Dataset Information

Skin strata delineation in reflectance confocal microscopy images using recurrent convolutional networks with attention.

ABSTRACT: Reflectance confocal microscopy (RCM) is an effective non-invasive tool for cancer diagnosis. However, acquiring and reading RCM images requires extensive training and experience, and novice clinicians exhibit high discordance in diagnostic accuracy. Quantitative tools to standardize image acquisition could reduce both required training and diagnostic variability. To perform diagnostic analysis, clinicians collect a set of RCM mosaics (RCM images concatenated in a raster fashion to extend the field view) at 4-5 specific layers in skin, all localized in the junction between the epidermal and dermal layers (dermal-epidermal junction, DEJ), necessitating locating that junction before mosaic acquisition. In this study, we automate DEJ localization using deep recurrent convolutional neural networks to delineate skin strata in stacks of RCM images collected at consecutive depths. Success will guide to automated and quantitative mosaic acquisition thus reducing inter operator variability and bring standardization in imaging. Testing our model against an expert labeled dataset of 504 RCM stacks, we achieved [Formula: see text] classification accuracy and nine-fold reduction in the number of anatomically impossible errors compared to the previous state-of-the-art.

SUBMITTER: Bozkurt A

PROVIDER: S-EPMC8206415 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BACKGROUND:The surgical removal of non-melanoma skin cancers (NMSCs) is guided by the pathologic examination of margins. However, the preparation of histopathology is time consuming, labour-intensive and requires separate laboratory infrastructure. Furthermore, when histopathology indicates positive margins, patients must return for re-excisions. Reflectance confocal microscopy (RCM) with a new video-mosaicking approach can noninvasively delineate margins directly on patients and potentially guide surgery in real-time, augmenting the traditional approaches of histopathology. OBJECTIVE:To assess a new peri-operative RCM video-mosaicking approach for comprehensive delineation of NMSC margins on patients in vivo. METHODS:Thirty-five patients undergoing Mohs micrographic surgery (MMS) in the Mohs surgery unit at Memorial Sloan Kettering Cancer Center, New York, NY were included in the study. RCM imaging was performed before and after the first staged excision by acquiring videos along the surgical margins (epidermal, peripheral and deep dermal) of each wound, which were subsequently processed into video-mosaics. Two RCM evaluators read and assessed video-mosaics, and subsequently compared to the corresponding Mohs frozen histopathology. RESULTS:Reflectance confocal microscopy videos and video-mosaics displayed acceptable imaging quality (resolution and contrast), pre-operatively in 32/35 (91%) NMSC lesions and intra-operatively in 29/35 lesions (83%). Pre-operative delineation of margins correlated with the histopathology in 32/35 (91%) lesions. Intra-operative delineation correlated in 10/14 (71%) lesions for the presence of residual tumour and in 18/21 (86%) lesions for absence. Sensitivity/specificity were 71%/86% and 86%/81% for two RCM video-mosaic evaluators, and overall agreement was 80% and 83% with histopathology, with moderate inter-evaluator agreement (k = 0.59, P ≤ 0.0002). CONCLUSIONS:Peri-operative RCM video-mosaicking of NMSC margins directly on patients may potentially guide surgery in real-time, serve as an adjunct to histopathology, reduce time spent in clinic and reduce the need for re-excisions. Further testing in larger studies is needed.

Dataset Information

Skin strata delineation in reflectance confocal microscopy images using recurrent convolutional networks with attention.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets