Automatic segmentation in three-dimensional imaging of bone using artificial intelligent algorithms

The formation of three-dimensional (3-D) ultrasonic images requires the acquisition of the 3-D data set, the segmentation of the data into regions of different tissue types, and the rendering of the image. To produce 3-D ultrasonic images in real-time, the segmentation of tissues must be done automatically and rapidly. Two algorithms were developed based on artificial intelligence principles to segment data into the overlying soft-tissue and bone: an expert system utilizing the forward chaining application of rules, and a three-layer, counter-propagation neural network. The 3-D structure, a test object, and in-vivo bone were imaged with small error. The expert system performed better than the neural network and the neural network showed potential for future use