Multi-step breast mass image segmentation method based on MFC-PCNN

In order to solve the problem that pulse coupled neural network (PCNN) has low image segmentation accuracy and high computational complexity for image segmentation aspect, this paper proposes a multi-step medical image processing method that combines saliency detection and a modified PCNN model. First, an improved pulse coupled neural network model (MFC-MSPCNN) is proposed based on the FC-MSPCNN model. This method simplifies the related setting parameters, sets a new connection matrix, and improves the attenuation factor α according to the MFC-MSPCNN characteristics. The achieved steps of the method firstly use a GBVS algorithm based on the saliency detection mechanism to obtain the saliency region map of the mass, and then use it as an external input of the MFC-MSPCNN model to accurately segment the breast mass region. The experiments show that our proposed method can accurately segment breast masses and has low computational complexity than other prevalent methods.