Reconstruction of Medical Images from Sparse Data: A Deep Learning Approach

For diagnosis, medical imaging technology plays a significant role in today’s scenario. Accordingly, enormous amounts of medical data are generated daily on research centers and hospitals. Storing the medical images needs a lot of space where in 3D (Three dimensional) medical data acquires a huge space over it. All the medical image details need to be stored digitally for efficient diagnosis. The Digital Imaging and Communications in Medicine (DICOM) standard was established as the global norm for digitally storing patient images from diverse medical imaging modalities. Efficient compression over medical images is required to maintain efficacy in storage space. Compression over images like JPEG, PNG, etc. may not be efficient in bringing an effective lossless image compression in all the types of imaging modalities. Hence, an efficient methodology of Image compression over medical images are required in today’s scenario. This research study compares the performance of different methods of compression standard for medical images with the proposed method called DL Based Compressive sensing on Medical Image compression. The primary aim of Compressive Sensing (CS) is to achieve compression of a sparse signal at a sampling rate below the Nyquist rate. Results are analyzed using performance metrics like PSNR, MSE, BPP and CR. And it proves that the proposed algorithm achieves a higher PSNR value and states that it is efficient in bringing a lossless medical image compression.