Skip to main content
. Author manuscript; available in PMC: 2021 Mar 17.
Published in final edited form as: J Invest Dermatol. 2019 Dec 12;140(6):1214–1222. doi: 10.1016/j.jid.2019.10.018

Figure 4. Coregistered dermoscopic and DEJ mosaic image pairs with color overlays showing the extent of uninformative areas obscuring the perilesional skin.

Figure 4.

Uninformative areas cover (a) 27.3%, (b) 32.0%, (c) 13.0%, and (d) 34% of the entire mosaic with limited lesional involvement. MED-Net segmentation results are color-overlaid on RCM mosaics (row 3) and coregistered dermoscopic images (row 2). Raw images without overlays are shown in rows 1 and 4. In (a), poor skin-to-window contact on the left is due to curved body contour. In (b) and (d), the papular lesion contour results in similar poor window contact in perilesional skin. In (c), alternating areas of saturation and underillumination (black) are due to patient motion. In (c), the lesional area is free of uninformative areas and reimaging is unnecessary. In (a), (b), and (d), reimaging is unnecessary, but collecting additional mosaics to capture needed anatomic skin levels in all portions of the lesion is recommended. DEJ, dermal-epidermal junction; MED-Net, Multiscale Encoder Decoder Network; RCM, reflectance confocal microscopy.