“…DL has been determined to be applicable in a variety of aspects for PAM such as image reconstruction [21] , [22] , [23] , [24] , [25] , image classification [26] , quantitative imaging [27] , image detection, [28] and, especially, image segmentation [26] , [29] , [30] , [31] , [32] , [33] . However, those studies almost have been reported about segmentation of the C-scan image (MAP image domain) [26] , [29] , [31] , [32] , [33] , not widely applied to 3D PA images for separating skin and blood vessels areas. Unlike other works, Chlis et al [30] used a DL method for processing the cross-sectional B-scan image to avoid the rigorous and time-consuming manual segmentation.…”