Reproducibility of the scleroderma pattern assessed by wide-field capillaroscopy in subjects suffering from Raynaud's phenomenon

Analysis of nailfold capillaroscopy images with artificial intelligence: Data from literature and performance of machine learning and deep learning from images acquired in the SCLEROCAP study

OBJECTIVE

To evaluate the performance of machine learning and then deep learning to detect a systemic scleroderma (SSc) landscape from the same set of nailfold capillaroscopy (NC) images from the French prospective multicenter observational study SCLEROCAP.

METHODS

NC images from the first 100 SCLEROCAP patients were analyzed to assess the performance of machine learning and then deep learning in identifying the SSc landscape, the NC images having previously been independently and consensually labeled by expert clinicians. Images were divided into a training set (70 %) and a validation set (30 %). After features extraction from the NC images, we tested six classifiers (random forests (RF), support vector machine (SVM), logistic regression (LR), light gradient boosting (LGB), extreme gradient boosting (XGB), K-nearest neighbors (KNN)) on the training set with five different combinations of the images. The performance of each classifier was evaluated by the F1 score. In the deep learning section, we tested three pre-trained models from the TIMM library (ResNet-18, DenseNet-121 and VGG-16) on raw NC images after applying image augmentation methods.

RESULTS

With machine learning, performance ranged from 0.60 to 0.73 for each variable, with Hu and Haralick moments being the most discriminating. Performance was highest with the RF, LGB and XGB models (F1 scores: 0.75-0.79). The highest score was obtained by combining all variables and using the LGB model (F1 score: 0.79 ± 0.05, p < 0.01). With deep learning, performance reached a minimum accuracy of 0.87. The best results were obtained with the DenseNet-121 model (accuracy 0.94 ± 0.02, F1 score 0.94 ± 0.02, AUC 0.95 ± 0.03) as compared to ResNet-18 (accuracy 0.87 ± 0.04, F1 score 0.85 ± 0.03, AUC 0.87 ± 0.04) and VGG-16 (accuracy 0.90 ± 0.03, F1 score 0.91 ± 0.02, AUC 0.91 ± 0.04).

CONCLUSION

By using machine learning and then deep learning on the same set of labeled NC images from the SCLEROCAP study, the highest performances to detect SSc landscape were obtained with deep learning and in particular DenseNet-121. This pre-trained model could therefore be used to automatically interpret NC images in case of suspected SSc. This result nevertheless needs to be confirmed on a larger number of NC images.

Automatic assessment of nailfold capillaroscopy software: a pilot study

Introduction

Capillaroscopy is a simple method of nailfold capillary imaging, used to diagnose diseases from the systemic sclerosis spectrum. However, the assessment of the capillary image is time-consuming and subjective. This makes it difficult to use for a detailed comparison of studies assessed by various physicians. This pilot study aimed to validate software used for automatic capillary counting and image classification as normal or pathological.

Material and methods

The study was based on the assessment of 200 capillaroscopic images obtained from patients suffering from systemic sclerosis or scleroderma spectrum diseases and healthy people. Dinolite MEDL4N Pro was used to perform capillaroscopy. Each image was analysed manually and described using working software. The neural network was trained using the fast.ai library (based on PyTorch). The ResNet-34 deep residual neural network was chosen; 10-fold cross-validation with the validation and test set was performed, using the Darknet-YoloV3 state of the art neural network in a GPU-optimized (P5000 GPU) environment. For the calculation of 1 mm capillaries, an additional detection mechanism was designed.

Results

The results obtained under neural network training were compared to the results obtained in manual analysis. The sensitivity of the automatic tool relative to manual assessment in classification of correct vs. pathological images was 89.0%, specificity 89.4% for the training group, in validation 89.0% and 86.9% respectively. For the average number of capillaries in 1 mm the precision of real images detected within the region of interest was 96.48%.

Conclusions

The pilot software for fully automatic capillaroscopic image assessment can be a useful tool for the rapid classification of a normal and altered capillaroscopy pattern. In addition, it allows one to quickly calculate the number of capillaries. In the future, the tool will be developed and will make it possible to obtain full imaging characteristics independent of the experience of the examiner.

Treatment with mycophenolate mofetil is associated with improved nailfold vasculature in systemic sclerosis

OBJECTIVE

To investigate the evolution of nailfold capillary density in patients with SSc in relation to immunosuppressive treatment and autoantibodies.

METHODS

This was a prospective study cohort. Consecutive newly diagnosed SSc patients were included into this study who, in a retrospective review, had at least two nailfold capillary microscopy measurements performed during the first 48 months of follow-up. Capillary density per 3 mm was measured with widefield nailfold capillary microscopy. Improvement of capillary density per finger and mean capillary density were analysed. Longitudinal measurements of mean capillary density were analysed by generalized estimating equation.

RESULTS

Eighty patients (68 women, 12 men) met the inclusion criteria. The median follow-up time was 27 months. Twenty-eight patients had an improved capillary density in per-finger analysis. MMF was associated with fewer numbers of fingers that had worsened in capillary density. Anti-topoisomerase antibodies were associated with low mean capillary density. Anti-RNA polymerase III antibodies were associated with improvement and anti-centromere antibodies with worsening of capillary density in per-finger analysis. MMF treatment was associated with less steep capillary density decline in a moderated generalized estimating equation model including presence of anti-topoisomerase antibodies and the interaction of MMF with follow-up time.

CONCLUSION

Nailfold capillary density improved over time in a substantial proportion of SSc patients. MMF treatment had a positive impact on the evolution of capillary density in these patients. SSc autoantibody phenotype may affect the capillary density development. The data support previous hypotheses that early immunosuppression may favourably affect vascular regeneration in SSc.

Innate Immune Modulation Induced by EBV Lytic Infection Promotes Endothelial Cell Inflammation and Vascular Injury in Scleroderma

Microvascular injury is considered an initial event in the pathogenesis of scleroderma and endothelial cells are suspected of being the target of the autoimmune process seen in the disease. EBV has long been proposed as a trigger for autoimmune diseases, including scleroderma. Nevertheless, its contribution to the pathogenic process remains poorly understood. In this study, we report that EBV lytic antigens are detected in scleroderma dermal vessels, suggesting that endothelial cells might represent a target for EBV infection in scleroderma skin. We show that EBV DNA load is remarkably increased in peripheral blood, plasma and circulating monocytes from scleroderma patients compared to healthy EBV carriers, and that monocytes represent the prominent subsets of EBV-infected cells in scleroderma. Given that monocytes have the capacity to adhere to the endothelium, we then investigated whether monocyte-associated EBV could infect primary human endothelial cells. We demonstrated that endothelial cells are infectable by EBV, using human monocytes bound to recombinant EBV as a shuttle, even though cell-free virus failed to infect them. We show that EBV induces activation of TLR9 innate immune response and markers of vascular injury in infected endothelial cells and that up-regulation is associated with the expression of EBV lytic genes in infected cells. EBV innate immune modulation suggests a novel mechanism mediating inflammation, by which EBV triggers endothelial cell and vascular injury in scleroderma. In addition, our data point to up-regulation of EBV DNA loads as potential biomarker in developing vasculopathy in scleroderma. These findings provide the framework for the development of novel therapeutic interventions to shift the scleroderma treatment paradigm towards antiviral therapies.

Nailfold capillaroscopy findings in cutaneous lupus erythematosus patients with or without digital lesions and comparison with dermatomyositis patients: A prospective study

BACKGROUND

Differential diagnosis between cutaneous lupus erythematosus (CLE) and dermatomyositis (DM) may be challenging if digital lesions occur.

OBJECTIVES

To compare nailfold capillaroscopy (NFC) findings in CLE patients with or without digital involvement, and to compare capillaroscopic findings between CLE patients with digital lesions and DM patients.

METHODS

Prospective monocentric study including CLE and DM patients. NFC was performed and standardized items were recorded.

RESULTS

Fifty-one CLE patients and 10 DM patients with digital lesions were included. A scleroderma pattern was found in 6 patients (12%): in 5 out of 17 patients with digital lesions, compared with only 1 out of 34 patients without digital lesions (p = 0.01). In multivariate analysis, CLE digital lesions and digital ulcerations were statistically associated with scleroderma pattern. CLE digital lesions were significantly associated with architectural disorganization (p = 0.0003) and capillary rarefaction (p = 0.0038). A scleroderma pattern was significantly more frequent in DM patients (80%) than in CLE patients with digital lesions (30%, p = 0.018). Capillaroscopic findings were not significantly different between CLE patients with digital lesions and DM patients.

CONCLUSION

Although scleroderma pattern is more frequent in DM patients than in CLE patients with digital lesions, NFC cannot formally distinguish CLE from DM.

Capillaroscopy: a useful tool in the early diagnosis of connective tissue disease and nonscleroderma spectrum disorders

BACKGROUND

Detection of early capillaroscopic alterations in the preclinical phase may prove useful in patients with non-scleroderma connective tissue disease (CTD). We aimed to verify whether certain capillaroscopic alterations, alone or in combination, might be predictive of CTD.

METHODS

We retrospectively collected data on patients with Raynaud's phenomenon who underwent capillaroscopy conducted by highly expert examiners with a degree in vascular medicine at our institutions. Included subjects were divided in two groups: those developing rheumatic disease during follow-up, and those without subsequent diagnosis of CTD. Notably, we excluded subjects who presented with an evident scleroderma pattern or rheumatic disease during their initial examination.

RESULTS

We included a total of 76 patients, 60 who developed CTD during follow-up, which spanned in this group 23±7 months, and 16 who did not develop CTD during follow-up, which spanned 23±9 months. The following features were significantly associated with Raynaud's phenomenon: 1) angiotectonic disorder (p<0.001), 2) nonhomogeneous loop morphology (p<0.001), 3) avascular areas (p<0.001), 4) pseudo-avascular areas (p<0.001), and, albeit to a lesser degree, 5) ectasias (p=0.050). Notably, the initial capillaroscopic pattern did not undergo any changes in subsequent tests.

CONCLUSIONS

Although certain pathological characteristics of the capillaroscopic pattern are nonspecific and not diagnostic if considered individually, they can be significantly suggestive for latent CTD when found in combination. At the very least, they warrant an in-depth diagnostic analysis and a lengthy follow-up.

Nailfold capillaroscopy in systemic sclerosis - state of the art: The evolving knowledge about capillaroscopic abnormalities in systemic sclerosis

Capillaroscopy is a unique method for morphological evaluation of the nailfold capillaries that plays a crucial role for early diagnosis of systemic sclerosis. The first description of the pathological capillaroscopic changes in systemic sclerosis was made by Brown and O'Leary in 1925. Several decades later they have been validated and accepted as a diagnostic criterion in the current 2013 EULAR/ACR classification criteria. This article summarizes the evolving knowledge about the use of nailfold capillaroscopy in systemic sclerosis. Initially, Maricq et al. suggested two major categories of capillaroscopic findings in systemic sclerosis - an 'active'and 'slow' capillaroscopic pattern. Their description and terminology suggested a correlation between capillaroscopic changes and disease activity and progression. In the later classification of Cutolo et al., three phases were defined, i.e. 'early', 'active' and 'late' that reflect the time-related evolution of the capillaroscopic changes suggesting their association with disease duration. Current knowledge about the microvascular changes in systemic sclerosis supports both associations with disease activity and disease duration. The general opinion about the association of capillaroscopic findings with clinical involvement and disease activity in systemic sclerosis is not uniform. This is supposedly because the phase changes of systemic sclerosis-related microangiopathy are almost a universal feature in scleroderma and are not specific for a certain type of an accompaning clinical manifestation. Thus, the speed of progression of microvascular alterations might be a decisive criterion, and in cases of rapid dynamics of capillaroscopic findings, it could be considered as an indicator of disease activity. Interestingly, vascular 'recovery' has been observed after treatment with immunosuppressive drugs, haematopoietic stem cell transplantation, and the endothelin receptor antagonist - bosentan. The evolving knowledge about nailfold capillaroscopy in systemic sclerosis will further spread its application from a mainly diagnostic tool to an established, reliable method for evaluation of disease activity, prognosis and therapeutic response.

Automated assessment of absolute nailfold capillary number on videocapillaroscopic images: Proof of principle and validation in systemic sclerosis

BACKGROUND

Absolute nailfold capillary number should be a putative biomarker in selected rheumatic diseases but could be time-consuming and not highly repeatable.

OBJECTIVE

To validate an automated software for absolute nailfold capillary number and density evaluation, on NVC images in SSc.

METHODS

An automated software to count nailfold capillary number (AUTOCAPI) had been constructed, through an exploratory image set. Subsequently, application rules have been created to define the ROI in NVC images, through a training images set. The software reliability was assessed through calculation of the ICC between automatic and manual counting, by four independent observers, on the same NVC images.

RESULTS

The following ICC's were obtained per observer, for the patients with SSc (40 images), the healthy (20 images), and the PRP subgroups (20 images), respectively: 0.94, 0.81, and 0.62 (observer 1); 0.94, 0.91, and 0.67 (observer 2); 0.88, 0.56, and 0.64 (observer 3); and 0.88, 0.85, and 0.85 (observer 4).

CONCLUSIONS

The validation of an automated software for measuring absolute nailfold capillary number and density in SSc was achieved. The integration into the pre-existing imaging software should make the assessment of the capillary number in NVC easier, quicker, and standardized.