Lateral Resolution Enhancement of Ultrasound Image Using Neural Networks

Ultrasound imaging techniques are widely applied for medical diagnosis. However its main disadvantage compared with other techniques are low resolution and speckle artifacts. Therefore, improving the spatial resolution of ultrasound images has been a very active research area. System parameters that affect resolution include the size of the active transducer aperture, the center frequency and the bandwidth of the transducer, and the selected transmit focal depth. This paper presents an approach to enhance the lateral resolution by reproducing the underlying mapping between a low resolution image and a high resolution image obtained with different aperture size. Artificial neural networks are used to build the mapping. A generalization parameter is added to neural network and fuzzy logic is used to perform the resulting data fusion. In addition, a local histogram algorithm is used to filter the speckle like sample data when training network.