Skin Cancer Classification using Delaunay Triangulation and Graph Convolutional Network

Oftentimes, many people or even medical workers misdiagnose skin cancer, which may lead to malpractice and thus, resulting in delayed recovery life-threatening complications. In this research, a Graph Convolutional Network (GCN) method is proposed as classification model Delaunay triangulation its feature extraction classify various types of cancers. serves the purpose boundary extraction, implementation allows focus only on cancerous lesion ignore around it. This way, cancer can be predicted more accurately. Furthermore, GCN offers advantages image analysis over traditional CNN models. interactions between different regions structures an perform messaging nodes, whereas not explicitly designed do such thing. Other than that, also leverage transfer learning few-shot techniques address challenges limited annotated datasets. However, result shows that tends overfit unable generate correct predictions for new images. There are several reasons could overfit, imbalance data, incorrect insufficient features data prediction, containing noise.