Automated texture-based segmentation of hair follicles in volumetric optical coherence tomography images of dermatologic burn wounds

Adnexal structures in skin, such as sweat glands and hair follicles, harbor stem cells that are indispensable for wound healing following acute injuries such as burns of various degrees [1]. Developing a method to identify and visualize these structures in the dermal layer of tissue would greatly supplement imaging studies probing burn injury depth-extent, to assess whether intact adnexal structures will contribute to wound healing and provide invaluable prognostic information in the clinic. Two methods were recently attempted for this purpose in literature but held limitations of resolution and lack of proper filtering [2, 3], thus here we present an automated segmentation algorithm for volumetrically delineating such adnexal structures as hair follicles from surrounding skin tissue by performing texture-based analysis of optical coherence tomography (OCT) images of burn wounds. OCT, a label-free imaging modality, has been previously used for burn wound examination and is ideally suited for dermatological studies demanding high resolution, real-time, and noncontact evaluation to avoid infliction of further injury and patient discomfort [4]. The proposed method is based on Gamma distribution fits of OCT speckle pattern histograms [5], with demonstrated sensitivity of the fitting parameters to follicle locations, the boundary detection of which is confirmed with histology. Our technique demonstrates high accuracy in hair follicle delineation and strong applicability to subsequent studies to assess skin regeneration potential based on follicle location in measured depth-extent of burn injury, enabling future clinical translation not only to burn wounds, but other types of skin injuries.