Reproducibility of capillaroscopic classifications of systemic sclerosis: results from the SCLEROCAP study

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.

Nailfold dermoscopy predicts the absence of a capillaroscopy sclerodermic pattern: The multicentre, prospective VASCUL-R trial

BACKGROUND

Nailfold capillaroscopy is recommended to diagnose primary or secondary Raynaud's phenomenon (RP). Capillaroscopy is normal in primary RP, which is the most frequent. Screening for RP capillary anomalies with nailfold dermoscopy has been promising.

OBJECTIVE

To determine whether normal nailfold dermoscopy-based on the absence of five criteria that define a sclerodermic pattern-is able to predict normal capillaroscopy with good positive-predictive value (PPV).

METHODS

Prospective, 2-phase (monocentre and multicentre) study on patients at first consultation for RP undergoing nailfold video capillaroscopy (NVC) and nailfold dermoscopy by two different 'blinded' trained observers, respectively, a vascular specialist and a dermatologist, not familiar with capillaroscopy. The five criteria noted were as follows: disorganization, megacapillaries, low capillary density, avascular areas and haemorrhages.

RESULTS

Based on 105 patients, the dermoscopy PPV for a normal NVC was 100% (p = 0.015), with 37.9% sensitivity, when no criterion was observed. Excluding haemorrhages, the PPV remained 100% (p < 0.0001), with sensitivity rising to 73.7% and 100% specificity.

CONCLUSIONS

Normal nailfold dermoscopy with the absence of four easy-to-observe criteria predicts normal NVC with an excellent PPV.

Heating of the neck or elbows alleviates Raynaud's phenomenon but has different effects on different types of patients with systemic sclerosis

OBJECTIVES

We previously reported that heating of the neck or elbows alleviated Raynaud's phenomenon in patients with systemic sclerosis and upregulated capillary extension factor angiopoietin-1 (Angpt-1) in the fingertips. In this study, we investigated which cases responded better to the effect of heating of the neck or elbows.

METHODS

The pre- to postheating change in the visual analogue scale (ΔVAS) for Raynaud's phenomenon was examined for correlation with age, disease duration, autoantibodies, disease types, corticosteroid dose, capillaroscopic nailfold capillary damage, fingertip Angpt-1 concentrations at baseline, and increased rate of Angpt-1 concentration.

RESULTS

The ΔVAS for elbow heating correlated positively with the baseline Angpt-1 concentration, whereas opposite correlation was observed for neck heating. The other items did not significantly correlate with the ΔVAS; however, the ΔVAS for elbow heating tended to be larger in patients with advanced capillary damage, whereas an opposite trend was observed for neck heating.

CONCLUSIONS

Elbow and neck heating alleviated Raynaud's phenomenon to a similar extent, but their mechanism was different. Heating of the elbows had a greater effect on patients with advanced capillary damage and lower fingertip Angpt-1 concentrations.

"Scleroderma" and "Scleroderma-like" Capillaroscopic Pattern-Differences and Similarities

INTRODUCTION

The "scleroderma" type capillaroscopic pattern is a reference pattern in rheumatology that is a diagnostic sign for systemic sclerosis (SSc) in an appropriate clinical context and is observed in more than 90% of scleroderma patients. Similar microvascular changes, the so-called "scleroderma-like", have been described albeit in a lower proportion of patients with other rheumatic diseases, such as dermatomyositis (DM), undifferentiated connective tissue diseases (UCTD), systemic lupus erythematosus (SLE), etc. Three distinct stages of "scleroderma" pattern have been suggested by Cutolo et al., i.e., "early", "active", and "late". However, disease duration is just one of the factors that contributes to the progression of microvascular changes, and in this regard, "active" or even "late" pattern could be observed in patients with shorter disease duration. In addition, stable microvascular changes could be found for long periods in other cases.

OBJECTIVE

The aim of the study was to assess the presence of differentiating features between "scleroderma" pattern in SSc and "scleroderma-like" pattern in other rheumatic diseases.

METHODS

684 capillaroscopic images demonstrating a "scleroderma" and "scleroderma-like" pattern have been analysed in the current retrospective cross-sectional study. 479 capillaroscopic pictures were obtained from 50 SSc patients, 105 from 7 DM patients, 38 from 10 rheumatoid arthritis (RA) patients, 36 images from 5 patients with SLE, and 26 images from 9 patients with UCTD. All capillaroscopic images used in the current analysis have fulfilled the criteria for "sclerderma/scleroderma-like" pattern, as the pathological changes in the capillaroscopic parameters have also been confirmed by quantitative measurement of capillary diameters, capillary density, and intercapillary distance. All the images have been categorized into one of the following groups, i.e., "early", "active" and "late" phases (according to the definition of Cutolo et al.), or "other" findings, the latter being specifically described as they could not be attributed to one of the other three categories.

RESULTS

479 capillaroscopic pictures were obtained from 50 scleroderma patients. 31 of them showed an "early", 391 an "active" phase, and 57 a "late" phase "scleroderma" type microangiopathy. In 69 images assessed as an "active" pattern, neoangiogenesis was found. In 43 out of 105 capillaroscopic pictures from DM patients, an "active" phase was detected; in 2 of the images, a "late" pattern was found, and in 60 capillaroscopic pictures, neoangiogenesis in combination with giant capillary loops was observed. Early microangiopathy was not found in this group. Among capillaroscopic images from SLE patients, "late" phase microangiopathy was not found. "Early" phase was present in 3 images, "active" phase in 29, neoangiogenesis in "active" phase in 4 pictures. Early microangiopathy was detected in 11 capillaroscopic pictures from RA patients (8 out of 9 patients), an "active" phase in 4 images (3 patients), and in 23 capillaroscopic images, neoangiogenesis with mild capillary derangement and capillary loss and single giant capillaries ("rheumatoid neoangiogenic pattern") were observed. Classic "late" type microangiopathy was not found in RA patients as well as among patients with UCTD. The predominant capillaroscopic pattern in UCTD patients was early microangiopathy (n = 23). The rest images from UCTD exhibited features of the "active" phase.

CONCLUSION

In conclusion, early microangiopathy was observed in RA, SLE, and UCTD patients, but not in patients with DM. An "active" phase "scleroderma" type capillaroscopic pattern was detected in all patient groups other than SSc, i.e., DM, SLE, RA, and UCTD. "Late" phase "scleroderma" type microangiopathy was present in patients with scleroderma and DM and was not observed in SLE, RA, and UCTD. Despite the fact that in some cases, microangiopathy in scleroderma and other rheumatic diseases may be indistinguishable, the results of the current research have shown the presence of some differentiating features between "scleroderma" and "scleroderma-like" microangiopathy that might be a morphological phenomenon associated with differences in the pathogenesis and the degree of microvascular pathology in various rheumatic diseases.

Altered ocular microvasculature in patients with systemic sclerosis and very early disease of systemic sclerosis using optical coherence tomography angiography

The vascular hypothesis of systemic sclerosis (SSc) would predict microvascular alterations should also affect anatomical regions like ocular microvasculature. The objective of this study was to evaluate retinal and choriocapillary vessel density (VD) in patients with definite SSc and very early disease of systemic sclerosis (VEDOSS) using optical coherence tomography angiography (OCTA). 22 eyes of 22 patients and 22 eyes of 22 healthy subjects were included in this study. Patients were classified into patients with definite SSc and patients with VEDOSS. VD data of the superficial OCT angiogram (OCTA-SCP), deep OCT angiogram (OCTA-DCP) and choriocapillaris (OCTA-CC) were analysed. VD in the OCTA-SCP and OCTA-CC was lower in patients with SSc (p < 0.05). In VEDOSS patients, VD in the OCTA-CC was still reduced compared to controls (p < 0.05). Correlation analysis revealed a positive correlation between nailfold capillaroscopy and VD of OCTA-CC (Spearman correlation coefficient (rSp) 0.456, p < 0.05) and a negative correlation between skin score and VD of OCTA-SCP (p < 0.05). Ocular perfusion seems to be impaired in patients with SSc and even VEDOSS. VD correlated with disease severity. OCTA could be a new useful diagnostic and predictive parameter for monitoring patients with different stages of the disease.

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.

Retinal microcirculation abnormalities in patients with systemic sclerosis: an explorative optical coherence tomography angiography study

OBJECTIVES

To investigate subclinical or clinical abnormalities in retinal and choroidal vascular plexuses in patients with systemic sclerosis (SSc) by means of optical coherence tomography angiography (OCT-A).

METHODS

A total of 20 consecutive SSc patients were recruited and compared to 20 healthy subjects. Quantitative analysis of vessel density (VD), choriocapillaris plexus flow index (CCP-FI) and choroidal vascularity index were performed on OCT-A images in the superficial capillary plexus (SCP), deep capillary plexus (DCP) and choriocapillaris for all patients. Images were further reviewed by two independent readers for the assessment of qualitative abnormalities, including tortuosity, rarefaction areas, megacapillaries and macular-foveal capillaries (MFC).

RESULTS

The DCP-VD in the whole scan and in parafoveal, superior, inferior, nasal and temporal regions was significantly lower in the SSc group. The CC-FI was significantly higher in SSc patients. When comparing SSc patients with and without digital ulcers (DUs), significantly reduced SCP-VD was demonstrated in the whole, parafoveal, superior, inferior, temporal and nasal regions. No difference in any of the OCT-A parameters was observed when comparing patients with or without interstitial lung disease (ILD). Qualitative analysis of OCT-A revealed at least one abnormality in 95% of patients.

CONCLUSIONS

We showed the ability of OCT-A to disclose early ocular vascular abnormalities in patients with SSc. Our results may represent a hypothesis-generating basis for exploring the potential role of OCT-A in diagnosis, monitoring and prognosis stratification in SSc.

New perspectives in the imaging of Raynaud's phenomenon

The last 10-20 years have seen huge strides in imaging science. The aim of this review article is to share with the reader the key recent advances in non-invasive imaging of the digital (finger) vasculature in patients with Raynaud's phenomenon (RP), including in systemic sclerosis (SSc)-related digital vasculopathy. For the rheumatologist, seeing a patient with RP is an opportunity for early diagnosis of an underlying SSc-spectrum disorder or (conversely) for reassuring the patient with primary (idiopathic) RP. Non-invasive imaging techniques can help to provide diagnostic certainty. In addition, they can provide new insights into pathophysiology and have the potential to facilitate the development of much needed effective treatments by providing primary and secondary endpoints for randomized controlled trials: validation studies are ongoing. This review article focuses on nailfold capillaroscopy, thermography, and laser Doppler methods but also discusses (briefly) other technologies, including optical coherence tomography, multispectral imaging, and photoacoustic imaging. Key recent advances are the increasing use/availability of nailfold capillaroscopy (and better understanding of the role of low-cost hand-held devices), increased accessibility of thermography (including mobile phone thermography), and increased application of laser Doppler methods to the study of RP/digital vasculopathy (in particular of laser Doppler imaging and laser speckle contrast imaging, both of which measure blood flow over an area rather than at a single site). In an era of precision medicine, non-invasive imaging techniques can help stratify risk of (a) SSc in the patient with RP and (b) digital vascular disease progression in the patient with an SSc-spectrum disorder.

Fast track algorithm: How to differentiate a "scleroderma pattern" from a "non-scleroderma pattern"

OBJECTIVES

This study was designed to propose a simple "Fast Track algorithm" for capillaroscopists of any level of experience to differentiate "scleroderma patterns" from "non-scleroderma patterns" on capillaroscopy and to assess its inter-rater reliability.

METHODS

Based on existing definitions to categorise capillaroscopic images as "scleroderma patterns" and taking into account the real life variability of capillaroscopic images described standardly according to the European League Against Rheumatism (EULAR) Study Group on Microcirculation in Rheumatic Diseases, a fast track decision tree, the "Fast Track algorithm" was created by the principal expert (VS) to facilitate swift categorisation of an image as "non-scleroderma pattern (category 1)" or "scleroderma pattern (category 2)". Mean inter-rater reliability between all raters (experts/attendees) of the 8th EULAR course on capillaroscopy in Rheumatic Diseases (Genoa, 2018) and, as external validation, of the 8th European Scleroderma Trials and Research group (EUSTAR) course on systemic sclerosis (SSc) (Nijmegen, 2019) versus the principal expert, as well as reliability between the rater pairs themselves was assessed by mean Cohen's and Light's kappa coefficients.

RESULTS

Mean Cohen's kappa was 1/0.96 (95% CI 0.95-0.98) for the 6 experts/135 attendees of the 8th EULAR capillaroscopy course and 1/0.94 (95% CI 0.92-0.96) for the 3 experts/85 attendees of the 8th EUSTAR SSc course. Light's kappa was 1/0.92 at the 8th EULAR capillaroscopy course, and 1/0.87 at the 8th EUSTAR SSc course.

CONCLUSION

For the first time, a clinical expert based fast track decision algorithm has been developed to differentiate a "non-scleroderma" from a "scleroderma pattern" on capillaroscopic images, demonstrating excellent reliability when applied by capillaroscopists with varying levels of expertise versus the principal expert and corroborated with external validation.

Position article and guidelines 2018 recommendations of the Brazilian Society of Rheumatology for the indication, interpretation and performance of nailfold capillaroscopy

Nailfold capillaroscopy (NFC) is a reproducible, simple, low-cost, and safe imaging technique used for morphological analysis of nail bed capillaries. It is considered to be extremely useful for the investigation of Raynaud's phenomenon and for the early diagnosis of systemic sclerosis (SSc). The capillaroscopic pattern typically associated with SSc, scleroderma ("SD") pattern, is characterized by dilated capillaries, microhemorrhages, avascular areas and/or capillary loss, and distortion of the capillary architecture. The aim of these recommendations is to provide orientation regarding the relevance of NFC, and to establish a consensus on the indications, nomenclature, the interpretation of NFC findings and the technical equipments that should be used. These recommendations were formulated based on a systematic literature review of studies included in the database MEDLINE (PubMed) without any time restriction.

Abnormal Nailfold Capillaroscopy Is Common in Patients with Connective Tissue Disease and Associated with Abnormal Pulmonary Function Tests

OBJECTIVE

To assess the presence of a systemic sclerosis (SSc) pattern on nailfold capillary microscopy (NCM) in patients with Raynaud phenomenon (RP) and to explore its association with abnormal pulmonary function tests (PFT).

METHODS

NCM patterns were assessed in 759 consecutive patients with RP. Patterns were classified as normal (n = 354), nonspecific (n = 159), or SSc pattern (n = 246). Abnormal PFT was defined as forced vital or diffusion capacity < 70%. Patients were classified as primary RP (n = 245), or secondary: no definite diagnosis (n = 391), SSc (n = 40), primary Sjögren syndrome (pSS; n = 30), systemic lupus erythematosus (SLE; n = 30), mixed connective tissue disease (MCTD; n = 7), rheumatoid arthritis (RA; n = 15).

RESULTS

An SSc pattern on NCM was frequently observed in most patients with a definite diagnosis: SSc (88%), pSS (33%), SLE (17%), MCTD (71%), and RA (13%). In patients without definite diagnosis, 17% had a normal NCM pattern, 35% nonspecific, and 48% SSc pattern. Abnormal PFT was more frequent in patients with an SSc pattern (35.9% vs 19.5%, p = 0.002), even when corrected for SSc diagnosis (p = 0.003). Absence of an SSc pattern had high negative predictive value (88%); positive predictive values were low.

CONCLUSION

SSc pattern on NCM is common in patients with RP, and in those with connective tissue diseases other than SSc. It is associated with a higher prevalence of abnormal PFT, independent of the presence of an SSc diagnosis. Although these data need validation in a prospective setting, they underline the importance of NCM in RP and putative value to stratify the risk of pulmonary involvement in early stages of disease.

Advances in nailfold capillaroscopic analysis in systemic sclerosis

Systemic sclerosis is an autoimmune connective tissue disease characterized by early and persistent microvascular impairment which leads to functional and organic manifestations, with progressive fibrosis of the skin and internal organs. Morphological and functional assessment of the peripheral microvasculature is a must, not only for diagnosis but also for the prognosis and therapeutical follow-up of systemic sclerosis patients, as reported in recent studies. Nailfold videocapillaroscopy is the validated technique for the study of scleroderma microangiopathy as it is able to detect peripheral microvascular morphology and both classify and score the capillary abnormalities into different microangiopathy patterns ('Early', 'Active' and 'Late'). Indeed, the possibility to early diagnose and follow the microvascular changes and the safety of the technique have made nailfold videocapillaroscopy a mandatory tool for patient evaluation and included its assessment in the new systemic sclerosis classification criteria. Important links between nailfold videocapillaroscopy patterns and systemic sclerosis clinical manifestations have been described.

Can nailfold videocapillaroscopy images be interpreted reliably by different observers? Results of an inter-reader and intra-reader exercise among rheumatologists with different experience in this field

Nailfold videocapillaroscopy (VCP) allows non-invasive assessment of the microcirculation. Adequate training in this field is relevant for rheumatologists. There is increasing evidence of the reliability of VCP findings among different readers. Objective: To evaluate inter- and intra-reader agreement of rheumatologists to identify normal images and systemic sclerosis (SSc) patterns on VCP ("early," "active," and "late" proposed by Cutolo et al.). Thirteen rheumatologists with different experience in nailfold VCP received training to standardize reading criteria. They rated 60 VCP images from healthy and SSc patients at baseline and 4 weeks later, using an electronic platform. The reading of an expert was considered the gold standard. Data were analyzed using Cohen's kappa for concordance and Student's t test and ANOVA to compare kappa means for inter-reader, intra-reader, and inter-pattern readings. Mean inter-reader and intra-reader kappa were 0.45 and 0.49, respectively, (moderate agreement). Kappa scores were higher among experienced vs inexperienced readers (inter-reader kappa 0.58 vs 0.34, p = 0.001, intra-reader kappa 0.65 vs 0.37, p = 0.01). Agreement was substantial (kappa = 0.61) for the identification of normal vs abnormal images and higher for the identification of active (0.48, p = 0.009) and late SSc patterns (0.56, p = 0.008) than for the early SSc pattern (0.35, p = 0.003). There is moderate agreement among rheumatologists for the identification of SSc videocapillaroscopy patterns (higher among experienced rheumatologists) and substantial agreement, regardless of previous experience in VCP, in the identification of normal and abnormal images. Agreement for the identification of active and late patterns is higher than for the early pattern.

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

Objectives

The aim of this work was to study inter- and intra-observer agreement for the diagnosis of scleroderma pattern by wide-field capillaroscopy.

Methods

Images were taken from 50 patients known to have SSc and 50 controls consulting for RP who did not have SSc. These images were rated simultaneously by 11 experienced vascular medicine physicians as scleroderma pattern or not. Two weeks later, 7 of the 11 observers again rated the same images.

Results

Inter-observer agreement was almost perfect between the 11 observers (κ 0.86 ± 0.01), and the proportion of concordant observations was 79% (70-87). When each observer was compared with the reference, agreement was also almost perfect: κ coefficient 0.92 ± 0.03 and proportion of concordant observations 79% (70-87). Intra-observer agreement was also almost perfect: median κ coefficient 0.94 (0.78-0.96) and median proportion of concordant observations 97% (89-98).

Conclusion

Excellent inter- and intra-observer agreement was obtained in experienced vascular physicians for the diagnosis of capillaroscopic landscape by wide-field nailfold capillary microscopy.