Comparing the results of manual and automated quantitative corneal neuroanalysing modules for beginners

Diabetic corneal neuropathy and its relation to the severity of retinopathy in patients with type 2 diabetes mellitus: an in vivo confocal microscopy study

PURPOSE

To investigate corneal neuropathy and corneal nerve alterations in type 2 diabetes mellitus (DM) patients with different diabetic retinopathy (DR) status.

METHODS

A total of 87 eyes of 87 patients with DM and 28 eyes of 28 healthy control subjects were included in the study. DM patients were further classified into 3 groups: patients without DR (NDR), patients with non-proliferative DR (NPDR), and patients with proliferative DR (PDR). PDR patients were classified into 2 groups regarding having undergone retinal argon laser photocoagulation treatment (ALP). Ocular surface disease index score (OSDI), average tear break-up time (A-BUT), corneal sensitivity and cornea nerve fiber length (CNFL), cornea nerve fiber density (CNFD), and cornea nerve branch density (CNBD) of the cornea subbasal nerve plexus (SBNP) were measured using in vivo confocal microscopy (IVCM).

RESULTS

OSDI scores increased and A-BUT decreased in DM patients compared to the control group, but no significant difference was found between DM patient groups. Corneal sensitivity decreased in DM patients who developed DR, compared to both the controls and the NDR group. CNFD and CNFL decreased in NPDR and PDR patients compared to controls. CNFD and CNBD decreased in patients who had developed PDR, compared to all three groups. All IVCM parameters decreased with DR progression.

CONCLUSION

IVCM can detect early structural corneal nerve changes in diabetic patients. The presence of DM affects ocular surface parameters, especially in long-term DM patients. Corneal sensitivity loss is increased with the presence of DR. All IVCM parameters decrease with DR development and its progression.

Corneal structural alterations in autism spectrum disorder: An in vivo confocal microscopy study

Individuals with autism spectrum disorder (ASD) often exhibit joint hypermobility and connective tissue disorders. However, it remains unclear if ASD individuals also have structural alterations in the connective tissue of the cornea. This study aims to determine whether the Kobayashi structure (K-structure) characteristics differ between adults with ASD and typically developing controls (TDC) and explore the clinical correlates of the K-structure abnormality. We recruited 30 ASD adults and 35 TDC. Corneal structures, particularly the K-structure in the Bowman's layer, of the participants were examined using in vivo confocal microscopy (IVCM), and a K-grading ranging from 1 to 4 was given to each eye based on the level of morphological mosaicism. The ASD participants' eyes received a significantly higher single-eye K-grading than that of the TDC eyes (p < 0.001), and the medians [25th, 75th percentile] of bilateral-eye summed K-grading were 8 [7, 8] and 5 [4, 6] in ASD and TDC, respectively (p < 0.001). A significantly higher K-grading in the ASD participants' eyes was still observed after adjusting for the within-subject inter-eye correlation (p < 0.001). Youden Index showed the optimal cutoffs to differentiate ASD from TDC by bilateral-eye summed K-grading and single-eye K-grading was >6 and >3, respectively. Additionally, a higher K-grading was associated with fewer visual sensation seeking in ASD (Spearman's correlation coefficient ρ = -0.518, p = 0.008) and low visual registration (i.e., higher sensory threshold) in TDC (ρ = 0.446, p = 0.023). This study provided novel evidence of corneal structural alterations in ASD by IVCM. Our findings may not only support the prior hypothesis of the association between ASD and connective tissue abnormalities but also shed light on the relationship between connective tissue disorder and neurodevelopmental disorders.

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.

Semiautomated and Automated Quantitative Analysis of Corneal Sub-Basal Nerves in Patients With DED With Ocular Pain Using IVCM

Purpose

Investigate the correlation and agreement between the results of semiautomated and fully automated quantitative analysis of the corneal sub-basal nerve plexus (SNP) in patients with dry eye disease (DED) with ocular pain using in vivo confocal microscopy (IVCM).

Method

A total of 50 voluntary participants were enrolled in this study, i.e., 25 DED patients with ocular pain and 25 healthy controls. Each patient underwent an evaluation of ocular symptoms that utilized: the Ocular Surface Disease Index (OSDI), the Ocular Pain Assessment Survey (OPAS), the tear film breakup time (TBUT) test, the Schirmer test, corneal staining, and IVCM. Five SNP images of the cornea of each eye were selected and analyzed using a semiautomated analysis software (NeuronJ) and a fully automated method (ACCMetrics) to quantify corneal nerve fiber density (CNFD), corneal nerve branch density (CNBD), and corneal nerve fiber length (CNFL).

Results

The intraclass correlation coefficient (ICC) of the CNFD (0.460 [0.382–0.532], p < 0.001), CNBD (0.608 [0.545–0.665], p < 0.001), and CNFL (0.851 [0.822–0.875], p < 0.001) represents the repeatability and consistency of measurements by the NeuronJ and ACCMetrics software. The CNFL values (r = 0.881, p < 0.001) obtained using the two methods have extremely high correlation, and similarly, the CNFD values (r = 0.669, p < 0.001) and CNBD values (r = 0.703, p < 0.001) are highly correlated. The CNFL had the biggest area under the curve (AUC; 0.747 [0.700–0.793], p < 0.001) when using ACCMetrics. In DED patients with ocular pain, the mean CNFD values for semiautomated and fully automated quantization were 23.5 ± 8.1 and 23.8 ± 8.6 n/mm²; the mean CNBD values were 46.0 ± 21.3, 35.7 ± 23.3 n/mm²; and the mean CNFL values were 19.3 ± 4.3 and 15.2 ± 3.8 mm/mm², which were significantly lower than healthy subjects (p < 0.001).

Conclusion

There is a significant correlation between the measurements obtained via ACCMetrics and NeuronJ, especially for CNFL, which can be considered as the primary indicator in the diagnosis of DED with ocular pain. The SNP of the disease was significantly lower than that of healthy subjects.