External clinical validation of automated software to identify structural abnormalities and microhaemorrhages in nailfold videocapillaroscopy images.

OBJECTIVES: Automated systems to analyse nailfold videocapillaroscopy (NVC) images are needed to promptly and comprehensively characterise patients with systemic sclerosis (SSc) or Raynaud's phenomenon (RP). We previously developed, and validated in-house, a deep convolutional neural network-based algorithm to classify NVC-captured images according to the presence/absence of structural abnormalities and/or microhaemorrhages. We present its external clinical validation. METHODS: A total of 1,164 NVC images of RP patients were annotated by 5 trained capillaroscopists according to the following categories: normal capillary; dilation; giant capillary; abnormal shape; tortuosity; microhaemorrhage. The images were also presented to the algorithm. Matches and discrepancies between algorithm predictions and those annotations obtained by consensus of ≥3 or ≥4 interobservers were analysed. RESULTS: Consensus among ≥3 capillaroscopists was achieved in 86.9% of images, 75.8% of which were correctly predicted by the algorithm. Consensus among ≥4 experts occurred in 52.0% of cases, in which 87.1% of the algorithm's results matched with those of the expert panel. The algorithm's positive predictive value was >80% for microhaemorrhages and unaltered, giant or abnormal capillaries. Sensitivity was >75% for dilations and tortuosities. Negative predictive value and specificity were >89% for all categories. CONCLUSIONS: This external clinical validation suggests that this algorithm is useful to assist in the diagnosis and follow-up of SSc or RP patients in a timely manner. It may also be helpful in the management of patients with any pathology presenting with microvascular changes, as the algorithm has been designed to also be useful for research aiming at extending the usage of nailfold capillaroscopy to more conditions.

The challenge of comprehensive nailfold videocapillaroscopy practice: a further contribution.

OBJECTIVES: Although classification systems and scores for capillaroscopy interpretation have been published, there is a lack of homogenization for the procedure, especially in the way and place the images are taken, the counting of the capillaries and the measuring of their size. Our objective is to provide a deep learning-based software to obtain objective and exhaustive data for the whole nailfold without increasing the time or effort needed to do the examination, or requiring expensive equipment. METHODS: An automated software to count nailfold capillaries has been designed, through an exploratory image dataset of 2,713 images with 18,000 measurements of 3 different types. Subsequently, application rules have been created to detect the morphology of nailfold videocapillaroscopy images, through a training set of images. The software reliability has been evaluated with standard metrics used in the machine learning field for object detection tasks, comparing automatic and manual counting on the same NVC images. RESULTS: A mean average precision (mAP) of 0.473 is achieved for detecting and classifying capillaries and haemorrhages by their shape, and a mAP of 0.515 is achieved for detecting and classifying capillaries by their size. A precision of 83.84% and a recall of 92.44% in the identification of capillaries was estimated. CONCLUSIONS: Deep learning is a useful tool in nailfold videocapillaroscopy that allows to analyse objectively and homogeneously images taken with multiple devices. It should make the assessment of the capillary morphology in nailfold video capillaroscopy easier, quicker, more complete and accessible to everyone.