Automated detection of melanoma

摘要

Embodiments herein exploit the optical sectioning capability of reflectance confocal microscopy to non-invasively survey the dermal-epidermal junction (DEJ), noting the irregularities associated with malignancy. Methods are provided to aid a clinician in diagnosing melanoma through pattern recognition to extract pertinent diagnostic information from large 3D confocal images. Identifying the combination of pagetoid melanocytes and DEJ breakdown increases the accuracy of detection. A method may be used to process a 3D confocal volume of images taken by a clinician of a suspicious lesion and deduce the depth location z of the first reflective surface (FRS) at each x-y position. This FRS is where the most superficial melanin resides. In this manner, the stratum corneum and epidermis are digitally stripped and no longer distract the clinician from the more diagnostically relevant pigmented cell network. The FRS is putatively either the DEJ for benign nevi or the depth of a pagetoid melanocyte at x,y above the DEJ. Thus by creating a 3D surface plot of the FRS for each lateral x,y point, the presence of pagetoid cells is identified by a discreet jump in the FRS above the level of the DEJ. The DEJ map may also be used to determine if breakdown of the DEJ is occurring.