OBJECTIVE: Quantitative assessment of small fiber damage is key to the early diagnosis and assessment of progression or regression of diabetic sensorimotor polyneuropathy (DSPN). Intraepidermal nerve fiber density (IENFD) is the current gold standard, but corneal confocal microscopy (CCM), an in vivo ophthalmic imaging modality, has the potential to be a noninvasive and objective image biomarker for identifying small fiber damage. The purpose of this study was to determine the diagnostic performance of CCM and IENFD by using the current guidelines as the reference standard. RESEARCH DESIGN AND METHODS: Eighty-nine subjects (26 control subjects and 63 patients with type 1 diabetes), with and without DSPN, underwent a detailed assessment of neuropathy, including CCM and skin biopsy. RESULTS: Manual and automated corneal nerve fiber density (CNFD) (P < 0.0001), branch density (CNBD) (P < 0.0001) and length (CNFL) (P < 0.0001), and IENFD (P < 0.001) were significantly reduced in patients with diabetes with DSPN compared with control subjects. The area under the receiver operating characteristic curve for identifying DSPN was 0.82 for manual CNFD, 0.80 for automated CNFD, and 0.66 for IENFD, which did not differ significantly (P = 0.14). CONCLUSIONS: This study shows comparable diagnostic efficiency between CCM and IENFD, providing further support for the clinical utility of CCM as a surrogate end point for DSPN.

Purpose.: To assess the diagnostic validity of a fully automated image analysis algorithm of in vivo confocal microscopy images in quantifying corneal subbasal nerves to diagnose diabetic neuropathy. Methods.: One hundred eighty-six patients with type 1 and type 2 diabetes mellitus (T1/T2DM) and 55 age-matched controls underwent assessment of neuropathy and bilateral in vivo corneal confocal microscopy (IVCCM). Corneal nerve fiber density (CNFD), branch density (CNBD), and length (CNFL) were quantified with expert, manual, and fully-automated analysis. The areas under the curve (AUC), odds ratios (OR), and optimal thresholds to rule out neuropathy were estimated for both analysis methods. Results.: Neuropathy was detected in 53% of patients with diabetes. A significant reduction in manual and automated CNBD (P < 0.001) and CNFD (P < 0.0001), and CNFL (P < 0.0001) occurred with increasing neuropathic severity. Manual and automated analysis methods were highly correlated for CNFD (r = 0.9, P < 0.0001), CNFL (r = 0.89, P < 0.0001), and CNBD (r = 0.75, P < 0.0001). Manual CNFD and automated CNFL were associated with the highest AUC, sensitivity/specificity and OR to rule out neuropathy. Conclusions.: Diabetic peripheral neuropathy is associated with significant corneal nerve loss detected with IVCCM. Fully automated corneal nerve quantification provides an objective and reproducible means to detect human diabetic neuropathy.

Purpose: To establish intraobserver and interobserver repeatability, agreement, and symmetry of corneal nerve fiber (NF) morphology in healthy subjects using in vivo corneal confocal microscopy. Methods: Nineteen subjects underwent in vivo corneal confocal microscopy (Heidelberg Retinal Tomograph III Rostock Cornea Module) at baseline and 7 days apart. Bland–Altman plots were generated to assess agreement, and the intraclass correlation coefficient and coefficient of repeatability were calculated to estimate intraobserver and interobserver repeatability for corneal NF density (numbers per square millimeter), nerve branch density (NBD; numbers per square millimeter), NF length (millimeters per square millimeter), and NF tortuosity coefficient. Symmetry between the right and left eyes was also assessed. Results: Intraclass correlation coefficient and coefficient of repeatability for intraobserver repeatability were 0.66 to 0.74 and 0.17 to 0.64, for interobserver repeatability 0.54 to 0.93 and 0.15 to 0.85, and for symmetry 0.34 to 0.77 and 0.17 to 0.63, respectively. NBD demonstrated low repeatability. Conclusions: This study demonstrates good repeatability for the manual assessment of all major corneal NF parameters with the exception of NBD, which highlights the difficulty in defining nerve branches and suggests the need for experienced observers or automated image analysis to ensure optimal repeatability.

We describe and evaluate an automated software tool for nerve-fiber detection and quantification in corneal confocal microscopy (CCM) images, combining sensitive nerve-fiber detection with morphological descriptors. Method: We have evaluated the tool for quantification of Diabetic Sensorimotor Polyneuropathy (DSPN) using both new and previously published morphological features. The evaluation used 888 images from 176 subjects (84 controls and 92 patients with type 1 diabetes). The patient group was further subdivided into those with (n = 63) and without (n = 29) DSPN. Results: We achieve improved nervefiber detection over previous results (91.7% sensitivity and specificity in identifying nerve-fiber pixels). Automatic quantification of nerve morphology shows a high correlation with previously reported, manually measured, features. Receiver Operating Characteristic (ROC) analysis of both manual and automatic measurement regimes resulted in similar results in distinguishing patients with DSPN from those without: AUC of about 0.77 and 72% sensitivity-specificity at the equal error rate point. Conclusion: Automated quantification of corneal nerves in CCM images provides a sensitive tool for identification of DSPN. Its performance is equivalent to manual quantification, while improving speed and repeatability. Significance: CCM is a novel in vivo imaging modality that has the potential to be a noninvasive and objective image biomarker for peripheral neuropathy. Automatic quantification of nerve morphology is a major step forward in the early diagnosis and assessment of progression, and, in particular, for use in clinical trials to establish therapeutic benefit in diabetic and other peripheral neuropathies.

Deep learning has been employed to prognostic and health management of automotive and aerospace with promising results. Literature in this area has revealed that most contributions regarding deep learning is largely focused on the model’s architecture. However, contributions regarding improvement of different aspects in deep learning, such as custom loss function for prognostic and health management are scarce. There is therefore an opportunity to improve upon the effectiveness of deep learning for the system’s prognostics and diagnostics without modifying the models’ architecture. To address this gap, the use of two different dynamically weighted loss functions, a newly proposed weighting mechanism and a focal loss function for prognostics and diagnostics task are investigated. A dynamically weighted loss function is expected to modify the learning process by augmenting the loss function with a weight value corresponding to the learning error of each data instance. The objective is to force deep learning models to focus on those instances where larger learning errors occur in order to improve their performance. The two loss functions used are evaluated using four popular deep learning architectures, namely, deep feedforward neural network, one-dimensional convolutional neural network, bidirectional gated recurrent unit and bidirectional long short-term memory on the commercial modular aero-propulsion system simulation data from NASA and air pressure system failure data for Scania trucks. Experimental results show that dynamically-weighted loss functions helps us achieve significant improvement for remaining useful life prediction and fault detection rate over non-weighted loss function predictions.

Objective: Corneal confocal microscopy (CCM), an in vivo ophthalmic imaging modality, is a noninvasive and objective imaging biomarker for identifying small nerve fiber damage. We have evaluated the diagnostic performance of previously established CCM parameters to a novel automated measure of corneal nerve complexity called the corneal nerve fiber fractal dimension (ACNFrD). Methods: A total of 176 subjects (84 controls and 92 patients with type 1 diabetes) with and without diabetic sensorimotor polyneuropathy (DSPN) underwent CCM. Fractal dimension analysis was performed on CCM images using purpose-built corneal nerve analysis software, and compared with previously established manual and automated corneal nerve fiber measurements. Results: Manual and automated subbasal corneal nerve fiber density (CNFD) (P < 0.0001), length (CNFL) (P < 0.0001), branch density (CNBD) (P < 0.05), and ACNFrD (P < 0.0001) were significantly reduced in patients with DSPN compared to patients without DSPN. The areas under the receiver operating characteristic curves for identifying DSPN were comparable: 0.77 for automated CNFD, 0.74 for automated CNFL, 0.69 for automated CNBD, and 0.74 for automated ACNFrD. Conclusions: ACNFrD shows comparable diagnostic efficiency to identify diabetic patients with and without DSPN.