Automatic analysis of corneal nerves imaged using in vivo confocal microscopy

[Clinical profile of corneal sensitivity in diabetic patients: A case-control study]

PURPOSE

To evaluate the corneal sensitivity of black diabetic patients and identify factors associated with changes in corneal sensitivity.

METHODOLOGY

We conducted a cross-sectional comparative case-control study at the National Obesity Center of the Yaounde Central Hospital and the Djoungolo District Hospital from March 1 to July 31, 2022. Corneal sensitivity was measured using the Cochet-Bonnet esthesiometer in all diabetic patients over 18 years of age, matched for age and sex to a clinically healthy control population. Data were analyzed using SPSS version 23.0 software. A P-value of less than 5% was considered significant.

RESULTS

A total of 111 diabetic and 111 non-diabetic patients participated in the study. The mean age was 53.46±12.74 years for diabetics and 52.85±11.77 years for non-diabetics (P=0.901). The mean duration of diabetes was 6.4±5.30 years. Corneal sensitivity in diabetics was lower (44.56±9.59mm) compared to non-diabetics (53.59±6.30mm) with a statistically significant difference (P=0.000). Factors associated with decrease in corneal sensitivity in diabetics were duration of diabetes and poor glycemic control.

CONCLUSION

Decrease in corneal sensitivity related to diabetes is a complication to be systematically screened for during the ophthalmologic follow-up of diabetic patients.

Objective Analysis of Corneal Nerves and Dendritic Cells

Corneal nerves and dendritic cells are increasingly being visualised to serve as clinical parameters in the diagnosis of ocular surface diseases using intravital confocal microscopy. In this review, different methods of image analysis are presented. The use of deep learning algorithms, which enable automated pattern recognition, is explained in detail using our own developments and compared with other established methods.

Morphological Differentiation of Corneal Inflammatory Cells: Proposal of Pragmatic Protocol

PURPOSE

Corneal confocal microscopy is a noninvasive imaging technique to analyze corneal nerve fibers and corneal inflammatory cells (CICs). The amount of CICs is a potential biomarker of disease activity in chronic autoinflammatory diseases. To date, there are no standardized criteria for the morphological characterization of CICs. The aim was to establish a protocol for a standardized morphological classification of CICs based on a literature search and to test this protocol for applicability and reliability.

METHODS

A systematic review of the literature about definitions of CICs was conducted. Existing morphological descriptions were translated into a structured algorithm and applied by raters. Subsequently, the protocol was optimized by reducing and defining the criteria of the cell types. The optimized algorithm was applied by 4 raters. The interrater reliability was calculated using Fleiss kappa (K).

RESULTS

A systematic review of the literature revealed no uniform morphological criteria for the differentiation of the individual cell types in CICs. Our first protocol achieved only a low level of agreement between 3 raters (K = 0.09; 1062 rated cells). Our revised protocol was able to achieve a higher interrater reliability with 3 (K = 0.64; 471 rated cells) and 4 (K = 0.61; 628 rated cells) raters.

CONCLUSIONS

The indirect use of criteria from the literature leads to a high error rate. By clearly defining the individual cell types and standardizing the protocol, reproducible results were obtained, allowing the introduction of this protocol for the future evaluation of CICs in the corneal confocal microscopy.

Impact of SGLT2 Inhibitors on Corneal Nerve Morphology and Dendritic Cell Density in Type 2 Diabetes

PURPOSE

This study aims to determine the effects of SGLT2 inhibitors on corneal dendritic cell density and corneal nerve measures in type 2 diabetes.

METHODS

Corneal dendritic cell densities and nerve parameters were measured in people with type 2 diabetes treated with SGLT2 inhibitors (T2DM-SGLT2i) [n = 23] and those not treated with SGLT2 inhibitors (T2DM-no SGLT2i) [n = 23], along with 24 age and sex-matched healthy controls.

RESULTS

There was a reduction in all corneal nerve parameters in type 2 diabetes groups compared to healthy controls (All parameters: p < 0.05). No significant differences in corneal nerve parameters were observed between T2DM-SGLT2i and T2DM-no SGLT2i groups (All parameters: p > 0.05). Central corneal dendritic cells were significantly reduced [mature (p = 0.03), immature (p = 0.06) and total (p = 0.002)] in the T2DM-SGLT2i group compared to the T2DM-no SGLT2i group. Significantly, higher mature (p = 0.04), immature (p = 0.004), total (p = 0.002) dendritic cell densities in the T2DM-no SGLT2i group were observed compared to the healthy controls. In the inferior whorl, no significant difference in immature (p = 0.27) and total dendritic cell densities (p = 0.16) between T2DM-SGLT2i and T2DM-no SGLT2i were observed except mature dendritic cell density (p = 0.018). No differences in total dendritic cell density were observed in the central (p > 0.09) and inferior whorl (p = 0.88) between T2DM-SGLT2i and healthy controls.

CONCLUSION

The present study showed a reduced dendritic cell density in people with type 2 diabetes taking SGLT2 inhibitors compared to those not taking these medications.

Neurophysiology of corneal neuropathic pain and emerging pharmacotherapeutics

The altered activity generated by corneal neuronal injury can result in morphological and physiological changes in the architecture of synaptic connections in the nervous system. These changes can alter the sensitivity of neurons (both second-order and higher-order projection) projecting pain signals. A complex process involving different cell types, molecules, nerves, dendritic cells, neurokines, neuropeptides, and axon guidance molecules causes a high level of sensory rearrangement, which is germane to all the phases in the pathomechanism of corneal neuropathic pain. Immune cells migrating to the region of nerve injury assist in pain generation by secreting neurokines that ensure nerve depolarization. Furthermore, excitability in the central pain pathway is perpetuated by local activation of microglia in the trigeminal ganglion and alterations of the descending inhibitory modulation for corneal pain arriving from central nervous system. Corneal neuropathic pain may be facilitated by dysfunctional structures in the central somatosensory nervous system due to a lesion, altered synaptogenesis, or genetic abnormality. Understanding these important pathways will provide novel therapeutic insight.

Fully Automatic, Semiautomatic, and Manual Corneal Nerve Fiber Analysis in Patients With Sarcoidosis

Purpose

No guidelines are available on the preferred method for analyzing corneal confocal microscopy (CCM) data. Manual, semiautomatic, and automatic analyzes are all currently in use. The purpose of the present study was threefold. First, we aimed to investigate the different methods for CCM analysis in patients with and without small fiber neuropathy (SFN). Second, to determine the correlation of different methods for measuring corneal nerve fiber length (CNFL) and nerve fiber area (NFA). Finally, we investigated the added value of automatic NFA analysis.

Methods

We included 20 healthy controls and 80 patients with sarcoidosis, 31 with established SFN and 49 without SFN. The CNFL was measured using CCMetrics, ACCMetrics, and NeuronJ. NFA was measured with NFA FIJI and ACCMetrics NFA.

Results

CNFL and NFA could not distinguish sarcoidosis with and without SFN or healthy controls. CCMetrics, NeuronJ, and ACCMetrics CNFL highly correlated. Also, NFA FIJI and ACCMetrics NFA highly correlated. Reproducing a nonlinear formula between CNFL and NFA confirmed the quadratic relation between NFA FIJI and ACCMetrics CNFL. CCMetrics and NeuronJ instead showed a square root relationship and seem to be less comparable owing to differences between automatic and manual techniques.

Conclusions

ACCMetrics can be used for fully automatic analysis of CCM images to optimize efficiency. However, CNFL and NFA do not seem to have a discriminatory value for SFN in sarcoidosis. Further research is needed to determine the added value and normative values of NFA in CCM analysis.

Translational Relevance

Our study improves the knowledge about CCM software and pathophysiology of SFN.

Decreased corneal subbasal nerve fiber length and density in diabetic dogs with cataracts using in vivo confocal microscopy

OBJECTIVE

To determine whether there is a difference in corneal sensitivity and corneal subbasal nerve plexus (CSNP) morphology in cataractous dogs with diabetes mellitus (DM) versus without DM.

ANIMALS STUDIED

Twenty six domestic dogs with cataracts of various breeds presented for phacoemulsification, 13 with DM and 13 without DM.

PROCEDURE

The inclusion criteria for the study were dogs with bilateral cataracts and no clinical evidence of corneal disease. The diabetic group had documented hyperglycemia and was currently treated with insulin. The non-diabetic group had no evidence of DM on examination and bloodwork. Complete ophthalmic examination, corneal esthesiometry, and in vivo confocal microscopy of the CSNP was performed for both eyes of each dog. The CSNP was evaluated using a semi-automated program and statistically analyzed.

RESULTS

The mean (±SD) CSNP fiber length was significantly decreased in diabetic (3.8 ± 3.0 mm/mm2 ) versus non-diabetic (6.7 ± 1.9 mm/mm2 ) dogs. Likewise, the mean (±SD) fiber density was significantly decreased in diabetic (8.3 ± 3.1 fibers/mm2 ) versus non-diabetic (15.5 ± 4.9 fibers/mm2 ) dogs. The corneal touch threshold was significantly reduced in diabetic (2.1 ± 0.8 cm) versus non-diabetic (2.8 ± 0.4 cm) dogs. There was a non-significant trend towards subclinical keratitis in diabetic (9/13) versus non-diabetic (4/13) dogs.

CONCLUSIONS

Morphological and functional abnormalities of the CSNP were present in dogs with DM, including decreased fiber length, fiber density, and corneal sensitivity. These findings are consistent with diabetic neuropathy and could contribute to clinically significant corneal complications after cataract surgery.

Conjunctival microcirculation in ocular and systemic microvascular disease

The conjunctival microcirculation is an accessible complex network of micro vessels whose quantitative assessment can reveal microvascular haemodynamic properties. Currently, algorithms for the measurement of conjunctival haemodynamics use either manual or semi-automated systems, which may provide insight into overall conjunctival health, as well as in ocular and systemic disease. These algorithms include functional slit-lamp biomicroscopy, laser doppler flowmetry, optical coherence tomography angiography, orthogonal polarized spectral imaging, computer-assisted intravitral microscopy, diffuse reflectance spectroscopy and corneal confocal microscopy. Furthermore, several studies have demonstrated a relationship between conjunctival microcirculatory haemodynamics and many diseases such as dry eye disease, Alzheimer's disease, diabetes, hypertension, sepsis, coronary microvascular disease, and sickle cell anaemia. This review aims to describe conjunctival microcirculation, its characteristics, and techniques for its measurement, as well as the association between conjunctival microcirculation and microvascular abnormalities in disease states.

In-vivo corneal confocal microscopy: Imaging analysis, biological insights and future directions

In-vivo corneal confocal microscopy is a powerful imaging technique which provides clinicians and researcher with the capabilities to observe microstructures at the ocular surfaces in significant detail. In this Mini Review, the optics and image analysis methods with the use of corneal confocal microscopy are discussed. While novel insights of neuroanatomy and biology of the eyes, particularly the ocular surface, have been provided by corneal confocal microscopy, some debatable elements observed using this technique remain and these are explored in this Mini Review. Potential improvements in imaging methodology and instrumentation are also suggested.

In Vivo Corneal Confocal Microscopy in Multiple Sclerosis: Can it Differentiate Disease Relapse in Multiple Sclerosis?

PURPOSE

This study aims to investigate the role of in vivo corneal confocal microscopy (IVCCM) in the detection of corneal inflammatory activity and subbasal nerve alterations in patients with multiple sclerosis (MS) and to further determine whether IVCCM can be used to detect (acute) disease relapse.

DESIGN

Prospective cross-sectional study, with a subgroup follow-up.

METHODS

This single-center study included 58 patients with MS (MS-Relapse group [n = 27] and MS-Remission group [n = 31]), and 30 age- and sex-matched healthy control subjects. Patients with a history of optic neuritis or trigeminal symptoms were excluded. Corneal nerve fiber density (CNFD), corneal nerve branch density (CNBD), corneal nerve fiber length (CNFL), and dendritic cell (DC) density were evaluated in all patients with MS and control subjects by IVCCM. Patients in the MS-Relapse group who were in remission for ≥6 months after the MS incident underwent a repeat IVCCM.

RESULTS

No statistical difference was observed between the MS-Relapse and MS-Remission groups regarding age, sex, MS duration, and the number of relapses (P > .05). Compared with healthy control subjects, all subbasal nerve parameters were significantly lower (CNFD: P < .001, CNFL: P < .001, CNBD: P < .001), and the DC density was significantly higher (P = .023) in patients with MS. However, no significant difference was observed between MS-Relapse and MS-Remission groups in terms of CNFD (mean [SE] difference -2.05 [1.69] fibers/mm² [95% confidence interval {CI} -1.32 to 5.43]; P < .227), CNFL (mean [SE] difference -1.10 [0.83] mm/mm² [95% CI -0.56 to 2.75]; P < .190), CNBD (mean [SE] difference -3.91 [2.48] branches/mm² [95% CI -1.05 to 8.87]; P < .120), and DC density (median [IQR], 59.38 [43.75-85.0] vs 75.0 [31.25-128.75]; P = .596). The repeat IVCCM in relapse patients (n = 16 [59.3%]) showed a significant increase in CNFD (P = .036) and CNBD (P = .018), but no change was observed in CNFL (P = .075) and DC density (P = .469).

CONCLUSION

Although increased inflammation and neurodegeneration can be demonstrated in patients with MS compared with healthy control subjects, a single time point evaluation of IVCCM does not seem to be sufficient to confirm the occurrence of relapse in patients with MS. However, IVCCM holds promise for demonstrating early neuroregeneration in patients with MS.

Corneal Sub-Basal Nerve Plexus in Non-Diabetic Small Fiber Polyneuropathies and the Diagnostic Role of In Vivo Corneal Confocal Microscopy

In vivo corneal confocal microscopy (IVCM) allows the immediate analysis of the corneal nerve quantity and morphology. This method became, an indispensable tool for the tropism examination, as it evaluates the small fiber plexus in the cornea. The IVCM provides us with direct information on the health of the sub-basal nerve plexus and indirectly on the peripheral nerve status. It is an important tool used to investigate peripheral polyneuropathies. Small-fiber neuropathy (SFN) is a group of neurological disorders characterized by neuropathic pain symptoms and autonomic complaints due to the selective involvement of thinly myelinated Aδ-fibers and unmyelinated C-fibers. Accurate diagnosis of SFN is important as it provides a basis for etiological work-up and treatment decisions. The diagnosis of SFN is sometimes challenging as the clinical picture can be difficult to interpret and standard electromyography is normal. In cases of suspected SFN, measurement of intraepidermal nerve fiber density through a skin biopsy and/or analysis of quantitative sensory testing can enable diagnosis. The purpose of the present review is to summarize the current knowledge about corneal nerves in different SFN. Specifically, we explore the correlation between nerve density and morphology and type of SFN, disease duration, and follow-up. We will discuss the relationship between cataracts and refractive surgery and iatrogenic dry eye disease. Furthermore, these new paradigms in SFN present an opportunity for neurologists and clinical specialists in the diagnosis and monitoring the peripheral small fiber polyneuropathies.

Impact of Chronic Kidney Disease on Corneal Neuroimmune Features in Type 2 Diabetes

Aim: To determine the impact of chronic kidney disease on corneal nerve measures and dendritic cell counts in type 2 diabetes. Methods: In vivo corneal confocal microscopy images were used to estimate corneal nerve parameters and compared in people with type 2 diabetes with chronic kidney disease (T2DM-CKD) (n = 29) and those with type 2 diabetes without chronic kidney disease (T2DM-no CKD) (n = 29), along with 30 healthy controls. Corneal dendritic cell densities were compared between people with T2DM-CKD and those with T2DM-no CKD. The groups were matched for neuropathy status. Results: There was a significant difference in corneal nerve fiber density (p < 0.01) and corneal nerve fiber length (p = 0.04) between T2DM-CKD and T2DM-no CKD groups. The two diabetes groups had reduced corneal nerve parameters compared to healthy controls (all parameters: p < 0.01). Immature central dendritic cell density was significantly higher in the T2DM-CKD group compared to the T2DM-no CKD group ((7.0 (3.8−12.8) and 3.5 (1.4−13.4) cells/mm2, respectively, p < 0.05). Likewise, central mature dendritic cell density was significantly higher in the T2DM-CKD group compared to the T2DM-no CKD group (0.8 (0.4−2.2) and 0.4 (0.6−1.1) cells/mm2, respectively, p = 0.02). Additionally, total central dendritic cell density was increased in the T2DM-CKD group compared to T2DM-no CKD group (10.4 (4.3−16.1) and 3.9 (2.1−21.0) cells/mm2, respectively, p = 0.03). Conclusion: The study showed that central corneal dendritic cell density is increased in T2DM-CKD compared to T2DM-no CKD, with groups matched for peripheral neuropathy severity. This is accompanied by a loss of central corneal nerve fibers. The findings raise the possibility of additional local factors exacerbating central corneal nerve injury in people with diabetic chronic kidney disease.

Assessing the corneal sub-basal nerve plexus by in vivo confocal microscopy in patients with blepharoptosis

BACKGROUND

To assess in vivo confocal microscopy features of corneal sub-basal nerve plexus in patients with congenital or aponeurogenic blepharoptosis using a fully automated software (ACCMetrics).

PATIENTS AND METHODS

This prospective study included 33 patients with blepharoptosis and 17 normal controls. The corneal sub-basal nerve plexus was assessed using in vivo confocal microscopy, and the ocular surface status was evaluated by tear break-up times.

RESULTS

The mean age of 33 patients with blepharoptosis and 17 normal controls were 38.77 ± 22.81 years and 48.35 ± 17.15 years, respectively. The mean duration of blepharoptosis was 16.42 ± 15.60 years. In 13 patients with unilateral blepharoptosis, there was no significant difference between affected eyes and contralateral eyes (all ps > .05), except for wider corneal nerve fibre width (CNFW) in affected eyes (0.024 ± 0.001 versus 0.023 ± 0.001 mm/mm², p = .021). In 20 patients with bilateral blepharoptosis, there was no significant difference between the eyes. No significant difference was detected between 19 cases with congenital blepharoptosis and 14 cases with aponeurogenic blepharoptosis. When compared with normal controls, eyes with both unilateral and bilateral blepharoptosis had significantly wider CNFW. But from the point of aetiology, only eyes with congenital blepharoptosis presented with wider CNFW (p = .001), rather than the eyes with aponeurogenic blepharoptosis (p = .093). Besides, four young patients with congenital blepharoptosis revealed very sparse sub-basal nerve plexus.

CONCLUSIONS

These data suggested that corneal confocal microscopy demonstrated no significant changes in patients with blepharoptosis as compared with normal controls, except for relatively wider CNFW in congenital affected eyes. However, in some children and young adults with congenital blepharoptosis, the density of corneal sub-basal nerve plexus was evidently decreased, which needs to be cautioned when ones with congenital blepharoptosis want to take corneal surgeries or wear contact lens.Key messagesWhen compared with normal controls, no significant effect was found in the influence of blepharoptosis on the most of corneal nerve parameters, except for corneal nerve fibre width (CNFW) in the group of congenital blepharoptosis.The age of onset of blepharoptosis may influence corneal nerve fibres, so timely surgical treatment of congenital blepharoptosis is not only conducive to the development of normal vision, but also beneficial to the reduction of corneal nerve lesions to some extent.We noted that some young blepharoptosis patients revealed sparse corneal nerve, which should be taken precaution when ones with congenital blepharoptosis who want to take corneal surgeries or wear contact lens.

Impact of Peripheral and Corneal Neuropathy on Markers of Ocular Surface Discomfort in Diabetic Chronic Kidney Disease

SIGNIFICANCE

There is a reduction in corneal nerve fiber density and length in type 2 diabetes mellitus with chronic kidney disease compared with type 2 diabetes mellitus alone; however, this difference does not result in worse ocular surface discomfort or dry eye disease.

PURPOSE

This study aimed to determine the clinical impact of corneal nerve loss on ocular surface discomfort and markers of ocular surface homeostasis in people with type 2 diabetes mellitus without chronic kidney disease (T2DM-no CKD) and those with type 2 diabetes mellitus with concurrent chronic kidney disease (T2DM-CKD).

METHODS

Participants were classified based on estimated glomerular filtration rates into two groups: T2DM-CKD (n = 27) and T2DM-no CKD (n = 28).

RESULTS

There was a significant difference between the T2DM-CKD and T2DM-no CKD groups in corneal nerve fiber density (14.9 ± 8.6 and 21.1 ± 7.1 no./mm 2 , respectively; P = .005) and corneal nerve fiber length (10.0 ± 4.6 and 12.3 ± 3.7 mm/mm 2 , respectively; P = .04). Fluorescein tear breakup time was significantly reduced in T2DM-CKD compared with T2DM-no CKD (8.1 ± 4.4 and 10.7 ± 3.8 seconds, respectively; P = .01), whereas ocular surface staining was not significantly different (3.5 ± 1.7 and 2.7 ± 2.3 scores, respectively; P = .12). In terms of ocular surface discomfort, there were no significant differences in the ocular discomfort score scores (12.5 ± 11.1 and 13.6 ± 12.1, respectively; P = .81) and Ocular Pain Assessment Survey scores (3.3 ± 5.4 and 4.3 ± 6.1, respectively; P = .37) between the T2DM-CKD and T2DM-no CKD.

CONCLUSIONS

The current study demonstrated that corneal nerve loss is greater in T2DM-CKD than in T2DM-no CKD. However, these changes do not impact ocular surface discomfort or markers of ocular surface homeostasis.

Segmentation and Evaluation of Corneal Nerves and Dendritic Cells From In Vivo Confocal Microscopy Images Using Deep Learning

Purpose

Segmentation and evaluation of in vivo confocal microscopy (IVCM) images requires manual intervention, which is time consuming, laborious, and non-reproducible. The aim of this research was to develop and validate deep learning–based methods that could automatically segment and evaluate corneal nerve fibers (CNFs) and dendritic cells (DCs) in IVCM images, thereby reducing processing time to analyze larger volumes of clinical images.

Methods

CNF and DC segmentation models were developed based on U-Net and Mask R-CNN architectures, respectively; 10-fold cross-validation was used to evaluate both models. The CNF model was trained and tested using 1097 and 122 images, and the DC model was trained and tested using 679 and 75 images, respectively, at each fold. The CNF morphology, number of nerves, number of branching points, nerve length, and tortuosity were analyzed; for DCs, number, size, and immature–mature cells were analyzed. Python-based software was written for model training, testing, and automatic morphometric parameters evaluation.

Results

The CNF model achieved on average 86.1% sensitivity and 90.1% specificity, and the DC model achieved on average 89.37% precision, 94.43% recall, and 91.83% F₁ score. The interclass correlation coefficient (ICC) between manual annotation and automatic segmentation were 0.85, 0.87, 0.95, and 0.88 for CNF number, length, branching points, and tortuosity, respectively, and the ICC for DC number and size were 0.95 and 0.92, respectively.

Conclusions

Our proposed methods demonstrated reliable consistency between manual annotation and automatic segmentation of CNF and DC with rapid speed. The results showed that these approaches have the potential to be implemented into clinical practice in IVCM images.

Translational Relevance

The deep learning–based automatic segmentation and quantification algorithm significantly increases the efficiency of evaluating IVCM images, thereby supporting and potentially improving the diagnosis and treatment of ocular surface disease associated with corneal nerves and dendritic cells.

Corneal in vivo Confocal Microscopy for Assessment of Non-Neurological Autoimmune Diseases: A Meta-Analysis

Purpose

This study aimed to evaluate the features of corneal nerve with in vivo confocal microscopy (IVCM) among patients with non-neurological autoimmune (NNAI) diseases.

Methods

We systematically searched PubMed, Web of Science, and Cochrane Central Register of Controlled Trials for studies published until May 2021. The weighted mean differences (WMDs) of corneal nerve fiber length (CNFL), corneal nerve fiber density (CNFD), corneal nerve branch density (CNBD), tortuosity, reflectivity, and beadings per 100 μm with a 95% CI between NNAI and control group were analyzed using a random-effects model.

Results

The results showed 37 studies involving collective totals of 1,423 patients and 1,059 healthy controls were ultimately included in this meta-analysis. The pooled results manifested significantly decreased CNFL (WMD: −3.94, 95% CI: −4.77–−3.12), CNFD (WMD: −6.62, 95% CI: −8.4–−4.85), and CNBD (WMD: −9.89, 95% CI: −14–−5.79) in NNAI patients. In addition, the NNAI group showed more tortuous corneal nerve (WMD: 1.19, 95% CI:0.57–1.81). The comparison between NNAI patients and healthy controls in beadings per 100 μm corneal nerve length was inconsistent. No significant difference was found in the corneal nerve fiber reflectivity between NNAI and the control group (WMD: −0.21, 95% CI: −0.65–0.24, P = 0.361).

Conclusions

The parameters and morphology of corneal nerves observed by IVCM proved to be different in NNAI patients from healthy controls, suggesting that IVCM may be a non-invasive technique for identification and surveillance of NNAI diseases.

Generative Adversarial Network Based Automatic Segmentation of Corneal Subbasal Nerves on In Vivo Confocal Microscopy Images

Purpose

In 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.

Methods

We 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.

Results

The 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.

Conclusions

This 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 Relevance

Generative 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.

Neurotrophic keratitis after penetrating keratoplasty for lattice dystrophy

Purpose

To report clinical outcomes of a patient with unilateral neurotrophic keratitis following penetrating keratoplasty for lattice dystrophy treated with topical recombinant human nerve growth factor.

Observations

A 75-year-old male with lattice dystrophy and history of herpes simplex keratitis, presented with recurrent neurotrophic ulceration in the right eye two years following penetrating keratoplasty. The patient was successfully treated with topical recombinant human nerve growth factor.

Conclusion

Neurotrophic keratitis is a rare chronic disorder that affects quality of life due to the risk of vision loss. Topical recombinant human nerve growth factor is a novel and effective treatment option that may help improve optical quality and patient's satisfaction as shown in this case of recurrent neurotrophic keratitis.

Impact of corneal parameters, refractive error and age on density and morphology of the subbasal nerve plexus fibers in healthy adults

The purpose of this study was to analyze corneal sub-basal nerve plexus (SBNP) density and morphology and their relationships with corneal parameters and refractive status. In this single center study, in vivo confocal microscopy (IVCM) was performed in 76 eyes of 38 healthy subjects aged 19–87 (mean age 34.987 ± 1.148). Nerve fiber analysis was performed using Confoscan 4 microscope with semi-automated software (Nidek Technologies, Italy) The nerve fiber length (NFL) µm/mm², nerve fiber density (NFD) no./mm², tortuosity coefficient (TC), and nerve beadings density (NBD) no./mm were considered. Relationship between SBNP parameters and corneal curvature, thickness, diameter, and refraction were analyzed. Additionally, the association with gender, laterality and age were determined. NFL was inversely correlated with age (r = − 0.528, p < 0.001), myopic refractive error (spherical value) (r = − 0.423, p < 0.001), and cylindrical power (r = − 0.340, p = 0.003). NFD was inversely correlated with age (r = − 0.420, p < 0.001) and myopic refractive error (r = − 0.341, p = 0.003). NBD showed a low inverse correlation with cylindrical power (r = − 0.287, p = 0.012) and a slight positive correlation with K (r = 0.230, p = 0.047). TC showed a significant negative correlation between age (r = − 0.500, p < 0.001) and myopic refractive error (r = − 0.351, p = 0.002). Additionally, there were strong positive correlations between NFL and NFD (r = 0.523, p < 0.001), NFL and TI (r = 0.603, p < 0.001), and NFD and TC (r = 0.758, p < 0.001). Multiple regression analysis revealed age to be the most significant factor affecting SBNP density (B = − 0.467, p = 0.013) and length (B = − 61.446, p < 0.001); myopic refractive error reduced both SBNP density (B = − 2.119, p = 0.011) and length (B = − 158.433, p = 0.016), while gender and laterality had no significant effects (p > 0.005). SBNP fiber length decreases with age, myopic refractive error and cylindrical power. SBNP fiber density reduces with age and myopic refractive error. Corneal nerve parameters are not influenced by gender or laterality.

Diabetic Corneal Neuropathy

Diabetic keratopathy (DK) is a common, but underdiagnosed, ocular complication of diabetes mellitus (DM) that has a significant economic burden. It is characterised by progressive damage of corneal nerves, due to DM-induced chronic hyperglycaemia and its associated metabolic changes. With advances in corneal nerve imaging and quantitative analytic tools, studies have shown that the severity of diabetic corneal neuropathy correlates with the status of diabetic peripheral neuropathy. The corneal nerve plexus is, therefore, considered as an important surrogate marker of diabetic peripheral neuropathy and helps in the evaluation of interventional efficacy in the management of DM. The clinical manifestations of DK depend on the disease severity and vary from decreased corneal sensitivity to sight-threatening corneal infections and neurotrophic ulcers. The severity of diabetic corneal neuropathy and resultant DK determines its management plan, and a step-wise approach is generally suggested. Future work would focus on the exploration of biomarkers for diabetic corneal neuropathy, the development of new treatment for corneal nerve protection, and the improvement in the clinical assessment, as well as current imaging technique and analysis, to help clinicians detect diabetic corneal neuropathy earlier and monitor the sub-clinical progression more reliably.

CS2-Net: Deep learning segmentation of curvilinear structures in medical imaging

Automated detection of curvilinear structures, e.g., blood vessels or nerve fibres, from medical and biomedical images is a crucial early step in automatic image interpretation associated to the management of many diseases. Precise measurement of the morphological changes of these curvilinear organ structures informs clinicians for understanding the mechanism, diagnosis, and treatment of e.g. cardiovascular, kidney, eye, lung, and neurological conditions. In this work, we propose a generic and unified convolution neural network for the segmentation of curvilinear structures and illustrate in several 2D/3D medical imaging modalities. We introduce a new curvilinear structure segmentation network (CS²-Net), which includes a self-attention mechanism in the encoder and decoder to learn rich hierarchical representations of curvilinear structures. Two types of attention modules - spatial attention and channel attention - are utilized to enhance the inter-class discrimination and intra-class responsiveness, to further integrate local features with their global dependencies and normalization, adaptively. Furthermore, to facilitate the segmentation of curvilinear structures in medical images, we employ a 1×3 and a 3×1 convolutional kernel to capture boundary features. Besides, we extend the 2D attention mechanism to 3D to enhance the network's ability to aggregate depth information across different layers/slices. The proposed curvilinear structure segmentation network is thoroughly validated using both 2D and 3D images across six different imaging modalities. Experimental results across nine datasets show the proposed method generally outperforms other state-of-the-art algorithms in various metrics.

Corneal nerves in diabetes-The role of the in vivo corneal confocal microscopy of the subbasal nerve plexus in the assessment of peripheral small fiber neuropathy

The cornea's intense innervation is responsible for corneal trophism and ocular surface hemostasis maintenance. Corneal diabetic neuropathy affects subbasal nerve plexus, with progressive alteration of nerves' morphology and density. The quantitative analysis of nerve fibers can be performed with in vivo corneal confocal microscopy considering the main parameters such as corneal nerve fibers length, corneal nerve fibers density, corneal nerve branching density, tortuosity coefficient, and beadings frequency. As the nerve examination permits the detection of early changes occurring in diabetes, the invivo corneal confocal microscopy becomes, over time, an important tool for diabetic polyneuropathy assessment and follow-up. In this review, we summarize the actual evidence about corneal nerve changes in diabetes and the relationship between the grade of alterations and the duration and severity of the disease. We aim at understanding how diabetes impacts corneal nerves and how it correlates with sensorimotor peripheral polyneuropathy and retinal complications. We also attempt to analyze the safety of the most common surgical procedures such as cataract and refractive surgery in diabetic patients and to highlight the specific risk factors. We believe that information about the corneal nerve fibers' condition obtained from the in vivo subbasal nerve plexus investigation may be crucial in monitoring peripheral small fiber polyneuropathy and that it will help with decision-making in ophthalmic surgery in diabetic patients.

Automated Tortuosity Analysis of Nerve Fibers in Corneal Confocal Microscopy

Precise characterization and analysis of corneal nerve fiber tortuosity are of great importance in facilitating examination and diagnosis of many eye-related diseases. In this paper we propose a fully automated method for image-level tortuosity estimation, comprising image enhancement, exponential curvature estimation, and tortuosity level classification. The image enhancement component is based on an extended Retinex model, which not only corrects imbalanced illumination and improves image contrast in an image, but also models noise explicitly to aid removal of imaging noise. Afterwards, we take advantage of exponential curvature estimation in the 3D space of positions and orientations to directly measure curvature based on the enhanced images, rather than relying on the explicit segmentation and skeletonization steps in a conventional pipeline usually with accumulated pre-processing errors. The proposed method has been applied over two corneal nerve microscopy datasets for the estimation of a tortuosity level for each image. The experimental results show that it performs better than several selected state-of-the-art methods. Furthermore, we have performed manual gradings at tortuosity level of four hundred and three corneal nerve microscopic images, and this dataset has been released for public access to facilitate other researchers in the community in carrying out further research on the same and related topics.

Corneal nerve tortuosity grading via ordered weighted averaging-based feature extraction

PURPOSE

Tortuosity of corneal nerve fibers acquired by in vivo Confocal Microscopy (IVCM) are closely correlated to numerous diseases. While tortuosity assessment has conventionally been conducted through labor-intensive manual evaluation, this warrants an automated and objective tortuosity assessment of curvilinear structures. This paper proposes a method that extracts the image-level features for corneal nerve tortuosity grading.

METHODS

For an IVCM image, all corneal nerve fibers are first segmented and then, their tortuosity are calculated by morphological measures. The ordered weighted averaging (OWA) approach, and the k-Nearest-Neighbor guided dependent ordered weighted averaging (kNNDOWA) approach are proposed to aggregate the tortuosity values and form a set of extracted features. This is followed by running the Wrapper method, a supervised feature selection, with an aim to identify the most informative attributes for tortuosity grading.

RESULTS

Validated on a public and an in-house benchmark data sets, experimental results demonstrate superiority of the proposed method over the conventional averaging and length-weighted averaging methods with performance gain in accuracy (15.44% and 14.34%, respectively).

CONCLUSIONS

The simultaneous use of multiple aggregation operators could extract the image-level features that lead to more stable and robust results compared with that using average and length-weighted average. The OWA method could facilitate the explanation of derived aggregation behavior through stress functions. The kNNDOWA method could mitigate the effects of outliers in the image-level feature extraction.

An artificial intelligence-based deep learning algorithm for the diagnosis of diabetic neuropathy using corneal confocal microscopy: a development and validation study

AIMS/HYPOTHESIS

Corneal confocal microscopy is a rapid non-invasive ophthalmic imaging technique that identifies peripheral and central neurodegenerative disease. Quantification of corneal sub-basal nerve plexus morphology, however, requires either time-consuming manual annotation or a less-sensitive automated image analysis approach. We aimed to develop and validate an artificial intelligence-based, deep learning algorithm for the quantification of nerve fibre properties relevant to the diagnosis of diabetic neuropathy and to compare it with a validated automated analysis program, ACCMetrics.

METHODS

Our deep learning algorithm, which employs a convolutional neural network with data augmentation, was developed for the automated quantification of the corneal sub-basal nerve plexus for the diagnosis of diabetic neuropathy. The algorithm was trained using a high-end graphics processor unit on 1698 corneal confocal microscopy images; for external validation, it was further tested on 2137 images. The algorithm was developed to identify total nerve fibre length, branch points, tail points, number and length of nerve segments, and fractal numbers. Sensitivity analyses were undertaken to determine the AUC for ACCMetrics and our algorithm for the diagnosis of diabetic neuropathy.

RESULTS

The intraclass correlation coefficients for our algorithm were superior to those for ACCMetrics for total corneal nerve fibre length (0.933 vs 0.825), mean length per segment (0.656 vs 0.325), number of branch points (0.891 vs 0.570), number of tail points (0.623 vs 0.257), number of nerve segments (0.878 vs 0.504) and fractals (0.927 vs 0.758). In addition, our proposed algorithm achieved an AUC of 0.83, specificity of 0.87 and sensitivity of 0.68 for the classification of participants without (n = 90) and with (n = 132) neuropathy (defined by the Toronto criteria).

CONCLUSIONS/INTERPRETATION

These results demonstrated that our deep learning algorithm provides rapid and excellent localisation performance for the quantification of corneal nerve biomarkers. This model has potential for adoption into clinical screening programmes for diabetic neuropathy.

DATA AVAILABILITY

The publicly shared cornea nerve dataset (dataset 1) is available at http://bioimlab.dei.unipd.it/Corneal%20Nerve%20Tortuosity%20Data%20Set.htm and http://bioimlab.dei.unipd.it/Corneal%20Nerve%20Data%20Set.htm.

Corneal nerve fiber loss in diabetes with chronic kidney disease

AIMS

Patients with chronic kidney disease (CKD) in type 2 diabetes typically manifest with severe peripheral neuropathy. Corneal confocal microscopy is a novel technique that may serve as a marker of nerve injury in peripheral neuropathy. This study examines the changes that occur in corneal nerve morphology as a result of peripheral neuropathy due to renal dysfunction in people with type 2 diabetes.

METHODS

Sixty-two participants (mean age, 62 ± 12 years) with type 2 diabetes and 25 age-matched healthy controls underwent a comprehensive assessment of neuropathy using the total neuropathy score (TNS). The corneal sub-basal nerve plexus was imaged using corneal confocal microscopy. Corneal nerve fiber length, fiber density, branch density, total branch density, nerve fractal dimension, inferior whorl length and inferior whorl nerve fractal dimension were quantified. Based on the eGFR, participants were classified into those with diabetic CKD (eGFR < 60; n = 22) and those without CKD (eGFR ≥ 60; n = 40).

RESULTS

Participants with diabetic CKD had significantly lower corneal nerve fiber density (P = 0.037), length (P = 0.036) and nerve fractal dimension (P = 0.036) compared to those without CKD. Multiple linear regression analysis revealed that reduced corneal nerve fiber density (ß coefficient = 0.098, P = 0.017), length (ß coefficient = 0.006, P = 0.008) and nerve fractal dimension (ß coefficient = 0.001, P = 0.007) was associated with low eGFR levels when adjusted for age, duration of diabetes and severity of neuropathy.

CONCLUSION

Corneal confocal microscopy detects corneal nerve loss in patients with diabetic CKD and reduction in corneal nerve parameters is associated with the decline of kidney function.

Early Alterations of Corneal Subbasal Plexus in Uncomplicated Type 1 Diabetes Patients

Purpose

The purpose of our study is to describe the in vivo corneal confocal microscopy characteristics of subbasal nerve plexus in a highly selected population of patients affected by type 1 diabetes mellitus (T1DM) without any microvascular diabetes complications.

Methods

We included 19 T1DM patients without diabetic peripheral neuropathy, diabetic autonomic neuropathy, diabetic retinopathy, and microalbuminuria. All patients underwent in vivo corneal confocal microscopy and blood analysis to determine subbasal nerve plexus parameters and their correlation with clinical data. We compared the results with 19 healthy controls.

Results

The T1DM group showed a significant decrease of the nerve fiber length (P=0.032), the nerve fiber length density (P=0.034), the number of fibers (P=0.005), and the number of branchings (P=0.028), compared to healthy subjects. The nerve fiber length, nerve fiber length density, and number of fibers were directly related to the age at onset of diabetes and inversely to the duration of DM. BMI (body mass index) was highly related to the nerve fiber length (r = −0.6, P=0.007), to the nerve fiber length density (r = −0.6, P=0.007), and to the number of fibers (r = −0.587, P=0.008). No significant correlations were found between the corneal parameters and HbA1c.

Conclusions

Early subclinical fiber corneal variation could be easily detected using in vivo corneal confocal microscopy, even in type 1 diabetes without any microvascular diabetes complications, including diabetic peripheral neuropathy, diabetic autonomic neuropathy, diabetic retinopathy, and microalbuminuria.

3D confocal laser-scanning microscopy for large-area imaging of the corneal subbasal nerve plexus

The capability of corneal confocal microscopy (CCM) to acquire high-resolution in vivo images of the densely innervated human cornea has gained considerable interest in using this non-invasive technique as an objective diagnostic tool for staging peripheral neuropathies. Morphological alterations of the corneal subbasal nerve plexus (SNP) assessed by CCM have been shown to correlate well with the progression of neuropathic diseases and even predict future-incident neuropathy. Since the field of view of single CCM images is insufficient for reliable characterisation of nerve morphology, several image mosaicking techniques have been developed to facilitate the assessment of the SNP in large-area visualisations. Due to the limited depth of field of confocal microscopy, these approaches are highly sensitive to small deviations of the focus plane from the SNP layer. Our contribution proposes a new automated solution, combining guided eye movements for rapid expansion of the acquired SNP area and axial focus plane oscillations to guarantee complete imaging of the SNP. We present results of a feasibility study using the proposed setup to evaluate different oscillation settings. By comparing different image selection approaches, we show that automatic tissue classification algorithms are essential to create high-quality mosaic images from the acquired 3D datasets.