Adaptive Architecture Neural Nets for Medical Image Compression

In this paper a technique is proposed for medical image compression using neural network based vector quantizers. There exist hundreds of modalities of medical images and each modality has hundreds of sub classes for different organs and different sizes, in such a situation it is difficult to generalize a neural network for all modalities. To tackle this problem and having a prior knowledge about similar nature and size of acquisition for a single type of medical image, we propose a flag byte which is automatically set by the image size and some features. This flag byte is then used to select the size of the net codebook to be used. Proposed method not only leads to dynamic architectures of neural nets used but also towards an adaptive selection of codebooks. This method yields high compression ratios with much better quality than existing standards