Storage space reduction in Picture Archiving and Communication System using Generative Adversarial Network

Abstract This paper presents a new architecture of Picture Archiving and Communication System based on Conditional Generative Adversarial Network and Fractal Image compression. The Conditional Generative Adversarial Network architecture is based on the Convolutional Neural Network which enables the system to capture the similarity measures without using any handcrafted functions. Performance of the proposed design is evaluated by comparing it with the commonly used compression techniques in Picture Archiving and Communication System such as JPEG, PNG and TIFF. The efficiency of the proposed architecture is tested by using a custom client program that sends the modality images to the Picture Archiving and Communication System server. The simulation runs on computers in multiple networks to gather the data similar to real time healthcare institutions. The results show that the storage space consumption of the proposed design is only 30% in comparison with Picture Archiving and Communication System, which uses the latest Machine learning and conventional non fractal compression methods. It is also observed that the Generative Adversarial Network based Fractal Image compression can drastically reduce the compression time compared to the conventional fractal and non-fractal compression methods. The empirical analysis shows that the proposed Generative Adversarial Network architecture can be a promising method to reduce the space complexity of the system such as Picture Archiving and Communication System.