Automated Analysis of Skin Images from in vivo Reflectance Confocal Microscopy
Localization of Dermal Epidermal Junction Surface in 3D Reflectance Microscopy Images of Skin:
Reflectance confocal microscopy (RCM) continues to be translated toward the detection of skin cancers in vivo. Automated image analysis may help clinicians and accelerate clinical acceptance of RCM. For screening and diagnosis of cancer, the dermal/epidermal junction (DEJ), at which melanomas and basal cell carcinomas originate, is an important feature in skin.
Localization of Dermal Epidermal Junction in Lightly Pigmented Skin Types:
In 3D RCM image stacks of lightly pigmented skin types (fair skin), the DEJ is marked by optically subtle changes and features and is difficult to detect purely by visual examination. Challenges for automation of DEJ detection include heterogeneity of skin tissue, high inter-, intra-subject variability and low optical contrast. To cope with these challenges, for fair skin, we introduced a semi-automated hybrid sequence segmentation/classification algorithm that partitions z-stacks of tiles into homogenous segments by fitting a model of skin layer dynamics and then classifies tile segments as epidermis, dermis, or transitional DEJ region using texture features. (JBO’11, ISBI’08, ISBI’09)
Localization of Dermal Epidermal Junction in Pigmented Skin Types:
We then extended the algorithm to also dark skin types. The extended algorithm first decides the skin type and then applies the appropriate DEJ localization method. In dark skin, strong backscatter from the pigment melanin causes the basal cells above the DEJ to appear with high contrast. To locate those high contrast regions, the algorithm operates on small tiles (regions) and finds the peaks of the smoothed average intensity depth profile of each tile. However, for some tiles, due to heterogeneity of skin tissue, multiple peaks in the depth profile exist and the strongest peak might not be the basal layer peak. To select the correct peak, basal cells are represented with a vector of texture features. The peak with most similar features to this feature vector is selected. Finally the detected surface is fitted to a spline surface with regularization for smoothing. (SPIE BIOS’11)
Resulting DEJ surfaces:
These algorithms work on a similar basis. They all use feature profiles as a function of depth:
• Skin type detection: Mean intensity profile in depth direction to detect peak slice
• Dark skin DEJ Detection: Mean intensity profile for each tile in depth direction to detect multiple peaks. Selects correct peak next.
• Fair skin DEJ transition region detection: Use mean texture features for each tile in depth direction. Partition into homogenous segments according to dynamics and classify each segment as epidermis or dermis to detect DEJ transition zone boundaries.
Fair skin:
Surface plot of the epidermis and the dermis boundaries trapping DEJ in between in 3-D.
• The multi-colored surfaces indicate the algorithm boundaries.
• The color maps indicate the distance from the expert boundary.
Mean error for 4 stacks: 8-10um (Less than size of a cell which is 12 um)
Dark skin:
Surface plot of the DEJ in 3-D.
Mean error for 5 stcaks: 3-8 um
Video showing fair skin results.
Video showing fair skin results.
Fair skin results: Two boundaries trapping DEJ in between.
Dark skin results: DEJ boundary located by the algorithm and by the expert .
