Medical image analysis by probabilistic modular neural networks
Yue Wang, Tülay Adalı, Sun‐Yuan Kung · 2002
A probabilistic neural network based technique is presented for unsupervised quantification and segmentation of the brain tissues from magnetic resonance image. The problem is formulated as distribution learning and relaxation labeling that may be particularly useful in quantifying and segmenting abnormal brain tissues where the distribution of each tissue type heavily overlaps. The new technique utilizes suitable statistical models for both the pixel and context images. The quantification is achieved by model-histogram fitting of probabilistic self-organizing mixtures and the segmentation by global consistency labeling through a probabilistic constraint relaxation network. Experimental results show the efficient and robust performance of the new algorithm.