupil dilation.⁹ However, it is a contact technique and depends somewhat on the skills of the investigator.¹⁰ 

As a relatively objective and noncontact technique, optical coherence tomography (OCT) has been reported to be useful for PVD evaluation.^9,10 Advances in OCT-assisted vitreous imaging technique include positioning the retinal layer inferiorly to allow maximum imaging depth into the vitreous and adjusting image brightness to improve visualization^11,12 Previous studies have compared the diagnostic ability between spectral-domain OCT (SD-OCT) with ultrasound,^9,10,13 slit-lamp biomicroscopy,^9,13 and intraoperative findings in detecting PVD status. A high agreement for complete PVD diagnosis was observed among SD-OCT, ultrasound, and slit-lamp biomicroscopy,^9,10 whereas OCT was less reliable.¹³ Compared to SD-OCT, swept-source OCT (SS-OCT) has an increased scan speed, deeper penetration, and a wider capture window, which enable better visualization of not only PVD status but also additional details of the vitreoretinal interface.^14,15 Accurate preoperative evaluation of PVD status could play a crucial role in minimizing the occurrence of iatrogenic retinal breaks during surgery while also facilitating further investigations into the role of PVD in the pathogenesis of epiretinal membrane and macular hole. Therefore, the aim of this study was to evaluate the accuracy of preoperative SS-OCT for assessing PVD status in comparison with intraoperative findings observed during vitrectomy. 

Methods

Patients and Study Design

We performed a retrospective review of medical records of patients who underwent surgery for epiretinal membrane or macular hole at Zhongshan Ophthalmic Center from February 2022 to August 2023. The study adhered to the tenets of the Declaration of Helsinki, and was approved by the institutional review board of Zhongshan Ophthalmic Center (2023KYPJ347). 

The inclusion criteria were as follows: patients diagnosed with epiretinal membrane or macular hole identified by an experienced specialist on biomicroscopic examination and confirmed by SD-OCT (SPECTRALIS; Heidelberg Engineering, Heidelberg, Germany) and age between 18 and 85 years old. The exclusion criteria were as follows: coexistence of retinal detachment, previous history of vitrectomy, spherical equivalent ≤−6.0 diopters (D), or poor image quality of OCT (signal strength index ≤6). All patients underwent routine preoperative examinations, including best-corrected visual acuity (BCVA), slit-lamp biomicroscopy with a 90-D lens, SD-OCT, and SS-OCT (VG200S; SVision Imaging, Guangdong, China). If both of the eyes of a patient who underwent vitrectomy during this period met eligibility criteria, then one eye was selected randomly. 

PVD Grading by SS-OCT

Within a week prior to surgery, all of the eyes were imaged through a dilated pupil with a new swept-source OCT (VG200S) by two trained operators. The VG200S is equipped with an eye-tracking system based on an integrated confocal scanning laser ophthalmoscope (cSLO) to eliminate eye-motion artifacts. The product uses a swept-source laser with a central wavelength of approximately 1050 nm and an acquisition rate of 200,000 A-scans per second. It operates at a designed power level of 3.4 mW with a scan depth of 3 mm. The axial resolution and lateral resolution were 5 µm and 13 µm, respectively. The OCT images, centered on the foveola, were captured through a 16 × 8-mm window, covering both the macular area and the optic nerve head. The scan range involves the acquisition of 33 horizontal B-scans with a spacing of 0.25 mm between each scan line. To achieve maximum imaging depth into the vitreous, the retinal layers were positioned inferiorly. Built-in software allowed for optimizing the brightness of the OCT images to visualize the vitreous. The brightness was adjusted similarly for all images. The method for capturing SLO and SS-OCT images is shown in Figure 1. 

Figure 1.

 

A normal eye was imaged through a dilated pupil using SS-OCT coupled with a simultaneous SLO device. The SLO image (16 × 16 mm) and the OCT image were both centered on the foveola (the intersection of the two white lines). The 5.5-mm-diameter red circle is the macula. Thirty-three OCT horizontal line scans were acquired through a capture window of 16 × 8 mm (the field between the two green lines), covering both the macula and the optic nerve head. The retinal layers were positioned inferiorly to achieve maximum imaging depth into the vitreous. Built-in software allowed for optimizing the brightness of the OCT images to visualize the vitreous.

In addition to SS-OCT, SD-OCT images of each eye were captured as a routine examination within 1 month before surgery. The surgeon reviewed the SD-OCT images before surgery but was blind to the SS-OCT images. SD-OCT was captured with SPECTRALIS OCT using an 870-nm wavelength laser with a scan depth of 1.8 mm and scan speed of 40,000 A-scans per second. The axial resolution was 7 µm, and the lateral resolution was 14 µm. Horizontal and vertical 9-mm B-scans centered on the foveola were captured. The retinal layers were positioned in the center of the images. 

Two examiners independently evaluated PVD stage by reviewing all 33 horizontal B-scans using a modified version of a previously described staging system,¹⁶ which was defined as follows: stage 0, no PVD present either overlying the macula or optic nerve head; stage 1, vitreous separation at the extrafoveal posterior pole while the vitreous remained attached at the fovea and optic nerve head; stage 2, partial foveal vitreous separation was observed while the vitreous remained attached at the foveola and optic nerve head; stage 3, complete foveal vitreous separation with total foveolar separation was observed while the vitreous remained attached at the optic nerve head; stage 4, complete PVD. Complete PVD was identified when the hyaloid condensation was visible clearly on any B-scan or when the vitreous cortex reflectivity was not visible on all 33 B-scans. In our study, because it is difficult to distinguish among stages 0, 1, and 2 during surgery, we combined these three stages into “stages 0–2” for clarity. Stages 0 to 3 were categorized together as attached vitreous, and stage 4 was categorized separately as complete PVD. PVD stages assessed by SS-OCT are shown in Figure 2. OCT images were similarly optimized by adjusting the brightness and contrast using built-in software. Representative cases are shown in Figure 3 to illustrate the results of PVD grading before and after brightness optimization. 

Figure 2.

 

PVD stages in four patients with epiretinal membranes (A–D) and in four patients with macular holes (E–H). The attached vitreous includes stages 0–2 (A, B, E, F) and stage 3 (C, G). Complete attached vitreous with premacular bursa is shown in a 31-year-old male with epiretinal membrane (A) and in an 8-year-old female with macular hole (E). Attached vitreous to both the macula and the optic nerve was observed in a 71-year-old male with epiretinal membrane (B) and a 58-year-old female with macular hole (F). Detached hyaloid in the whole premacular space and attached hyaloid to the optic nerve head were seen in a 67-year-old male with epiretinal membrane (C) and a 65-year-old male with macular hole (G). Vitreous cortex reflectivity was not visible on all 33 scans in a 54-year-old female with epiretinal membrane (D) and a 58-year-old female with macular hole (H), which were classified as stage 4 (complete PVD).

Figure 3.

 

Representative cases demonstrating PVD staging before and after brightness enhancement. Case 1 was classified as PVD stage 3 before enhancing the brightness, with the sign of vitreous attachment to the optic nerve head (A1). After the brightness was enhanced, Case 1 was categorized as stages 0–2 because the vitreous was attached to both the optic nerve head and the fovea (A2). Case 2 was graded as PVD stage 4 because of absence of vitreous cortex reflectivity on all 33 B-scans (B1). After the brightness was enhanced, Case 2 was graded as PVD stage 4 because the hyaloid condensation was clearly visible (B2).

PVD Grading by Surgical Examination

All 145 surgeries were performed by one experienced surgeon. Triamcinolone acetonide (TA) was used to assist assessing PVD stages during vitrectomy. Standard three-port sclerotomies were created using 25-gauge instruments (Alcon USA, Fort Worth, TX). The surgeon used low suction pressure to prevent the vitreous from detaching during core vitrectomy. After core vitrectomy, 0.2 to 0.3 mL TA was injected onto the posterior pole to help visualize posterior vitreous status. We focused on the presence or absence of attached vitreous in the macular area and the optic nerve head. Attached vitreous was recorded when a firm attachment had to be removed by vitreous cutter suction. Intraoperative PVD grading was as per the PVD grading by SS-OCT described above. The stages 0–2 category was defined as when attached vitreous was seen both on any location of both the macular area and the optic nerve head. Stage 3 was defined as when attached vitreous was detected on the optic nerve head but not be seen on the macular area. Stage 4 was defined as when attached vitreous was not detected on the macular area or the optic nerve head. Stages 0 to 3 were categorized together as attached vitreous, and stage 4 was categorized separately as complete PVD. The operating surgeon noted the status of vitreous attachment at the macular and the optic nerve head immediately after surgery. 

Statistical Analysis

Continuous data are reported as mean ± standard deviation, and percentage distributions were computed for the demographic variables. Interobserver agreement was calculated using kappa correlation. The accuracy of SS-OCT in detecting complete PVD, compared with intraoperative findings, was assessed by sensitivity, specificity, positive predictive value, and negative predictive value. The rate of complete PVD was compared between epiretinal membranes and macular holes using Pearson’s χ² test. All statistical analyses were performed using SPSS Statistics 27.0 (IBM, Chicago, IL). 

Results

Baseline Characteristics of the Patients

A total of 145 eyes from 145 patients (97 women, 48 men) with an average age of 63.83 years (range, 31–80 years) were included in the study. Of the 145 eyes, 90 eyes (62.1%) were diagnosed as epiretinal membrane and 55 eyes (37.9%) as macular hole. Of these 145 eyes, 130 eyes were phakic and 15 eyes were pseudophakic. Demographic data are summarized in Table 1. 

Table 1.

 

Baseline Characteristics of the Patients

 

PVD Staging by SS-OCT and Interobserver Agreement

The interobserver agreement for PVD grading based on SS-OCT was high (kappa = 0.985, P < 0.001) (Table 2). Among the 145 eyes assessed, the PVD grading of 144 eyes was consistent between the two observers (99.3% agreement) (Table 2). By SS-OCT, 11 eyes (7.6%) were graded as PVD stages 0–2, 33 eyes (22.8%) as stage 3, and 101 eyes (69.7%) as stage 4. Forty-four eyes (30.3%) were graded as attached vitreous and 101 eyes (69.7%) as complete PVD (Table 3). 

Table 2.

 

Interobserver Agreement for Grading PVD by SS-OCT

 

Table 3.

 

PVD Staging by SS-OCT Versus Intraoperative Examination

 

PVD Staging by Intraoperative Examination

During vitrectomy, 12 eyes (8.3%) were graded as PVD stages 0–2, 35 eyes (24.1%) as stage 3, and 98 eyes (67.6%) as stage 4. Forty-seven eyes (32.4%) showed attached vitreous, and 98 eyes (67.6%) showed complete PVD (Table 3). 

PVD Staging by SS-OCT Versus Intraoperative Examination

Out of the 101 eyes classified as complete PVD by SS-OCT, 97 eyes were found to have pre-existing complete PVD at the time of surgery (true-positive results), and four eyes were found to have attached vitreous during surgery (false-positive results). Out of the 44 eyes categorized as attached vitreous by SS-OCT, 43 eyes showed attached vitreous at the time of surgery (true-negative results), and one eye showed pre-existing complete PVD at the time of surgery (false-negative results). The sensitivity of SS-OCT for detecting complete PVD was 99.0%, and the specificity was 91.5%. The positive predictive and negative predictive values were 96.0% and 97.7%, respectively. Detailed data are shown in Tables 3 and 4. 

Table 4.

 

Test Characteristics of SS-OCT for Detection of Complete PVD in Comparison With Intraoperative Findings

 

This study had one false-negative case and four false-positive cases. The false-negative case showed obvious vitreous attachment on eight of all 33 B-scans and detached vitreous on 25 of all 33 B-scans (Fig. 4). However, attached vitreous was not detected on the macular area or the optic nerve head during PPV surgery in this case. Four false-positive cases showed an absence of the vitreous cortex reflectivity on all 33 B-scans, but attached vitreous was detected intraoperatively. These four false-positive cases were all phakic eyes; two cases were diagnosed with epiretinal membrane and two cases with macular hole. 

Figure 4.

 

Representative images of a false-negative case (A1, A2) and a false-positive case (B1, B2). The false-negative case was a 69-year-old female diagnosed with epiretinal membrane, showing both obvious vitreous attachment (A1) and detached vitreous (A2) on SS-OCT B-scans. However, in this case, attached vitreous was not detected in the macular area or the optic nerve head during PPV surgery. The false-positive case was an 80-year-old male diagnosed with epiretinal membrane, showing complete PVD on SS-OCT B-scans (B1). However, attached vitreous was detected after TA staining intraoperatively (B2).

Comparison of PVD Staging Between Eyes With Epiretinal Membranes and Eyes With Macular Holes

PVD status with epiretinal membranes and status with macular holes were compared. At the time of surgery, complete PVD was found significantly more frequently in eyes with epiretinal membranes (90.0%) compared with those with macular holes (30.9%; P < 0.001). In eyes with epiretinal membranes, the sensitivity of preoperative SS-OCT in diagnosing complete PVD was 98.8%, specificity was 77.8%, positive predictive value was 97.6%, and negative predictive value was 87.5% (Table 4, Supplementary Table S1). In eyes with macular holes, the sensitivity of preoperative SS-OCT in diagnosing complete PVD was 100.0%, specificity was 94.7%, positive predictive value was 89.5%, and negative predictive value was 100.0% (Table 4, Supplementary Table S2). 

Discussion

Posterior vitreous detachment plays a crucial role in the pathogenesis of various vitreoretinal disorders. Accurate evaluation of PVD is important for determining the stage of idiopathic macular holes and planning appropriate treatment strategies.¹⁷ In cases of macular holes without PVD, incomplete removal of the vitreous during surgery can affect hole closure. Moreover, PVD was regarded as the most critical event leading to the formation of idiopathic epiretinal membranes.¹⁸ However, the existing theories cannot explain how the epiretinal membrane forms in the absence of PVD. Thus, tracking the relationship between development of the epiretinal membrane and PVD status is essential for elucidating the pathogenic mechanisms underlying the epiretinal membrane. 

Both biomicroscopy and US depend on the investigator's skills to assess PVD status, and the results can be subjective. Rapid advancements in OCT, characterized by enhanced scanning depth and superior resolution, not only allow the visualization of a larger vitreous field but also offer additional details regarding the vitreoretinal interface. These advancements include the identification of such features as premacular bursa, Cloquet's canal, and vitreoschisis, which has proven to be valuable in assessment of the status of PVD.¹⁴ With the added benefits of non-contact operation and rapid scanning capabilities, OCT has emerged as an invaluable tool for assessing the status of PVD. 

This study aimed to compare the identification of complete PVD using widefield SS-OCT with intraoperative findings. Our data reveal a high sensitivity (99.0%) and specificity (91.5%) for SS-OCT in diagnosing complete PVD. In a retrospective study of 175 patients, Hwang et al.¹⁹ conducted a comparison between SD-OCT of the macular region (6-mm scan length) and intraoperative observations to assess PVD status. They reported a sensitivity of 71% and a specificity of 88% for SD-OCT in detecting complete PVD. The authors underscored that the SD-OCT depth of field (approximately 1 mm) was a notable limitation of their study.¹⁹ This restricted depth may result in an incomplete depiction of premacular bursa or posterior vitreous cortex, potentially leading to a false-positive diagnosis of complete PVD. Furthermore, the limited visualization of vitreous over the macular region in SD-OCT may lead to misinterpreting partial PVD as complete PVD, especially in cases lacking evidence of premacular bursa or posterior vitreous cortex but existing persistent vitreous adhesion to the optic nerve head. Moon et al.¹⁰ complemented these findings by integrating peripapillary OCT, which enhanced PVD determination in cases where vitreous status appears ambiguous on transverse images, thereby improving the diagnostic capabilities of SD-OCT. Li et al.²⁰ developed novel deep learning algorithms to automate the detection of PVD using OCT volume scans covering a 6 × 6-mm area centered on the foveola, achieving an accuracy of 90.7% with a sensitivity of 100% and a specificity of 74.5%. 

Compared to SD-OCT, SS-OCT has a significantly increased scan speed and axial depth, thus enabling rapid and precise imaging of both the macula and optic nerve by encompassing a vertical span from the deep retina to the vitreous cavity.¹⁴ The greatest challenge for OCT in PVD evaluation is differentiating between completely attached vitreous and complete PVD. In our study, SS-OCT images were captured with a scan depth of 3 mm and a scan window of 16 × 8 mm. The increased scan depth reduced the false-positive rate of complete PVD diagnosis, as it facilitated the visualization of the entire outline of the premacular bursa. The wider scan window covered both the macular region and the optic nerve head, which significantly increased the sensitivity of SS-OCT for detecting complete PVD.¹⁴ In previous studies using SD-OCT, the diagnosis of complete PVD depended on the absence of the premacular bursa and posterior vitreous cortex. In this study, the increased scan depth and SS-OCT window improved the visibility of the hyaloid condensation, which contributed to identifying complete PVD with greater confidence. 

Previous studies have reported that SD-OCT offers advantages over biomicroscopy and ultrasound in detecting flat and shallow partial PVD.^9,10 In a prospective study involving 95 eyes, Wang et al.¹⁵ reported that SS-OCT with a 16-mm horizontal line scan exhibited high inter-rater reliability and was comparable to B-scan ultrasound imaging and biomicroscopy in diagnosing PVD. Consistent with these findings, our study utilized a total of 33 SS-OCT horizontal line scans that comprehensively covered both the entire macular area and the optic nerve head. This approach yielded high-quality images of the vitreoretinal interface, establishing SS-OCT as a reliable tool for evaluating conditions such as partial or complete PVD. 

There was one false-negative case where obvious vitreous attachment was seen on OCT B-scans but was not detected during vitrectomy. There could be two reasons for this. First, there was a 6-day gap between the OCT scanning and the surgery, during which the PVD could have occurred. Second, the vitreous may become completely detached during the very beginning of core vitrectomy. 

Four false-positive cases showed an absence of vitreous cortex reflectivity on all 33 B-scans, but vitreous attachment was detected during surgery. There are two reasons for this. One is that it is a great challenge for OCT to differentiate completely attached vitreous from complete PVD, and the other is that, with a spacing of 0.25 mm between each scan line, any attached vitreous in the gaps would not be detected on OCT B-scans. Therefore, further research could focus on increasing the scanning density or total number of scans, as well as shortening the interval between OCT scanning and surgery. 

This study does have some limitations. Despite the use of TA to aid in determining the vitreous status, there is a possibility that a small area of mildly adherent vitreous might detach during the vitrectomy procedure. In this study, foveal adhesion was indeed clearly observed in seven cases by OCT preoperatively but no vitreous attachment to the macula was detected after TA staining during vitrectomy. 

In conclusion, our study demonstrates that SS-OCT, characterized by a deeper depth of field and a wider capture window, is proving to be a reliable tool for detecting the presence or absence of complete PVD. Moreover, given its relative objectivity and user-friendly operation, SS-OCT could also serve as an efficient tool in evaluating PVD status preoperatively and exploring the mechanisms of vitreoretinal diseases. 

Acknowledgments

Supported by grants from the National Natural Science Foundation of China (82070972 and 82271093). 

Disclosure: Z. Lin, None; K. Gao, None; R. Tuxun, None; C.-L. Tsai, None; Z. Xu, None; L. Jiang, None; Y. Liu, None; Z. Chen, None; Z. Chen, None; B. Liu, None; Y. Ma, None; X. Wei, None; K. Lai, None; T. Li, None