Applications of Artificial Intelligence in Corneal Nerve Images in Ophthalmology.

Background: This study aimed to compare the corneal nerve structural abnormalities detected using in vivo confocal microscopy (IVCM) in patients with neuropathic corneal pain (NCP) secondary to primary meibomian gland dysfunction (MGD) or autoimmune dry eye (AIDE). Methods: A two-stage retrospective nested case–control study was conducted. First, data from patients with either MGD or AIDE were assessed, selecting only cases with no corneal pain (VAS = 0) or severe pain (VAS ≥ 8). Ocular signs and symptoms of the 238 selected patients were compared between painful and painless cases. Next, painful patients with no corneal damage (Oxford score ≤ 1) were selected within each study group, defining the cases with NCP (i.e., “pain without stain”). IVCM images from all groups were compared with prospectively-recruited healthy controls, focusing on dendritiform cell density and nerve abnormalities (density, tortuosity, microneuromas). Results: AIDE patients had more ocular signs/symptoms than MGD patients. Compared with healthy controls, AIDE-related NCP patients showed increased nerve tortuosity and number of neuromas, whereas MGD-related NCP patients had reduced nerve density and increased number, perimeter, and area of microneuromas. Microneuromas were also observed in healthy controls. Furthermore, a higher number of microneuromas was found in MGD-related NCP compared to AIDE-related NCP or painless MGD. Conclusions: MGD-related NCP was associated with significantly more corneal nerve abnormalities than AIDE-related NCP or healthy controls. Although IVCM can be useful to detect NCP-related corneal nerve changes in such patients, the diagnosis of dry eye disease-related NCP will require an association of several IVCM-based criteria without relying solely on the presence of microneuromas.

Deep-learning based analysis of in-vivo confocal microscopy images of the subbasal corneal nerve plexus’ inferior whorl in patients with neuropathic corneal pain and dry eye disease

PurposeThis study was designed to compare and contrast quantitative data of the human corneal sub-basal nerve plexus (SBP) evaluated by two different methods: in vivo confocal microscopy (IVCM), and immunohistochemical staining of ex vivo donor corneas.MethodsSeven parameters of the SBP in large-scale IVCM mosaicking images from healthy subjects were compared with the identical parameters in ex vivo donor corneas stained by β-III-tubulin immunohistochemistry. Corneal nerve fiber length (CNFL), corneal nerve fiber density (CNFD), corneal nerve branch density (CNBD), average weighted corneal nerve fiber tortuosity (CNFTo), corneal nerve connection points (CNCP), average corneal nerve single-fiber length (CNSFL), and average weighted corneal nerve fiber thickness (CNFTh) were calculated using a dedicated, published algorithm and compared.ResultsOur experiments showed significantly higher values for CNFL (50.2 vs 21.4 mm/mm2), CNFD (1358.8 vs 277.3 nerve fibers/mm2), CNBD (847.6 vs 163.5 branches/mm2), CNFTo (0.095 vs 0.081 μm-1), and CNCP (49.4 vs 21.6 connections/mm2) in histologically staining specimens compared with IVCM images. In contrast, CNSFL values were higher in IVCM images than in histological specimens (32.1 vs 74.1 μm). No significant difference was observed in CNFTh (2.22 vs 2.20 μm) between the two groups.ConclusionsThe results of this study have shown that IVCM has an inherently lower resolution compared with ex vivo immunohistochemical staining of the corneal SBP and that this limitation leads to a systematic underestimation of several SBP parameters. Despite this shortcoming, IVCM is a vital clinical tool for in vivo characterization, quantitative clinical imaging, and evaluation of the human corneal SBP.

ObjectiveTo establish an automated corneal nerve analysis system for corneal in vivo confocal microscopy (IVCM) images from both the whorl-like corneal nerves in the inferior whorl (IW) region and the...

PurposeThe diagnosis of neuropathic corneal pain (NCP) is challenging, as it is often difficult to differentiate it from conventional dry eye disease (DED). In addition to eye pain, NCP can present with similar signs and symptoms of DED. The purpose of this study is to find an objective diagnostic sign to identify patients with NCP, using in vivo confocal microscopy (IVCM). MethodsThis was a comparative, retrospective, case-control study. Patients with clinical diagnosis of NCP (n = 25), DED (n = 30), and age- and sex-matched healthy controls (n = 16), who underwent corneal imaging with IVCM (HRT3/RCM) were included. Central corneal IVCM scans were analyzed by 2 masked observers for nerve density and number, presence of microneuromas (terminal enlargements of subbasal corneal nerve) and/or nerve beading (bead-like formation along the nerves), and dendritiform cell (DC) density. ResultsThere was a decrease in total nerve density in both NCP (14.14 ± 1.03 mm/mm2) and DED patients (12.86 ± 1.04 mm/mm2), as compared to normal controls (23.90 ± 0.92 mm/mm2; p < 0.001). However, total nerve density was not statistically different between NCP and DED patients (p = 0.63). Presence of nerve beading was not significantly different between patients and normal controls (p = 0.15). Interestingly, microneuromas were observed in all patients with NCP, while they were not present in any of the patients with conventional DED (sensitivity and specificity of 100%). DC density was significantly increased in both NCP (71.89 ± 16.91 cells/mm2) and DED patients (111.5 ± 23.86 cells/mm2), as compared to normal controls (24.81 ± 4.48 cells/mm2 (p < 0.05). However, there was no significant difference in DC density between DED and NCP patients (p = 0.31). ConclusionIVCM may be used as an adjunct diagnostic tool for the diagnosis of NCP in the presence of neuropathic symptoms. Microneuromas may serve as a sensitive and specific biomarker for the diagnosis of NCP.

Neuropathic corneal pain (NCP) is a growing problem in ophthalmology, but its precise pathology and treatment are unclear. NCP is often associated with hyperalgesia, allodynia, photo-allodynia and spontaneous pain with a normal slit-lamp examination [1, 2] . Neuropathic pain is defined as pain arising as the direct consequence of a lesion or disease affecting the somatosensory system [1] . Clinically, it should be distinguished from nociceptive pain, where there is detectable pathology in the corneal tissue/ ocular surface [1] . Several patients who suffer from NCP present with dry eye disease (DED) symptoms. It is frequent to see patients with presumed DED and severe ocular Pain but no clinical signs on slit-lamp examination. Such patients show no fluorescein corneal staining. These patients are referred to as having 'Pain without Stain ' [3]. This group of patients with NCP suffer from poor quality of life (QoL), chronic fatigue, pain, and depression with an increased overall burden of symptoms [4] . According to a study conducted in the UK, such ocular surface diseases with DED are associated with a 30% reduction in workplace performance, productivity, and non-work activities [4] . In this issue, Leonardi et al. present their investigations on patients with NCP and attempt to characterise them into pathological entities of neuropathic pain. They divide NCP into subgroups based on ocular surface evaluations and neuropsychiatric tests. They found a positive correlation between pain severity, post-traumatic stress disorder, and depression in patients with NCP compared to nociceptive cases [5], i.e. cases where corneal pathology can be detected. In addition to providing a classification approach and treatment recommendations, the authors offer helpful therapeutic insights. Some patients with NCP found relief from systemic therapy such as duloxetine or pregabalin. This is interesting, as there is a lack of consensus regarding the efficacy of systemic neurological treatment [2] . Leonardi et al. also investigated the use of in vivo confocal microscopy (IVCM) to detect corneal nerve morphology in patients with NCP, attempting to identify a newly described clinical feature of 'corneal neuromas' on IVCM [6] . The presence of these 'corneal neuromas' has been controversial. Some investigators suggest that these are diagnostic biomarkers and explain NCP [7] . However, these features have also been reported in healthy corneas [8], suggesting they are an inconsistent finding. Leonardi et al. similarly could not find a correlation between corneal neuromas and NCP [5] . This study reinforces the idea that treating patients with NCP whose mental health issues are as important as treating their eye defects. Perhaps then, we can provide relief to patients with 'Pain without Stain'.

PurposeIn vivo confocal microscopy (IVCM) is a noninvasive, reproducible, and inexpensive diagnostic tool for corneal diseases. However, widespread and effortless image acquisition in IVCM creates serious image analysis workloads on ophthalmologists, and neural networks could solve this problem quickly. We have produced a novel deep learning algorithm based on generative adversarial networks (GANs), and we compare its accuracy for automatic segmentation of subbasal nerves in IVCM images with a fully convolutional neural network (U-Net) based method.MethodsWe have collected IVCM images from 85 subjects. U-Net and GAN-based image segmentation methods were trained and tested under the supervision of three clinicians for the segmentation of corneal subbasal nerves. Nerve segmentation results for GAN and U-Net-based methods were compared with the clinicians by using Pearson's R correlation, Bland-Altman analysis, and receiver operating characteristics (ROC) statistics. Additionally, different noises were applied on IVCM images to evaluate the performances of the algorithms with noises of biomedical imaging.ResultsThe GAN-based algorithm demonstrated similar correlation and Bland-Altman analysis results with U-Net. The GAN-based method showed significantly higher accuracy compared to U-Net in ROC curves. Additionally, the performance of the U-Net deteriorated significantly with different noises, especially in speckle noise, compared to GAN.ConclusionsThis study is the first application of GAN-based algorithms on IVCM images. The GAN-based algorithms demonstrated higher accuracy than U-Net for automatic corneal nerve segmentation in IVCM images, in patient-acquired images and noise applied images. This GAN-based segmentation method can be used as a facilitating diagnostic tool in ophthalmology clinics.Translational RelevanceGenerative adversarial networks are emerging deep learning models for medical image processing, which could be important clinical tools for rapid segmentation and analysis of corneal subbasal nerves in IVCM images.

To investigate the longitudinal alterations of subbasal corneal nerves in patients with infectious keratitis (IK) during the acute phase, cessation of treatment, and the recovery phase by in vivo confocal microscopy (IVCM). Prospective, longitudinal, case-control, single-center study. Fifty-six eyes of 56 patients with the diagnosis of bacterial (n=28), fungal (n=15), or Acanthamoeba (n=13) keratitis were included in the study. Thirty eyes of 30 normal volunteers constituted the control group. Corneal sensation and serial IVCM of the central cornea were performed prospectively using the Heidelberg Retina Tomograph 3/Rostock Cornea Module (Heidelberg Engineering, Heidelberg, Germany). The IVCM images were assessed at 3 time points: at the acute phase (first visit to the cornea service), at cessation of antimicrobial treatment, and up to 6 months after the resolution of infection. Total nerve number and length, main nerve trunks, branching, and corneal sensation were assessed during the follow-up period. Corneal nerves were reduced significantly during the acute phase in eyes with IK compared with controls across all subgroups, with total nerve length of 5.47±0.69 mm/mm2 versus 20.59±1.06 mm/mm2 (P<0.0001). At the cessation of treatment, corneal nerves in patients with IK had regenerated, including total nerve length (8.49±0.94 mm/mm2; P=0.02) and nerve branch length (4.80±0.37 mm/mm2; P=0.005). During the recovery phase, after resolution of infection, corneal nerves regenerated further, including total nerve length (12.13±1.97 mm/mm2; P=0.005), main nerve trunk length (5.80±1.00 mm/mm2; P=0.01), and nerve branch length (6.33±0.76 mm/mm2; P=0.003) as compared with the acute phase, but were still significantly lower when compared with controls (P<0.05 for all parameters). Corneal degeneration and regeneration correlated with corneal sensation (r=0.47; P=0.0009). Patients with IK who sustain profound loss of corneal nerves during the acute phase of infection demonstrate increased corneal nerve density during the first 6 months after the resolution of infection. However, despite significant nerve regeneration, corneal nerve density does not recover fully and remains low compared to controls. By providing an objective methodology to monitor corneal re-innervation, IVCM adds potentially important findings that may have implications for clinical management and surgical planning.

This study aimed to observe corneal subbasal nerves and Langerhans cells (LCs) using in vivo confocal microscopy (IVCM) in patients with dry eye, a tool for the evaluation of disease stage and severity and for treatment monitoring at the microstructural level. A total of 107 eyes from 62 patients were included. The Ocular Surface Disease Index (OSDI) questionnaire and other examinations were used to assess dry eye symptoms and signs. IVCM was performed to observe subbasal corneal nerves and LCs. Corneal nerves were graded using both objective and subjective methods. The correlations between dry eye symptoms and corneal nerve parameters, corneal nerve grading, and LC number were analyzed. Corneal nerve length was negatively correlated with sensitivity to light [correlation coefficient (CC)= -0.24, P < 0.05]; nerve width was positively correlated with the OSDI score, painful eyes, and blurred vision (CC = 0.41, 0.23, and 0.46, respectively, all P < 0.05); and nerve tortuosity was positively correlated with sensitivity to light (CC = 0.23, P < 0.05). Moreover, both total objective and subjective grading scores were positively correlated with OSDI scores (CC = 0.48 and 0.27, respectively, both P < 0.05). LC number was found not to be significantly correlated with dry eye symptoms (P > 0.05). IVCM is a useful tool to evaluate corneal subbasal nerve changes in patients with dry eye. Detailed nerve grading could help to understand and evaluate the pathophysiologic conditions of the disease and could be used for further treatment follow-up in the future.

Chemotherapy-induced neurotoxicity is an increasingly recognized clinical issue in oncology. in vivo confocal microscopy (IVCM) of corneal nerves has been successfully used to diagnose peripheral neuropathies, including diabetic neuropathy. The purpose of this study was to test if the combination of corneal nerve density and morphology assessed by IVCM is useful to monitor the neurotoxic effects of chemotherapy compared to epidermal nerve quantification. Overall, 95 adult patients with different cancer types were recruited from the oncology and hematology departments of the San Raffaele Hospital. Neurological examination, including clinical Total Neuropathy Score, and in vivo corneal confocal microscopy (IVCM), were performed before and after chemotherapy. In a group of 14 patients, skin biopsy was performed at the first and last visit. In the group of 14 patients who underwent both skin biopsy and corneal nerve imaging, clinical worsening (+69%, p = 0.0018) was paralleled by corneal nerve fiber (CNF) density reduction (−22%, p = 0.0457). Clinical Total neuropathy score significantly worsened from the first to the last visit (+62%, p < 0.0001). CNF length was not significantly reduced overall. However, CNF density/tortuosity ratio significantly decreased after therapy. Correlation analysis showed that the CNF density/tortuosity ratio was also correlated with the number of chemotherapy cycles (r = −0.04790, P = 0.0009). Our data confirm that in vivo corneal confocal microscopy is a helpful, non-invasive tool which shows promise for the diagnosis of chemotherapy-induced peripheral neuropathies. IVCM could allow a rapid, reproducible and non-invasive quantification of peripheral nerve pathology in chemotherapy-associated neuropathy.

Background/aimsThis systematic review critically evaluated peer-reviewed publications describing morphological features consistent with, or using terms related to, a ‘neuroma’ or ‘microneuroma’ in the human cornea using laser-scanning in vivo confocal...

PurposeThe purpose of this study was to establish a deep learning model for automated sub-basal corneal nerve fiber (CNF) segmentation and evaluation with in vivo confocal microscopy (IVCM).MethodsA corneal nerve segmentation network (CNS-Net) was established with convolutional neural networks based on a deep learning algorithm for sub-basal corneal nerve segmentation and evaluation. CNS-Net was trained with 552 and tested on 139 labeled IVCM images as supervision information collected from July 2017 to December 2018 in Peking University Third Hospital. These images were labeled by three senior ophthalmologists with ImageJ software and then considered ground truth. The areas under the receiver operating characteristic curves (AUCs), mean average precision (mAP), sensitivity, and specificity were applied to evaluate the efficiency of corneal nerve segmentation. The relative deviation ratio (RDR) was leveraged to evaluate the accuracy of the corneal nerve fiber length (CNFL) evaluation task.ResultsThe model achieved an AUC of 0.96 (95% confidence interval [CI] = 0.935–0.983) and an mAP of 94% with minimum dice coefficient loss at 0.12. For our dataset, the sensitivity was 96% and specificity was 75% in the CNF segmentation task, and an RDR of 16% was reported in the CNFL evaluation task. Moreover, the model was able to segment and evaluate as many as 32 images per second, much faster than skilled ophthalmologists.ConclusionsWe established a deep learning model, CNS-Net, which demonstrated a high accuracy and fast speed in sub-basal corneal nerve segmentation with IVCM. The results highlight the potential of the system in assisting clinical practice for corneal nerves segmentation and evaluation.Translational RelevanceThe deep learning model for IVCM images may enable rapid segmentation and evaluation of the corneal nerve and may provide the basis for the diagnosis and treatment of ocular surface diseases associated with corneal nerves.

Neuropathic corneal pain (NCP) is an underdiagnosed ocular disorder caused by aberrant nociception and hypersensitivity of corneal nerves, often resulting in chronic pain and discomfort even in the absence of noxious stimuli. Recently, microneuromas (aberrant growth and swelling of the corneal nerve endings) detected using in vivo confocal microscopy (IVCM) have emerged as a promising biomarker for NCP. However, this process is time-intensive and error-prone, limiting its clinical use and availability. In this work, we present a new NCP screening system based on a deep learning model trained to detect microneuromas using a multisite dataset with a combined total of 103,168 IVCM images. Our model showed excellent discriminative ability detecting microneuromas (AuROC: 0.97) and the ability to generalize to data from a new institution (AuROC: 0.90). Additionally, our pipeline provides an uncertainty quantification mechanism that allows it to communicate when its predictions are reliable, further increasing its clinical relevance.

NerveStitcher: Corneal confocal microscope images stitching with neural networks

Exploration of imaging and molecular biomarkers for differentiation of neuropathic corneal pain from dry eye syndrome.