**Purpose**:
We advance studies of subretinal treatments by developing a microscope-integrated optical coherence tomography (MIOCT) image-based method for measuring the volume of therapeutics delivered into the subretinal space.

**Methods**:
A MIOCT image-based volume measurement method was developed and assessed for accuracy and reproducibility by imaging an object of known size in model eyes. This method then was applied to subretinal blebs created by injection of diluted triamcinolone. Bleb volumes obtained from MIOCT were compared to the intended injection volume and the surgeon's estimation of leakage.

**Results**:
Validation of the image-based volume measurement method showed accuracy to ±1.0 μL (6.0% of measured volume) with no statistically significant variation under different imaging settings. When this method was applied to subretinal blebs, four of 11 blebs without surgeon-observed leakage yielded a mean volume of 32 ± 12.5 μL, in contrast to the intended 50 μL volume injected from the delivery device. This constituted a mean difference of −18 μL (mean percent error, 36 ± 25%). For all 11 blebs, the surgeon's estimations of leakage were significantly different from and showed no correlation with the volume loss based on image-based volume measurements (*P* < 0.001, paired *t*-test; intraclass correlation = 0).

**Conclusions**:
We validated an accurate and reproducible method for measuring subretinal volumes using MIOCT. Use of this method revealed that the intended volume might not be delivered into the subretinal space. MIOCT can allow for accurate assessment of subretinal dose delivered, which may have therapeutic implications in evaluating the efficacy and toxicity of subretinal therapies.

**Translational Relevance**:
Use of MIOCT can provide feedback on the accuracy of subretinal injection volumes delivered.

^{1–4}More recently, clinical trials involving subretinal injections of viral vectors as gene therapy for retinitis pigmentosa,

^{5}Leber's congenital amaurosis.

^{6,7}and Leber's hereditary optic neuropathy

^{8}have been conducted. Subretinal injections of recombinant tPA also have been used to manage submacular or subretinal hemorrhage.

^{9,10}However, subretinal injections carry risk of complications, including retinal detachment, choroidal hemorrhages, rupture of Bruch's membrane, and reflux of RPE cells or other therapeutics from the needle.

^{11}This may lead to insufficient delivery of stem cells into the subretinal space or leakage of viral vectors into the vitreous body, and increase the risk of postoperative vitritis.

^{2,5–8,12,13}With the rapidly increasing interest in subretinal therapies and progress in clinical trials, there is a need to test this assumption and directly and accurately measure the volumes of therapeutics.

^{14}and shown in previous studies using spectral domain OCT (SD-OCT) to image areas of subretinal injections immediately before and after injection, and three-dimensional (3D) volumes have been postprocessed and reconstructed from the collected data.

^{15–17}More recently, live spectral domain microscope-integrated OCT (SD-MIOCT) showing two-dimensional (2D) B-scans have been used intraoperatively to aid in delivery of viral vectors and RPE cells into the subretinal space.

^{18,19}Gregori et al.

^{20}demonstrated the benefits of SD-MIOCT in subretinal gene therapy delivery; the ultimate goal, he stated, would be to calculate the volume of vector injected. Current generation commercial retinal SD-MIOCT systems, however, have limitations, such as in speed and depth of imaging, that make intraoperative imaging and quantitative assessment of subretinal fluid volumes challenging.

^{15}

^{21,22}In previous swept-source MIOCT studies, live volumetric images of subretinal blebs were captured (Vajzovic LM, et al.

*IOVS*. 2017;58:ARVO E-Abstract 3122), thus prompting the question of whether we also could use these 3D images to quantify the volume of subretinal blebs. Accurate and reproducible quantification of subretinal volumes using SD-OCT images has been shown previously for measurement of drusen volumes in AMD,

^{23}and, thus, we hypothesize that swept-source MIOCT also can be used to image and measure volumes of subretinal blebs accurately and reproducibly within 10% of the actual volume. Additionally, we hypothesized that swept-source MIOCT image-based volume measurements will be comparable to the current standard, where the volume of the subretinal bleb is equal to the volume injected from a calibrated delivery device under the assumption of that no leakage is observed by the surgeon.

^{21}The system uses a 100 kHz Axsun swept-source laser centered at 1060 nm, has a peak sensitivity of 99 dB, allows for up to 7.4 mm depth imaging range, and has an axial resolution in air of 7.8 μm. An electrically-tunable lens allowed for dynamic focus adjustment ensuring that the OCT system was parfocal with the surgical microscope.

^{24}We acquired volumetric images with 1000 A-scans/B-scan and 128 B-scans/volume through a contact lens. Lateral dimensions of volumes consisted nominally of either 6 × 6 or 10 × 10 mm. Additionally, volumes were captured at 0° and 90° scan angles to establish intrasession reproducibility.

^{25}To assess the accuracy and repeatability of manual segmentation of B-scans, a grader repeatedly traced the top curve of the ceramic ball in the same set of B-scans from a single volumetric image. However, unlike the B-scans of a subretinal bleb in which the upper and lower boundaries were clear and easy to segment, the lower half of the ceramic ball was not visible due to shadowing; therefore, to avoid introducing irrelevant error to our method for measuring subretinal bleb volumes, we set the same line in all B-scans of the ball to serve as the lower boundary of the segmentations, producing segmented areas of semicircles throughout the B-scans (Figs. 1D, 1E). The areas of the semicircles then were doubled to convert the areas to those of full circles. The variance in manual segmentation was factored into the propagation of uncertainty

^{26}calculations to obtain an overall measure of uncertainty associated with the MIOCT image-based method for measuring volumes.

*X*,

*Y*, and

*Z*dimensions of a voxel (a cuboid on a three-dimensional grid):

*X*and

*Y*lateral dimensions, we measured the diameter of the ball on the en face MIOCT images in pixels, calibrated that to the known diameter of the ball in millimeters, and calculated the conversion from pixels to millimeters (Fig. 1C). To assess the variation in the

*X*and

*Y*dimensions of MIOCT images due to differences between porcine eyes, we then repeated these measures under the same imaging system settings in different eyes. The variance in the calibrated

*X*and

*Y*lateral dimensions were factored into the overall measure of uncertainty associated with the MIOCT image-based method for measuring volumes. The

*Z*pitch was calculated from a formula using the number of spectral samples and index of refraction of balanced salt solution (BSS). The accuracy in the Z pitch is an inherent property of the swept source laser clock and the digitizer, and is negligible compared to segmentation error.

^{27}

^{24}(within a visually optimal range ± 1.8 mm from the ideal focus while maintaining parfocality with the surgical microscope). Diameters in the

*X*and

*Y*lateral dimensions were measured 10 times for each en face image captured using each of the different settings, and then averaged to calculate the mean volume for each image. This method enabled us to exclude any error that may occur from manual segmentation of B-scans, and, therefore, attribute any error from the calculated volumes to the change in imaging system settings.

^{28}To perform subretinal injections, we used a 25 gauge/38 gauge MedOne PolyTip cannula (MedOne Surgical, Sarasota, FL) attached to a 1.0 mL syringe with flexible extension tubing. During the subretinal injection, we documented the surgeon's estimation of the percentage of leakage of triamcinolone from the retinotomy site or the cannula tip as well as any factors that may have caused any irregularities in the formation of the subretinal bleb. Each porcine eye was used only for one subretinal injection.

^{25}This involved tracing the upper boundary (lower margin of the neurosensory retina) and the lower boundary (upper border of the RPE; Fig. 2), from which we were able to obtain the total number of pixels occupied by the subretinal bleb. We multiplied the total segmented area by the calibrated voxel dimensions to calculate the volume of the subretinal bleb in microliters. We analyzed the subretinal bleb volumes in which the surgeon observed bleb formation with 50% or less leakage of triamcinolone, and excluded one volume with inadequate visualization of bleb margins.

^{26}to calculate the overall uncertainty from the multiple steps in our method. Propagation of uncertainty combined the variances in measurement associated with the

*X*and

*Y*voxel dimensions and manual segmentation.

*t*-tests. Image-based volumes acquired using different focal length settings for the electrically-tunable lens were compared using a multivariate analysis of pair-wise differences among three focal lengths simultaneously based on Wilk's λ. Intergrader reproducibility of manual segmentation of subretinal blebs was assessed using a Wilcoxon signed rank test.

*t*-test to assess for significant difference and intraclass correlation to assess for agreement.

*n*= 5) of the ceramic ball of 5,968,131 voxels in the same volumetric image was 23,259 voxels, the standard deviation in the

*X*voxel dimension was 0.009562 mm/pixel (

*n*= 14 eyes), and the standard deviation in the

*Y*voxel dimension was 0.06984 mm/pixel (

*n*= 14 eyes). Using propagation of uncertainty calculations to combine the individual variances, we reported that our MIOCT image-based volume measurements are accurate to ±1.0 μL (6.0% of measured volume).

*n*= 5 eyes,

*P*= 0.73, paired

*t*-test; Table 1) and intraclass correlation was 0.92 (95% confidence interval [CI], 0.53–0.99). There also was no significant difference when imaging with two different MIOCT scan angles (

*n*= 5 eyes,

*P*= 0.41, paired

*t*-test; Table 1), and intraclass correlation was 0.77 (95% CI, 0.02–0.97). Varying the focal length of the electrically-tunable lens ±1.8 mm from ideal focus produced no significant differences among the three groups (

*n*= 7 eyes,

*P*= 0.26, multivariate analysis of pair-wise differences simultaneously based on Wilk's lambda; Table 1), and overall intraclass correlation was 0.74 (95% CI, 0.46–0.96). The image grader noted that some of the +1.8 mm images were not as sharply defined and, therefore, more difficult to measure.

*P*< 0.001,

*n*= 11, paired

*t*-test; intraclass correlation, 0; 95% CI, 0–0.37; Fig. 5). Observations made by the surgeon during the subretinal injections included that leakages of triamcinolone occurred from the subretinal cannula before injection, from the retinotomy site during and after injection, and from the subretinal cannula after withdrawal from the retinotomy, and all were included in the surgeon's total estimation of leakage.

*n*= 4, mean measured volumes range, 14.2–42.4 μL). Using the methods described above and changing the lateral scan dimension and scan angle, the standard deviation in measurements ranged from 0.5 to 2 μL (Table 3). We also evaluated intergrader reproducibility of manual segmentation using a random subset of subretinal blebs, and found no significant difference between bleb volumes segmented by two independent graders (

*n*= 5, mean difference 1.0 ± 0.43 μL,

*P*= 0.06, Wilcoxon signed rank test).

*n*= 20, percent error, −2%; coefficient of variation, 10%).

^{15,17,19}These studies, which used SD-OCT, discussed limitations that hindered the successful imaging of the area of interest, including significant time required for close interaction between the operator and surgeon to localize the instrument to align and capture the SD-OCT images, and the limited depth of capture on B-scans.

^{15}Accurate quantification of subretinal volumes requires imaging of all boundaries of the whole subretinal bleb. The use of swept-source MIOCT

^{21,22}addresses some of the limitations described previously in SD-OCT systems by enabling fast and dynamic localization of instruments and subretinal bleb boundaries as the bleb evolves, and also images depths up to 7.4 mm, more than the 2.5 mm depth offered by current generation posterior segment SD-OCT systems. One of the blebs in this report was 2.7 mm tall and would not have been imaged completely using an SD-OCT system. Additionally, increased imaging depth is useful particularly when blebs unexpectedly form more peripherally where the retina is curving upwards, increasing the axial height of the bleb. While SD-OCT can be used to visualize and possibly quantify subretinal bleb volumes, swept-source OCT does offer advantages that may benefit in future translation to in vivo imaging.

^{29}

**S.T. Hsu**, None;

**H. Gabr**, None;

**K. Sleiman**, None;

**C. Viehland**, None;

**H.T. Ngo**, None;

**O.M. Carrasco-Zevallos**, None;

**S.S. Stinnett**, None;

**L. Vajzovic**, Knights Templar Eye Foundation (R), PDC's ENABLE Award (R), Alcon (R), Janssen Pharmaceutical (R), Roche (R), Second Sight (R), DORC (R);

**R.P. McNabb**, Leica (P);

**J.A. IIzatt**, Leica (P, R);

**A.N. Kuo**, Leica (P), ClarVista (C);

**C.A. Toth**, Alcon (P)

*. 2016; 31: 25–29.*

*Semin Ophthalmol**. 2016; 57: ORSFc1-9.*

*Invest Ophthalmol Vis Sci**. 2013; 140: 2576–2585.*

*Development**. 2013; 54: ORSF68–ORSF80.*

*Invest Ophthalmol Vis Sci**. 2016; 135: 327–343.*

*Hum Genet**. 2009; 374: 1597–1605.*

*Lancet**. 2008; 358: 2231–2239.*

*N Engl J Med**. 2016; 123: 558–570.*

*Ophthalmology**. 2010; 248: 5–11.*

*Graefes Arch Clin Exp Ophthalmol**. 2001; 131: 208–215.*

*Am J Ophthalmol**. 2015: 52247.*

*J Vis Exp**. 2006; 13: 1074–1084.*

*Mol Ther**. 2012; 130: 65–75.*

*Arch Ophthalmol**. 1978; 207: 7–14.*

*Albrecht Von Graefes Arch Klin Exp Ophthalmol**. 2011; 52: 3153–3159.*

*Invest Ophthalmol Vis Sci**. 2014; 9: e105224.*

*PLoS One**. 2015; 46: 327–332.*

*Ophthal Surg Laserrs Imag Retina**. 2012; 29: 83–93.*

*Vis Neurosci**. 2017; 376: 1038–1046.*

*N Engl J Med**Retina*.

*. 2016; 6: 31689.*

*Sci Rep**. 2016; 57: OCT37–OCT50.*

*Invest Ophthalmol Vis Sci**. 2012; 53: 53–61.*

*Invest Ophthalmol Vis Sci**. 2014; 5: 1877–1885.*

*Biomed Opt Exp**. 2013; 54: 7595–7602.*

*Invest Ophthalmol Vis Sci**. 1966; 70C: 263–273.*

*J Res Natl Bureau Standards C Eng Instrum**. 2015: 65–94.*

*Opt Coher Tomogr**. 2013; 97: 1384–1386.*

*Br J Ophthalmol**.*

*Ophthalmology*