1. Introduction

Optical coherence tomography (OCT) allows for micrometer-scale imaging of the eye and has become an invaluable tool in ophthalmology [1–3]. OCT systems have since been integrated into surgical microscopes, allowing for real-time intrasurgical imaging of ocular tissues [4–6]. The technology has also been commercialized with systems such as the Leica Enfocus and Zeiss Rescan. While intraoperative OCT has significantly improved visualization of tool-tissue interactions inside the eye [7,8], new advancements in ophthalmic surgery have highlighted several key limitations of the existing technology. One of these advancements, subretinal delivery of therapeutics is becoming an increasingly prevalent form of treatment for a variety of previously untreatable genetic or degenerative retinal diseases. Debilitating diseases such as retinitis pigmentosa, Usher syndrome, Leber’s congenital amaurosis, choroidermia, and age related macular degeneration (which affect an estimated 950,000 people in the United States) can lead to severe visual impairment and blindness [9]. New therapeutics, including viral vectors, stem cells, and gene therapies, have shown great promise and are now in clinical trials or approved for clinical use for these diseases [10–19]. The approaches to deliver these therapies beneath the retina require: first, control of surgical maneuvers to position the delivery tip at a specific intraocular location without damaging delicate adjacent structures tens of microns away; and second, delivery of the required therapeutic dose at that location. Current delivery techniques present several risks and challenges which may limit their effectiveness. Positioning of the delivery tip relies on the clinician’s estimate from 3D microscope visualization. Incorrect assessment of delivery location is a recognized cause of suboptimal outcomes or complications such as penetration of Bruch’s membrane, retinal detachment, or injection into the vitreous cavity. With research swept-source, intraoperative OCT visualization is shown to aid placement of subretinal tools [20]. With current commercial intraoperative SDOCT systems, although it is difficult to align and track subretinal tool positioning in real time, the imaging provides additional valuable visualizations of the surgical instrument and how it interacts with surrounding tissue and the extent of the focal retinal detachment (bleb) [21–23]. Currently, surgeons also rely on a visual estimation of volume on the microscope [23] or OCT [21] to assess delivery success. Estimations of volume delivered can be highly variable, and material can reflux out of the subretinal space resulting in ocular inflammation or suboptimal therapeutic results, thus it has been suggested that new methods are needed to assess subretinal volumes delivered [21]. Two studies have measured the volume of subretinal therapeutics delivered; these were limited by the small fields-of-view or imaging depth of current commercial intrasurgical OCT systems, which limited measurements to volumes of less than 44.1 µL [24,25]. Effective therapeutic doses vary depending on the treatment, but some clinical trials or the approved gene therapy product Luxturna require up to 300 µL. Previously demonstrated techniques are well under such proposed injection volumes. Hsu et. al. used a flat contact lens to perform imaging, which kept the OCT scanning telecentric and distortion free, but limited the field-of-view (FOV). Roodaki et. al. used a spectral domain OCT system which severely limited the imaging depth. Furthermore, there is no validation of the volumes. In this paper, we demonstrated volume measurements up to 138.30 µL, but our described technique is suitable to quantify much larger volumes due to the large field-of-view (FOV) and imaging depth.

Retinal imaging instruments can be broadly divided into two categories: direct visualization where no intermediate or “aerial” image is formed (e.g. pocket ophthalmoscopes), and indirect visualization where an aerial image is formed (i.e., indirect ophthalmoscopy). Wide-field vitreoretinal surgery is typically conducted using indirect visualization performed using a stereo zoom surgical microscope suspended above the patient, with an additional removable optic placed in close proximity to the eye. The additional optic can be of a contact or non-contact design. Non-contact indirect retinal viewing designs (e.g. Oculus BIOM, Zeiss Resight, Leica RUV800, and Topcon OFFISS) provide good image quality and working space for the surgeon around the globe while maximizing the retinal field of view, and are the standard of care at many institutions including ours. Acting in concert with the patient’s own eye, the additional optic forms an inverted aerial image which is directly imaged by surgical microscope, as well as any OCT system built into or attached to the microscope. Since the additional optics refract the scanning OCT light to pivot about a point within the eye, such indirect viewing systems introduce lateral distortions into OCT visualizations, making the rectangular shape of the OCT display in typical clinical and research systems inaccurate. These distortions make it particularly challenging to obtain calibrated quantitative measurements. While there have been some publications examining distortion correction in ophthalmic OCT, their applications have been limited to curvature quantification of ocular structures [26,27]. Here, we present a novel approach to perform quantitative measurements of intraocular structures using intraoperative OCT with a non-contact indirect retinal viewing system (which is suspended above the eye), including specific, optimized procedures for producing an optical model for the system composing the combined optics of the eye and imaging assembly. Once our model was calibrated, we demonstrated measurements of surgical instrument tools and subretinal microinjection blebs. Finally, as a proof of concept, we measured surgical instrument tools in live human surgery. While quantitative measurement of the volume of subretinal blebs present a particularly compelling need for OCT based retinal quantitation, there are many other possible applications of OCT based retinal quantitation methods for which these methods may be applied. Other possible applications include precise measurements of macular holes to create retinal patch grafts or autologous amniotic membrane patches of matching size, which can then be used to repair the macular hole. There are also additional patch therapies in development that allow retinal detachment repair, and intraoperative sizing of retinal tears which lead to detachment will be important in the future.

2. Methods

In the following sections, we first describe the design of the intraoperative OCT system. We then discuss the optical modeling of the system and proposed optimal measurement procedure for an indirect retinal viewing system leveraging prior information on the optics of both the system and the human eye. The technique is first validated by measuring the volumes of spheres of known sizes in a custom made eye phantom, then extended to ex vivo porcine eyes. This is further extended to measure surgical tool diameters and subretinal injection volumes in ex vivo porcine eyes. Finally, as a proof-of-concept, we measured the diameter of surgical tools in human eyes undergoing standard surgical procedures.

2.1 Intra-operative OCT system design

The following novel measurement techniques were designed for use with an intraoperative OCT systems employing indirect retinal viewing optics (Fig. 1). For this paper, we used a custom intraoperative OCT system similar to the one previously reported by Carrasco-Zevallos, with additional modifications described as follows [8]. Our custom intraoperative OCT system consisted of a standard transmissive Mach-Zehnder interferometer based engine with a 400 kHz external cavity tunable “ping-pong” laser (consisting of two lasers with forward sweeps interlaced) (Axsun Technologies Inc.) centered at 1050 nm (Fig. 1(a)). Light was detected using a balanced photodetector (PDB481C-AC, Thorlabs, Inc.) and digitized using a high-speed digitizer (ATS9373, Alazar Technologies Inc.). The sample arm was based, in part, on a Leica Microsystems Enfocus OCT scanner, which we modified with optics optimized for 1050 nm operation (with assistance from Leica) and with high-speed galvanometer scanners (ScannerMAX Saturn 5B, Pangonlin Laser Systems Inc.) with 6 mm mirrors to support real-time volumetric (“4D”) OCT imaging (Fig. 1(b)). Raw OCT data was collected, processed, and rendered for display using a custom-built CUDA/C++ based application (BROCT, Duke University). OCT volumes, B-scans, and maximum intensity projections (MIPs) were displayed on an Ngenuity 3D visualization system (Alcon AG) alongside the stereoscopic surgical microscope video (Fig. 1(c)). For our non-contact viewing system, an Oculus BIOM 5c (Oculus Surgical, Inc., Fig. 1(d)) was modified with the addition of an axial translation measurement ruler to precisely measure the distance between correction lens (top) and wide field lens (bottom), the latter of which can be raised and lowered using the focusing knob provided. The correction lens and wide field lens were from the Oculus HD Professional 54411 BIOM Optic Set (Oculus Surgical, Inc.). A spacer lens (Ocular Flat Disposable Vitrectomy Lens 55110, Oculus Surgical, Inc.) was also used. Precise measurement of the distances in Fig. 3 is the subject of the detailed procedure we described in Section 2.3. Thus, they varied depending on differing experimental conditions (such as patient eye length), however in our experiments using porcine and human eyes they ranged from ${r_1}$: 40.1-40.1 mm, ${r_2}$: 94.7-102.4 mm, ${d_1}$: 11.96-12.36 mm, ${d_2}$: 11.96-12.35 mm, ${u_{eye}}$: 20.72-24.6 mm.

 

Fig. 1. (a) Duke custom intraoperative OCT system schematic. (b) OCT scanner mounted on a Leica Proveo surgical microscope with the Ngenuity 3D visualization system camera. (c) Ngenuity 3D visualization system display showing OCT and surgical microscope views. (d) Non-contact (Oculus BIOM 5c) indirect retinal viewing system with correction lens, wide field lens, focusing knob, and measurement ruler.

Download Full Size | PDF

2.2 Optical modeling of intraoperative OCT systems and measurement procedures

Normally, intraoperative OCT systems designed for direct imaging of tissue surfaces (such as the anterior eye or open surgical fields) perform telecentric scanning, wherein the chief rays are parallel and roughly normal to the desired imaging plane (Fig. 2(a)). With indirect retinal viewing systems, the telecentric scanning OCT light rays are refracted by the non-contact viewing system optics such that the rays pivot about a point within the eye just above the crystalline lens (Fig. 2(b)). This produces chief rays scanning over a range of angles, and thus the resulting retinal scan FOV at the back of the eye is dependent on several variables such as the aerial scan FOV at the intermediate aerial image plane, the axial length of the eye, and the exact placement of the viewing system optical components. For example, in a longer eye, the scan length increases as the angularly scanned light travels farther to reach the retina. To make accurate quantitative measurements, the values for these variables need to be determined. The aerial scan FOV at the telecentric image plane can be measured by simply using the OCT system to image a distance metric placed there (such as a microscopy ruler or other flat surface with known feature spacing). However, much more complex measurement techniques would be required to determine all of the parameters of an indirect retinal viewing system characterizing the positions and properties of all of the optics between the telecentric scan plane and the (curved) retinal surface. Once determined, these parameters could ideally be used to create an entire optical model of the OCT system, indirect retinal viewing optics, and the eye. However, such complete characterization of all optical elements is clearly incompatible with practical pre-operative measurement in a real operating room environment. In this work, we describe a compromise approach whereby we combine prior knowledge of the optics of the OCT system and indirect retinal viewing optics with a state of the art human eye model (Polans eye model [28]), and propose a minimal set of measurements to be taken in the operating room to optimize only the most important parameters and element spacing to provide a practically achievable overall model of the optical system. With such a model in hand, we simulate scanning an entire OCT volume by virtually sweeping the galvanometers to determine the position of each individual A-scan and the scan length in the back of the eye, which is then used to calculate the lateral voxel pitch. In contrast, the axial voxel pitch along the direction of the chief ray is derived from the known spectral sampling bandwidth of the OCT source and assumed eye refractive index as in standard retinal OCT. All optical modeling in the following sections was done using OpticStudio (Zemax LLC).

 

Fig. 2. Simplified optical model of an intraoperative OCT system performing (a) telecentric scanning of the anterior eye or surgical field wherein chief rays are parallel and (b) retinal scanning using an indirect retinal viewing system wherein rays are refracted to pivot about a point above the crystalline lens (indicated by yellow star). The axial length of the eye is the length from the corneal apex to the apex of the retina and (indicated by the bracket labelled eye). The retinal scan field-of-view (FOV) is indicated by the extent of the rays.

Download Full Size | PDF

2.3 Procedure for non-contact indirect viewing system

We developed an optical model and measurement procedure for a non-contact viewing system consisting of two lenses, a correction lens and a wide field lens (Fig. 3), which are suspended over the patient’s eye. We tested this procedure using an Oculus BIOM 5c indirect viewing system, which we modified by addition of a ruler to allow for direct measurement of the central distance (${d_1}$) between the top of the correction lens and the top of the wide field lens as the surgeon focuses on the retina. The following procedure assumes that ${k_1}$, the distance from the top of the correction lens surface to the intermediate aerial image plane, ${k_{lens}}$, the thickness of the wide field lens, and ${k_{spacer}}$, the thickness of the spacer lens (described later) are known from prior careful measurement or from the lens manufacturer’s specifications. However, there are still several unknown distances in the sample arm: the axial length of the eye (${u_{eye}}$) and the spacing between the bottom of the wide field lens and the top of the patient’s eye. By design, the OCT system and surgical microscope are parfocal and thus share the same focal plane. The procedure starts by placing a flat object, such as an IR laser viewing card (Thorlabs, Inc.) or business card at the aerial image plane so that it is in focus as viewed through the microscope. The OCT system reference arm is then adjusted to align the OCT image of the card at the bottom of the B-scan image, and the corresponding reference arm position ${r_1}$ is recorded. While this alignment is done manually, we believe its effect on measurement error is very small, since the manual alignment only serves as an initial reference point to match the bottom of the B-scan and reference arm position. Then, the focusing knob of the indirect viewing system (Fig. 1(d)) is adjusted by the surgeon to bring the retina into focus through the microscope, and ${d_1}$ is recorded. Next, the wide field lens is temporarily raised (using the focusing knob) while a spacer lens of known thickness is placed on the eye. The wide field lens is then manually lowered to touch the spacer lens, and distance ${d_2}$ is recorded (Fig. 3(a)). Then the spacer lens is removed, and the wide field lens is restored to its previous position at ${d_1}$ to allow the retina to be imaged (Fig. 3(b)). The reference arm position required to bring the OCT image of the retina to bottom of the B-scan image is adjusted, and is recorded as ${r_2}$. Since the length of the reference arm and sample arm need to match in order to observe the interferometric OCT signal, this requirement in combination with the measurements just described can be used to estimate the axial length of the eye. The difference ${r_2} - {r_1}$ of the two different reference arm positions represents how much the reference arm was moved from imaging the aerial image plane (at a known physical axial location in the OCT sample arm) to imaging the retina (at some unknown physical axial location). The difference ${d_2} - {d_1}$, plus the central thickness of the spacer lens, is thus the distance between the bottom of the wide field lens and the corneal apex when the retina is in focus. The axial length of the entire sample arm is now known up to the corneal apex. The remaining sample arm distance comes from the eye itself, and thus using the path length matching requirement:

 

Fig. 3. Illustration of measurement procedure for non-contact indirect viewing system (which consists of a correction lens and a wide field lens). Simplified optical system diagram (not to scale) of the non-contact viewing system with a spacer lens (in cyan) inserted (a) and after the spacer lens is removed and surgeon focuses on the retina (b). Depicted variables are: ${d_1}$ (distance between top of correction lens and top of wide field lens without the spacer lens in place), ${d_2}$ (distance between top of correction lens and top of widefield lens with the spacer lens in place), ${u_{eye}}$ (axial eye length), ${r_1}$ (reference arm length to intermediate aerial image plane), ${r_2}$ (reference arm length to apex of retina), ${k_1}$ (distance from top of correction lens to intermediate aerial image plane), ${k_{lens}}$ (thickness of wide field lens), and ${k_{spacer}}$ (thickness of spacer lens). All distances are center distances and are depicted as displaced only for readability of the figure. We have not provided specific measurements since the lens specifications are proprietary and the other measurements varied depending on the configuration used for a particular measurement.

Download Full Size | PDF

2.4 Model validation in eye phantoms

We next describe a procedure in which objects of known volume were placed in a fully characterized eye phantom, imaged, and measured to validate the accuracy of the model. We designed and fabricated a custom eye phantom using grey resin on a Form 2 (Formlabs) 3D printer (Fig. 4(a)). The phantom comprised an achromatic doublet lens (#65-436, Edmund Optics Inc.) placed 13.99 mm above a simulated retina which consisted of a grid of 3D printed features (cylinders with 0.2 mm radius and height, with 0.5 mm center-to-center spacing).

 

Fig. 4. (a) Exploded view of 3D printed eye phantom. Representative (b) OCT MIP view of 0.8 mm alumina ceramic sphere (indicated by red arrow) with a non-contact viewing system. The green line indicates the location of the B-scan scan shown as an inset of (c), the C-scan at the equator of the sphere (indicated by cyan line). The equator is set a calculated distance (indicated by the magenta arrow) from the top of the sphere. A representative segmentation of the sphere is shown as a yellow outline the C-scan.

Download Full Size | PDF

Alumina ceramic spheres (Boca Bearings, Inc.) of varying diameters (d = 0.8, 1.5, 3.0 mm) were placed at the back of the eye phantom. The non-contact indirect viewing system (Oculus BIOM 5c) was attached to the surgical microscope and adjusted to image the back surface of the phantom. The distance between the correction lens and wide field lens was noted using the attached measurement ruler. Then the spacer lens was placed on the eye phantom, and the new distance was noted. The spacer lens was removed, and the wide field lens was restored to the original position, as previously described in section 2.3, and OCT images (Fig. 4(b)) were acquired at 750 A-scans/B-scan x 750 B-scans/volume. The average lateral FOV (taking into consideration the varying size of voxels) was 16.17 × 16.08 mm. This process was repeated 10 times for each sphere diameter. The lateral voxel dimensions were determined as described in previous sections, except the Polans eye model was replaced with the eye phantom model. For the eye phantom, the axial length was the distance from the apex of the lens to the apex of the simulated retina. To adjust the eye phantom model for the measured axial eye length, we modified the distance between the posterior lens surface and simulated retina. For the validation of the sphere measurements, a research technician manually segmented the spheres using ImageJ (NIH) in each C-scan of the OCT volume. Since the bottom half of each sphere was shadowed in the OCT images, the lower segmentation boundary was set at the equator of the sphere. The equator was set a fixed number of pixels down from the top of the sphere by using the Z-voxel pitch and known diameter of the sphere (Fig. 4(c)). Using the X- and Y- voxel pitch determined from the optical model and the Z-voxel pitch, the volume of every voxel within the segmented region was calculated using MATLAB (MathWorks). These were summed up to obtain the volume of the hemisphere and doubled to get the full spherical volume.

2.5 Model validation in porcine eyes

Quantitative volume measurements were next performed using alumina ceramic spheres in porcine eyes. Freshly enucleated porcine eyes were purchased from an abattoir and imaging studies were performed within 12 hours. After a 23 gauge pars plana vitrectomy, a 2.5 mm keratome blade was used to make an incision parallel to and 4 mm behind the limbus. Three alumina ceramic spheres (diameter = 0.8, 1.0 and 1.5 mm) were inserted into the eye via the incision and placed at the back of the eye (Fig. 5). These spheres were smaller in diameter than the ones used in the eye phantoms due to limited incision length in order to prevent excessive fluid and intraocular pressure loss.

 

Fig. 5. Representative (a) volume and (b) MIP view of the 0.8, 1.0, and 1.5 mm alumina ceramic spheres from OCT and (c) surgical microscope in an ex vivo porcine eye.

Download Full Size | PDF

The same measurement procedure described in section 2.4 using the Oculus BIOM 5c indirect viewing system was repeated in the ex vivo porcine eyes. The measurements were used to generate an optical model to estimate the volumes of the spheres. OCT images (Fig. 5(a), b) were acquired at 750 A-scans/B-scan x 750 B-scans/volume. This procedure was repeated in 10 different porcine eyes for 10 measurements of each sphere diameter. The lateral voxel dimensions were determined as described in section 2.4 using the Polans eye model as an approximation for a porcine eye, and the distance between the posterior lens and retina was adjusted for the measured axial eye length of the specific porcine eye being imaged.

2.6 Measurement of surgical tool diameters in porcine eyes

For closer to real-world validation of the measurement techniques and optical models we developed, the diameters of common surgical tools in porcine eyes were calculated and compared to their specified gauge values. To accomplish this, we performed a 23 gauge pars plana vitrectomy in the porcine eyes, and 23 gauge and 25 gauge soft tip cannulas (Alcon Inc.) were inserted via the trocars. These tools were then held adjacent to the retina similar to what would be performed in human surgery. This procedure was repeated in 6 eyes. The measurement procedure described in section 2.4 was performed, and the tools were imaged using OCT (Fig. 6(a), b). To estimate the diameter of the instrument shafts, the C-scan at the tip of instrument was selected, and the shadow cast by the instrument was manually segmented using ImageJ. A line was drawn along the edge of the instrument shaft, and a perpendicular line spanning the width of the instrument was used as the final segmentation (Fig. 6(c)). The X- and Y- voxel pitch at the particular depth was then used to estimate the length of the segmented line representing the diameter of the instrument.

 

Fig. 6. Representative OCT view of (a) 23 gauge and (b) 25 gauge soft tip cannulas in an ex vivo porcine eye. (c) Representative C-scan (data slice perpendicular to A-scan direction) showing the segmented tool tip indicated with the cyan arrow.

Download Full Size | PDF

2.7 Measurement of microinjection volumes in porcine eyes

As a proof-of-concept of quantitation of retinal micro-injection bleb volumes, mock subretinal injections were performed in ex vivo porcine eyes. A 23 gauge pars plana vitrectomy was performed. An extendible 41 gauge subretinal injection needle in a 23 gauge cannula (DORC B.V.) was connected to an injection syringe (1000 µL) filled with balanced salt solution (BSS). The tip of the cannula was immersed in a dish of BSS to flush out potential bubbles and then introduced into the eye via a port. The needle was extended and guided into subretinal space in perifoveal location. Once the needle was in place, the measurement procedure described in section 2.4 was repeated. After the measurements were taken, the surgeon would use the markings on the injection syringe to inject a therapeutic dose (100 µL) of BSS into subretinal space and an OCT image of the bleb (Fig. 7) was captured.

 

Fig. 7. (a) B-scan, (b) volume render, and (c) MIP showing the bleb. Green line in MIP indicates position of the B-scan.

Download Full Size | PDF

The subretinal bleb was manually segmented in all the B-scans by a research technician using Avizo (Thermo Fisher Scientific, Inc.). The entire bleb volume was then determined using the determined voxel pitches. This experiment was repeated in 3 different porcine eyes.

2.8 Measurement of surgical tool diameters in in vivo human eyes

Finally, as a demonstration of potential applications in a real-world human surgical setting, the described techniques were applied to measurements of surgical tools in human retinal surgeries. All human research was performed under a study protocol (NCT03713268) approved by the Duke Medical Center Institutional Review Board and abided by the tenets of the Declaration of Helsinki. Patient consent to research participation was obtained prior to any research activities. Eight adult patients in the Intraoperative OCT Guidance of Intraocular Surgery II Study, who underwent standard clinical care vitrectomy procedures with the investigational microscope-integrated OCT imaging, had OCT imaging of a 25 gauge cannula (Alcon Inc.) when positioned immediately above the retinal surface. The measurement procedure was performed once on each patient and the obtained OCT images were processed as described in section 2.6 to determine the diameter of the instrument. The calculated diameter was compared to the specified diameter.

3. Results

In the following sections, we discuss the results of the experiments described previously. We first look at the volume of various sized spheres in eye phantoms, then in ex vivo porcine eyes. Then we examine the diameters of surgical tools and subretinal injections in ex vivo porcine eyes. Finally, we demonstrate measurements of surgical tool diameters in human eyes during surgery. With the described scanning protocol, the lateral voxel pitch ranged from 15.73 to 28.86 µm.

3.1 Model validation in eye phantoms

Using the image processing techniques previously described, the volumes of the alumina ceramic spheres in the model eye phantom were calculated. The experiment described in section 2.4 was performed to validate the accuracy of the optical model. These results are summarized in Table 1. Specified volume is the theoretical volume of the sphere as calculated using the nominal sphere diameter and tolerance (0.254 µm) specified by the manufacturer, and the calculated volume is the volume as measured using the intraoperative OCT measurement procedure (mean ± standard deviation). The following metrics are used to summarize the results. 1) The coefficient of variation: ratio of standard deviation to the mean. 2) The mean absolute error: mean of the absolute errors (absolute value of calculated volume - specified volume) for each observation. 3) The mean absolute percentage error (the mean difference between calculated volume and specified volume divided by specified volume).

Table 1. Validation volumes using non-contact indirect viewing system in eye phantom

View Table | View all tables in this article

3.2 Model validation in porcine eyes

This model was then validated using ex vivo porcine eyes as described in section 2.5. Alumina ceramic spheres were surgically inserted into the eye and placed on top of the retina. Once the spheres were in place, the measurement procedure was performed, OCT images were acquired, and the volumes of the spheres were calculated. These results are summarized in Table 2.

Table 2. Validation volumes using non-contact indirect viewing system in ex vivo porcine eyes

View Table | View all tables in this article

3.3 Measurement of surgical tool diameters in porcine eyes

Six porcine eyes underwent a pars plana vitrectomy and two different soft tip cannulas (23 gauge and 25 gauge) were inserted via the trocars and held adjacent to the retina similar to what would be performed in human surgery. The diameter of surgical tool dimensions was demonstrated following the standard of care for retinal surgeries. The dimensions of surgical tools in porcine eyes are summarized in Table 3.

Table 3. Calculated tool diameters using non-contact indirect viewing system in ex vivo porcine eyes

View Table | View all tables in this article

3.4 Measurement of microinjection volumes in porcine eyes

An additional measurement study of subretinal injection volumes was demonstrated using the non-contact approach following the standard of care for retinal surgeries. Subretinal injections of BSS were performed in three porcine eyes. The calculated microinjection volumes are summarized in Fig. 8 and compared to a targeted volume of 100 µL as observed by the surgeons using the markings on the injection syringe.

 

Fig. 8. A bar graph showing the calculated injection volume for each trial. The average volume (132.6 µL) is indicated by the solid orange line. The standard deviation is 7.18 µL. The target volume of 100 µL is represented by the dashed yellow line.

Download Full Size | PDF

3.5 Measurement of surgical tool diameters in in vivo human eyes

We also conducted initial studies of surgical tool measurements in eight human retinal surgeries, and these results are summarized in Table 4.

Table 4. Calculated tool diameters using non-contact indirect viewing system in human eyes

View Table | View all tables in this article

If we assumed an average distance between the posterior lens and the retina (16.04 mm in the unmodified Polans eye model), instead of the value retrieved using Eq. (1), we saw at worst a 16% difference in voxel size in the human data. We also modified the distance between the wide field lens and the cornea by ±2 mm and compared these voxel sizes to the voxel sizes computed using the measured distance. At worst, we saw a percent difference of 14% in voxel size.

4. Discussion and conclusion

In this paper, we have demonstrated a novel procedure to perform quantitative measurements of intraocular structures using intraoperative OCT with a non-contact indirect retinal viewing system. We devised novel measurement procedures to optimize a complete optical model of the OCT sample arm, the non-contact indirect retinal viewing system, and the patient’s eye. With a complete optical model, we simulated an OCT scan and derived voxel pitch throughout the entire OCT volume, which allowed us to perform quantitative measurements of imaged objects.

Validation was performed by estimating sizes of known objects in a controlled eye phantom and in ex vivo porcine eyes. The results of our validation studies showed concordance with specified dimensions within 7.32% for the eye phantoms. We aggregated different possible sources of error to determine a theoretical upper-bound on the accuracy of our measurement technique, so additional analysis was conducted using the initial pilot study with an eye phantom. A Monte-Carlo simulation was performed, which created a 5 pixel range of potential bias in each segmentation to account for segmentation error, which resulted in an 2.9% error. The Formlabs 3D printer used to create the eye phantoms had a layer resolution of 25 µm. Assuming this created an axial length difference of 25 µm from the expected axial length led to 5.4% error. The diameter tolerance of the ceramic spheres was ±0.254 µm which introduced 0.095% error. Additionally, errors in measuring the distances between the two lenses of the non-contact viewing system and the wide field lens of the viewing system and the top of the eye up to ±1 mm were modeled. This contributed an additional 5.6% of error. Adding everything in quadrature, the total error should be within 8.3% of the expected volume. Our results fall within this expected measurement error.

Subsequently, the technique was used to quantify dimensions of surgical tools intraoperatively and subretinal bleb injection volumes. The measurements of surgical tools fell within the expected error, while the calculated subretinal injection volume differed significantly from the expected volume. The subretinal injection volume difference could be attributed to other errors such as syringe fluid level readings which are prone to poor accuracy due to a variety of factors including the meniscus formed by the fluid, coarse markings on the syringe, and dim lighting conditions. This volume difference highlights the extreme variability currently present in subretinal injection procedures, which is likely related to syringe reading and volume delivery precision, and demonstrates the need for quantitative measurements. While we are currently performing manual segmentation post-surgery, clearly development of automated segmentation technology would be a key component to bring this technology closer to real-time use during procedures.

In summary, we demonstrated that our intraoperative OCT-based quantitative measurements technique only adds 2-4 minutes of extra operation room time, and accurately estimated the dimensions of surgical tools during human vitreoretinal surgery as a proof-of-concept and basis for further clinical utility.