Autonomous Needle Navigation in Subretinal Injections via iOCT.

Subretinal injection is an effective method for direct delivery of therapeutic agents to treat prevalent subretinal diseases. Among the challenges for surgeons are physiological hand tremor, difficulty resolving single-micron scale depth perception, and lack of tactile feedback. The recent introduction of intraoperative Optical Coherence Tomography (iOCT) enables precise depth information during subretinal surgery. However, even when relying on iOCT, achieving the required micron-scale precision remains a significant surgical challenge. This work presents a robot-assisted workflow for high-precision autonomous needle navigation for subretinal injection. The workflow includes online registration between robot and iOCT coordinates; tool-tip localization in iOCT coordinates using a Convolutional Neural Network (CNN); and tool-tip planning and tracking system using real-time Model Predictive Control (MPC). The proposed workflow is validated using a silicone eye phantom and ex vivo porcine eyes. The experimental results demonstrate that the mean error to reach the user-defined target and the mean procedure duration are within an acceptable precision range. The proposed workflow achieves a 100% success rate for subretinal injection, while maintaining scleral forces at the scleral insertion point below 15mN throughout the navigation procedures.

Steady-Hand Eye Robot 3.0: Optimization and Benchtop Evaluation for Subretinal Injection.

Subretinal injection methods and other procedures for treating retinal conditions and diseases (many considered incurable) have been limited in scope due to limited human motor control. This study demonstrates the next generation, cooperatively controlled Steady-Hand Eye Robot (SHER 3.0), a precise and intuitive-to-use robotic platform achieving clinical standards for targeting accuracy and resolution for subretinal injections. The system design and basic kinematics are reported and a deflection model for the incorporated delta stage and validation experiments are presented. This model optimizes the delta stage parameters, maximizing the global conditioning index and minimizing torsional compliance. Five tests measuring accuracy, repeatability, and deflection show the optimized stage design achieves a tip accuracy of < 30 µm, tip repeatability of 9.3 µm and 0.02°, and deflections between 20-350 µm/N. Future work will use updated control models to refine tip positioning outcomes and will be tested on in vivo animal models.

ASSESSMENT OF SIMULATED SURGICAL DEXTERITY AFTER MODIFIABLE EXTERNAL EXPOSURES AMONG NOVICE VERSUS EXPERIENCED VITREORETINAL SURGEONS.

PURPOSE: To evaluate novice and senior vitreoretinal surgeons after various exposures. Multiple comparisons ranked the importance of these exposures for surgical dexterity based on experience. METHODS: This prospective cohort study included 15 novice and 11 senior vitreoretinal surgeons (<2 and >10 years' practice, respectively). Eyesi-simulator tasks were performed after each exposure. Day 1, placebo, 2.5 mg/kg caffeine, and 5.0 mg/kg caffeine; day 2, placebo, 0.2 mg/kg propranolol, and 0.6 mg/kg propranolol; day 3, baseline simulation, breathalyzer readings of 0.06% to 0.10% and 0.11% to 0.15% blood alcohol concentrations; day 4, baseline simulation, push-up sets with 50% and 85% repetitions maximum; and day 5, 3-hour sleep deprivation. Eyesi-generated score (0-700, worst-best), out-of-tolerance tremor (0-100, best-worst), task completion time (minutes), and intraocular pathway (in millimeters) were measured. RESULTS: Novice surgeons performed worse after caffeine (-29.53, 95% confidence interval [CI]: -57.80 to -1.27, P = 0.041) and alcohol (-51.33, 95% CI: -80.49 to -22.16, P = 0.001) consumption. Alcohol caused longer intraocular instrument movement pathways (212.84 mm, 95% CI: 34.03-391.65 mm, P = 0.02) and greater tremor (7.72, 95% CI: 0.74-14.70, P = 0.003) among novices. Sleep deprivation negatively affected novice performance time (2.57 minutes, 95% CI: 1.09-4.05 minutes, P = 0.001) and tremor (8.62, 95% CI: 0.80-16.45, P = 0.03); however, their speed increased after propranolol (-1.43 minutes, 95% CI: -2.71 to -0.15 minutes, P = 0.029). Senior surgeons' scores deteriorated only following alcohol consumption (-47.36, 95% CI: -80.37 to -14.36, P = 0.005). CONCLUSION: Alcohol compromised all participants despite their expertise level. Experience negated the effects of caffeine, propranolol, exercise, and sleep deprivation on surgical skills.

Autonomous Needle Navigation in Retinal Microsurgery: Evaluation in ex vivo Porcine Eyes.

Important challenges in retinal microsurgery include prolonged operating time, inadequate force feedback, and poor depth perception due to a constrained top-down view of the surgery. The introduction of robot-assisted technology could potentially deal with such challenges and improve the surgeon's performance. Motivated by such challenges, this work develops a strategy for autonomous needle navigation in retinal microsurgery aiming to achieve precise manipulation, reduced end-to-end surgery time, and enhanced safety. This is accomplished through real-time geometry estimation and chance-constrained Model Predictive Control (MPC) resulting in high positional accuracy while keeping scleral forces within a safe level. The robotic system is validated using both open-sky and intact (with lens and partial vitreous removal) ex vivo porcine eyes. The experimental results demonstrate that the generation of safe control trajectories is robust to small motions associated with head drift. The mean navigation time and scleral force for MPC navigation experiments are 7.208 s and 11.97 mN, which can be considered efficient and well within acceptable safe limits. The resulting mean errors along lateral directions of the retina are below 0.06 mm, which is below the typical hand tremor amplitude in retinal microsurgery.

Robotic Navigation Autonomy for Subretinal Injection via Intelligent Real-Time Virtual iOCT Volume Slicing.

In the last decade, various robotic platforms have been introduced that could support delicate retinal surgeries. Concurrently, to provide semantic understanding of the surgical area, recent advances have enabled microscope-integrated intraoperative Optical Coherent Tomography (iOCT) with high-resolution 3D imaging at near video rate. The combination of robotics and semantic understanding enables task autonomy in robotic retinal surgery, such as for subretinal injection. This procedure requires precise needle insertion for best treatment outcomes. However, merging robotic systems with iOCT introduces new challenges. These include, but are not limited to high demands on data processing rates and dynamic registration of these systems during the procedure. In this work, we propose a framework for autonomous robotic navigation for subretinal injection, based on intelligent real-time processing of iOCT volumes. Our method consists of an instrument pose estimation method, an online registration between the robotic and the iOCT system, and trajectory planning tailored for navigation to an injection target. We also introduce intelligent virtual B-scans, a volume slicing approach for rapid instrument pose estimation, which is enabled by Convolutional Neural Networks (CNNs). Our experiments on ex-vivo porcine eyes demonstrate the precision and repeatability of the method. Finally, we discuss identified challenges in this work and suggest potential solutions to further the development of such systems.

Simultaneous Online Registration-Independent Stiffness Identification and Tip Localization of Surgical Instruments in Robot-assisted Eye Surgery.

Notable challenges during retinal surgery lend themselves to robotic assistance which has proven beneficial in providing a safe steady-hand manipulation. Efficient assistance from the robots heavily relies on accurate sensing of surgery states (e.g. instrument tip localization and tool-to-tissue interaction forces). Many of the existing tool tip localization methods require preoperative frame registrations or instrument calibrations. In this study using an iterative approach and by combining vision and force-based methods, we develop calibration- and registration-independent (RI) algorithms to provide online estimates of instrument stiffness (least squares and adaptive). The estimations are then combined with a state-space model based on the forward kinematics (FWK) of the Steady-Hand Eye Robot (SHER) and Fiber Brag Grating (FBG) sensor measurements. This is accomplished using a Kalman Filtering (KF) approach to improve the deflected instrument tip position estimations during robot-assisted eye surgery. The conducted experiments demonstrate that when the online RI stiffness estimations are used, the instrument tip localization results surpass those obtained from pre-operative offline calibrations for stiffness.

A Polysomnographic Study of Effects of Sleep Deprivation on Novice and Senior Surgeons during Simulated Vitreoretinal Surgery.

PURPOSE: To assess the impact of a 3-hour polysomnography (PSG)-recorded night of sleep deprivation on next-morning simulated microsurgical skills among vitreoretinal (VR) surgeons with different levels of surgical experience and associate the sleep parameters obtained by PSG with Eyesi-generated performance. DESIGN: Self-controlled cohort study. PARTICIPANTS: Eleven junior VR surgery fellows with < 2 years of surgical experience and 11 senior surgeons with > 10 years of surgical practice. METHODS: Surgical performance was assessed at 7am after a 3-hour sleep-deprived night using the Eyesi simulator and compared with each subject's baseline performance. MAIN OUTCOME MEASURES: Changes in Eyesi-generated score (0-700, worst to best), time for task completion (minutes), tremor-specific score (0-100, worst to best), and out-of-tolerance tremor percentage. Polysomnography was recorded during sleep deprivation. RESULTS: Novice surgeons had worse simulated surgical performance after sleep deprivation compared with self-controlled baseline dexterity in the total score (559.1 ± 39.3 vs. 593.8 ± 31.7; P = 0.041), time for task completion (13.59 ± 3.87 minutes vs. 10.96 ± 1.95 minutes; P = 0.027), tremor-specific score (53.8 ± 19.7 vs. 70.0 ± 15.3; P = 0.031), and out-of-tolerance tremor (37.7% ± 11.9% vs. 28.0% ± 9.2%; P = 0.031), whereas no performance differences were detected in those parameters among the senior surgeons before and after sleep deprivation (P ≥ 0.05). Time for task completion increased by 26% (P = 0.048) in the post-sleep deprivation simulation sessions for all participants with a high apnea-hypopnea index (AHI) and by 37% (P = 0.008) among surgeons with fragmented sleep compared with those with normal AHI and < 10 arousals per hour, respectively. Fragmented sleep was the only polysomnographic parameter associated with a worse Eyesi-generated score, with a 10% (P = 0.005) decrease the following morning. CONCLUSIONS: This study detected impaired simulated surgical dexterity among novice surgeons after acute sleep deprivation, whereas senior surgeons maintained their surgical performance, suggesting that the impact of poor sleep quality on surgical skills is offset by increased experience. When considering the 2 study groups together, sleep fragmentation and AHI were associated with jeopardized surgical performance after sleep deprivation. FINANCIAL DISCLOSURE(S): The authors have no proprietary or commercial interest in any materials discussed in this article.

Weight-adjusted caffeine and ß-blocker use in novice versus senior retina surgeons: a self-controlled study of simulated performance.

BACKGROUND/OBJECTIVES: Tremor and expertise are potentially influenced variables in vitreoretinal surgery. We investigated whether surgeon experience impacts the association of microsurgical performance with caffeine and ß-blockers weight-adjusted intake. SUBJECTS/METHODS: Novice and senior surgeons (<2 and >10 practice years, respectively) were recruited in this self-controlled, cross-sectional study. A simulator's task sequence was repeated over 2 days, 30 min after the following exposures: day 1, placebo, 2.5 mg/kg caffeine, 5.0 mg/kg caffeine, and 0.6 mg/kg propranolol; and day 2, placebo, 0.2 mg/kg propranolol, 0.6 mg/kg propranolol, and 5.0 mg/kg caffeine. Outcomes were total score (0-700, worst-best), simulation time (minutes), intraocular trajectory (centimeters), and tremor-specific score (0-100, worst-best). RESULTS: We recruited 15 novices (9 men [60%], 1.33 ± 0.49 practice years) and 11 seniors (8 men [72.7%], 16.00 ± 4.24 practice years). Novices performed worse after 2.5 mg/kg caffeine and improved following 0.2 mg/kg propranolol in total score (557 vs. 617, p = 0.009), trajectory (229.86 vs. 208.07, p = 0.048), time (14.9 vs. 12.7, p = 0.048), and tremor-score (55 vs. 75, p = 0.009). Surgical performance improved with propranolol post-caffeine but remained worse than 0.2 mg/kg propranolol in total score (570 vs. 617, p = 0.014), trajectory (226.59 vs. 208.07, p = 0.033), and tremor-score (50 vs. 75, p = 0.029). Seniors' tremor-score was lower after 2.5 mg/kg caffeine compared to 0.2 mg/kg propranolol (8 vs. 37, p = 0.015). Tremor-score following propranolol post-caffeine remained inferior to 0.6 mg/kg propranolol alone (17 vs. 38, p = 0.012). CONCLUSION: While caffeine and propranolol were associated with performance changes among novices, only tremor was affected in seniors, without dexterity changes. The pharmacologic exposure impact on surgical dexterity seems to be offset by increased experience.

Effects of fetal haemoglobin on systemic oxygenation in preterm infants and the development of retinopathy of prematurity PacIFiHER Report No. 2.

BACKGROUND/AIMS: Fetal haemoglobin (HbF) has an oxyhaemoglobin dissociation curve that may affect systemic oxygenation and the development of retinopathy of prematurity (ROP). The study aim is to characterise the effects of HbF levels on systemic oxygenation and ROP development. METHODS: Prospective study conducted from 1 September 2017 through 31 December 2018 at the Johns Hopkins NICU. Preterm infants with HbF measured at birth, 31, 34 and 37 weeks post-menstrual age (PMA), complete blood gas and SpO2 recorded up to 42 weeks PMA, and at least one ROP exam were included. RESULTS: Sixty-four preterm infants were enrolled. Higher HbF was associated with significantly higher SpO2, lower PCO2, lower FiO2 from birth to 31 weeks PMA and 31 to 34 weeks PMA (rs=0.51, rs=-0.62 and rs=-0.63; p<0.0001 and rs=0.71, rs=-0.58 and rs=-0.79; p<0.0001, respectively). To maintain oxygen saturation goals set by the neonatal intensive care unit, higher median FiO2 was required for HbF in the lowest tercile from birth compared with HbF in the highest tercile to 31 weeks and 31 to 34 weeks PMA; FiO2=35 (21-100) versus 21 (21-30) p<0.006 and FiO2=30 (28-100) versus 21 (21-30) p<0.001, respectively. Preterm infants with ROP had poorer indices of systemic oxygenation, as measured by median levels of SpO2 and PCO2, and lower levels of HbF (p<0.039 and p<0.0001, respectively) up to 34 weeks PMA. CONCLUSION: Low HbF levels correlated with poor oxygenation indices and increased risk for ROP. O2 saturation goals to prevent ROP may need to incorporate relative amount of HbF.

Human-Robot Interaction in Retinal Surgery: A Comparative Study of Serial and Parallel Cooperative Robots.

Cooperative robots for intraocular surgery allow surgeons to perform vitreoretinal surgery with high precision and stability. Several robot structural designs have shown capabilities to perform these surgeries. This research investigates the comparative performance of a serial and parallel cooperative-controlled robot in completing a retinal vessel-following task, with a focus on human-robot interaction performance and user experience. Our results indicate that despite differences in robot structure and interaction forces and torques, the two robots exhibited similar levels of performance in terms of general robot-to-patient interaction and average operating time. These findings have implications for the development and implementation of surgical robotics, suggesting that both serial and parallel cooperative-controlled robots can be effective for vitreoretinal surgery tasks.

ColibriDoc: An Eye-in-Hand Autonomous Trocar Docking System.

Retinal surgery is a complex medical procedure that requires exceptional expertise and dexterity. For this purpose, several robotic platforms are currently under development to enable or improve the outcome of microsurgical tasks. Since the control of such robots is often designed for navigation inside the eye in proximity to the retina, successful trocar docking and insertion of the instrument into the eye represents an additional cognitive effort, and is therefore one of the open challenges in robotic retinal surgery. For this purpose, we present a platform for autonomous trocar docking that combines computer vision and a robotic setup. Inspired by the Cuban Colibri (hummingbird) aligning its beak to a flower using only vision, we mount a camera onto the endeffector of a robotic system. By estimating the position and pose of the trocar, the robot is able to autonomously align and navigate the instrument towards the Trocar Entry Point (TEP) and finally perform the insertion. Our experiments show that the proposed method is able to accurately estimate the position and pose of the trocar and achieve repeatable autonomous docking. The aim of this work is to reduce the complexity of the robotic setup prior to the surgical task and therefore, increase the intuitiveness of the system integration into clinical workflow.

Delta Robot Kinematic Calibration for Precise Robot-Assisted Retinal Surgery.

High precision is required for ophthalmic robotic systems. This paper presents the kinematic calibration for the delta robot which is part of the next generation of Steady-Hand Eye Robot (SHER). A linear error model is derived based on geometric error parameters. Two experiments with different ranges of workspace are conducted with laser sensors measuring displacement. The error parameters are identified and applied in the kinematics to compensate for modeling error. To achieve better accuracy, Bernstein polynomials are adopted to fit the error residuals after compensation. After the kinematic calibration process, the error residuals of the delta robot are reduced to satisfy the clinical requirements.

Force-based Control for Safe Robot-assisted Retinal Interventions: In Vivo Evaluation in Animal Studies.

In recent years, robotic assistance in vitreoretinal surgery has moved from a benchtop environment to the operating rooms. Emerging robotic systems improve tool manoeuvrability and provide precise tool motions in a constrained intraocular environment and reduce/remove hand tremor. However, often due to their stiff and bulky mechanical structure, they diminish the perception of tool-to-sclera (scleral) forces, on which the surgeon relies, for eyeball manipulation. In this paper we measure these scleral forces and actively control the robot to keep them under a predefined threshold. Scleral forces are measured using a Fiber Bragg Grating (FBG) based force sensing instrument in an in vivo rabbit eye model in manual, cooperative robotic assistance with no scleral force control (NC), adaptive scleral force norm control (ANC) and adaptive scleral force component control (ACC) methods. To the best of our knowledge, this is the first time that the scleral forces are measured in an in vivo eye model during robot assisted vitreoretinal procedures. An experienced retinal surgeon repeated an intraocular tool manipulation (ITM) task 10 times in four in vivo rabbit eyes and a phantom eyeball, for a total of 50 repetitions in each control mode. Statistical analysis shows that the ANC and ACC control schemes restrict the duration of the undesired scleral forces to 4.41% and 14.53% as compared to 43.30% and 35.28% in manual and NC cases, respectively during the in vivo studies. These results show that the active robot control schemes can maintain applied scleral forces below a desired threshold during robot-assisted vitreoretinal surgery. The scleral forces measurements in this study may enable a better understanding of tool-to-sclera interactions during vitreoretinal surgery and the proposed control strategies could be extended to other microsurgery and robot-assisted interventions.

Surgical scene generation and adversarial networks for physics-based iOCT synthesis.

The development and integration of intraoperative optical coherence tomography (iOCT) into modern operating rooms has motivated novel procedures directed at improving the outcome of ophthalmic surgeries. Although computer-assisted algorithms could further advance such interventions, the limited availability and accessibility of iOCT systems constrains the generation of dedicated data sets. This paper introduces a novel framework combining a virtual setup and deep learning algorithms to generate synthetic iOCT data in a simulated environment. The virtual setup reproduces the geometry of retinal layers extracted from real data and allows the integration of virtual microsurgical instrument models. Our scene rendering approach extracts information from the environment and considers iOCT typical imaging artifacts to generate cross-sectional label maps, which in turn are used to synthesize iOCT B-scans via a generative adversarial network. In our experiments we investigate the similarity between real and synthetic images, show the relevance of using the generated data for image-guided interventions and demonstrate the potential of 3D iOCT data synthesis.

Force and Velocity Based Puncture Detection in Robot Assisted Retinal Vein Cannulation: In-Vivo Study.

OBJECTIVE: Retinal vein cannulation is a technically demanding surgical procedure and its feasibility may rely on using advanced surgical robots equipped with force-sensing microneedles. Reliable detection of the moment of venous puncture is important, to either alert or prevent the clinician from double puncturing the vessel and damaging the retinal surface beneath. This paper reports the first in-vivo retinal vein cannulation trial on rabbit eyes, using sensorized metal needles, and investigates puncture detection. METHODS: We utilized total of four indices including two previously demonstrated ones and two new indices, based on the velocity and force of the needle tip and the correlation between the needle-tissue and tool-sclera interaction forces. We also studied the effect of detection timespan on the performance of detecting actual punctures. RESULTS: The new indices, when used in conjunction with the previous algorithm, improved the detection rate form 75% to 92%, but slightly increased the number of false detections from 37 to 43. Increasing the detection window improved the detection performance, at the cost of adding to the delay. CONCLUSION: The current algorithm can supplement the surgeons' visual feedback and surgical judgment. To achieve automatic puncture detection, more measurements and further analysis are required. Subsequent in-vivo studies in other animals, such as pigs with their more human like eye anatomy, are required, before clinical trials. SIGNIFICANCE: The study provides promising results and the criteria developed may serve as guidelines for further investigation into puncture detection in in-vivo retinal vein cannulation.

OCT-guided Robotic Subretinal Needle Injections: A Deep Learning-Based Registration Approach.

Subretinal injection (SI) is an ophthalmic surgical procedure that allows for the direct injection of therapeutic substances into the subretinal space to treat vitreoretinal disorders. Although this treatment has grown in popularity, various factors contribute to its difficulty. These include the retina's fragile, nonregenerative tissue, as well as hand tremor and poor visual depth perception. In this context, the usage of robotic devices may reduce hand tremors and facilitate gradual and controlled SI. For the robot to successfully move to the target area, it needs to understand the spatial relationship between the attached needle and the tissue. The development of optical coherence tomography (OCT) imaging has resulted in a substantial advancement in visualizing retinal structures at micron resolution. This paper introduces a novel foundation for an OCT-guided robotic steering framework that enables a surgeon to plan and select targets within the OCT volume. At the same time, the robot automatically executes the trajectories necessary to achieve the selected targets. Our contribution consists of a novel combination of existing methods, creating an intraoperative OCT-Robot registration pipeline. We combined straightforward affine transformation computations with robot kinematics and a deep neural network-determined tool-tip location in OCT. We evaluate our framework's capability in a cadaveric pig eye open-sky procedure and using an aluminum target board. Targeting the subretinal space of the pig eye produced encouraging results with a mean Euclidean error of 23.8µm.

Mechanistic Insights into the Pathogenesis of Proliferative and Nonproliferative Vitreomacular Traction.

PURPOSE: To describe the vitreoretinal interface in vitreomacular traction (VMT) by using novel optical coherence tomography (OCT) methods; wide-angle montage, and pseudomotion OCT imaging systems. DESIGN: Observational case series. METHODS: Wide-angle montage OCT images of horizontal and vertical scans through the fovea were acquired in 50 eyes of 46 consecutive patients with VMT. Baseline fundus scans were obtained. These were followed by scans acquired with an eye-tracking system performed immediately after vertical and horizontal eye movements. Three scans were then superimposed to compare changes in the contour and position of the posterior vitreous. RESULTS: The subjects were classified as VMT with ("proliferative"; 48.0%) and without ("nonproliferative"; 52.0%) thickened posterior vitreous. Epiretinal membrane was observed in 26.9% of nonproliferative and 95.8% of proliferative VMT eyes (P = 3.6 × 10-7). No eye of proliferative and 57.7% of nonproliferative VMT eyes had wavy contoured posterior vitreous (P = 4.0 × 10-6). None with proliferative VMT, but 91.7% of nonproliferative VMT eyes, showed motion induced changes of posterior vitreous following eye movement (P = 2.0 × 10-8). The posterior vitreous detachment extended beyond the scanned area in 34.6% of nonproliferative and 8.3% of proliferative VMT eyes (P = .040). CONCLUSIONS: By dynamically evaluating the vitreoretinal interface of patients with VMT, the static contraction forces of a thickened posterior vitreous at the macula are implicated in proliferative VMT. This contractile force is not strongly implicated in the majority of VMT eyes with nontaut and more mobile vitreous (nonproliferative VMT). VMT and its associated complications are determined by at least 2 different pathophysiological mechanisms.

Posterior Vitreous Detachment in Normal Healthy Subjects Younger Than Age Twenty.

Purpose: To describe the initiation of posterior vitreous detachment (PVD) in the eyes of normal individuals, under 20 years of age, using wide-angle optical coherence tomography (OCT). Methods: This is an observational cross-sectional study. Montaged images of horizontal and vertical OCT scans were obtained in 63 healthy eyes of 35 consecutive subjects ranging in age from 4 to 17 years. Results: Forty-five eyes (71.4%) had obvious PVD, defined as a contiguous line of posterior cortical vitreous separated from the surface of the retina. Eighteen eyes (28.6%) had no PVD. The mean age of the individuals without PVD was significantly younger than those with PVD (P = 0.008). The spatial distribution of PVD initiation was highest in the superior quadrants, with the nasal, inferior, septum papillomaculae, and temporal quadrants following in descending order of frequency (P < 0.001). PVD was observed to begin anterior to the premacular liquefied lacuna, where the vitreous gel directly adheres to the vitreoretinal interface. In the majority of subjects (80.6%), PVD was initiated anterior to the vascular arcades. Conclusions: PVD can be observed by OCT to begin in the first and second decade of life. It begins in the mid-peripheral vitreous, most frequently in the superior quadrants anterior to the vascular arcades. In this study, all PVDs originated outside of the macular liquefied lacunae, where the vitreous gel adheres directly to the retina.

Towards Autonomous Eye Surgery by Combining Deep Imitation Learning with Optimal Control.

During retinal microsurgery, precise manipulation of the delicate retinal tissue is required for positive surgical outcome. However, accurate manipulation and navigation of surgical tools remain difficult due to a constrained workspace and the top-down view during the surgery, which limits the surgeon's ability to estimate depth. To alleviate such difficulty, we propose to automate the tool-navigation task by learning to predict relative goal position on the retinal surface from the current tool-tip position. Given an estimated target on the retina, we generate an optimal trajectory leading to the predicted goal while imposing safety-related physical constraints aimed to minimize tissue damage. As an extended task, we generate goal predictions to various points across the retina to localize eye geometry and further generate safe trajectories within the estimated confines. Through experiments in both simulation and with several eye phantoms, we demonstrate that our framework can permit navigation to various points on the retina within 0.089mm and 0.118mm in xy error which is less than the human's surgeon mean tremor at the tool-tip of 0.180mm. All safety constraints were fulfilled and the algorithm was robust to previously unseen eyes as well as unseen objects in the scene. Live video demonstration is available here: https://youtu.be/n5j5jCCelXk.