Knowledge-aware deep framework for collaborative skin lesion segmentation and melanoma recognition

Deep learning techniques have shown their superior performance in dermatologist clinical inspection. Nevertheless, melanoma diagnosis is still a challenging task due to the difficulty of incorporating useful knowledge into process. In this paper, we propose novel knowledge-aware deep framework that incorporates some collaborative two important tasks, i.e., skin lesion segmentation and recognition. Specifically, exploit morphological expressions region also periphery for identification, lesion-based pooling shape extraction (LPSE) scheme designed, which transfers structure information obtained from Meanwhile, pass recognition segmentation, an effective guided feature fusion (DGFF) strategy designed. Moreover, recursive mutual mechanism further promotes inter-task cooperation, thus iteratively improves joint capability model both Experimental results on publicly available datasets show effectiveness proposed method analysis.