Application of neural networks to compression of CT images

Efficient compression of medical images is needed to decrease the storage space and enable efficient image transfer over network for access of electronic patient records. Since the medical images contain diagnostically relevant information, it is necessary for the process of image compression to preserve high levels of image fidelity, especially when the images are compressed at low bit rates. This paper investigates the capacity of an artificial neural network framework for medical image compression. Specifically, the performance of the proposed image compression method is evaluated on a database of computed tomography images of lungs, where PSNR and MSE are used as the principal image quality metrics. The compressed image data is derived from the hidden layer outputs, where the artificial neural networks are trained to reconstruct the network input features. The results of image block segmentation are used as the network training features. The paper proposes the use of Kohonen's self-organizing maps for segmentation of feature space and the use of multiple finely tuned multi-layer perceptrons to achieve an improved compression performance. This paper presents a study on how the choice of block size, network architecture, and training method affect the compression performance. An attempt is made to optimize the artificial neural network framework for the compression of computed tomography lung images.