Discrete cosine transform–based compressive sensing recovery strategies in medical imaging

Abstract In medical imaging, it is crucial to reduce the exposure time of the patient to the medical modality. Medical image compression has a significant role, including telemedicine, medical imaging, and video conferencing for consultation. One of the effective compression techniques, compressive sensing (CS), defined as a hypothesis to represent the information at a rate below Nyquist-Shannon sampling rate for sparse and incoherence images/signals in a particular domain. CS effectively reduces the sampling rate without important information loses and breaks the canonical rules. Thus it has broad applications in the medical imaging for medical image compression with high image reconstructions quality and less data loss. Basically, the image is converted to the discrete cosine transform, for example, with the relief of sensing matrix to extract the essential coefficients having less dimensionality compared to the image dimensions. CS recovery strategies can be divided into greedy recovery and L1 minimization techniques, where the first depends on iteration calculations of the image’s coefficients till realizing the convergence criterion, while the second technique depends on solving a linear optimization problem. In the present chapter, several recovery strategies are applied for image reconstruction, namely l1-magic, orthogonal matching pursuit, compressive sampling matching pursuit, and CVX. Then, the performance of these techniques is studied by measuring the peak signal-to-noise ratio and the structural similarity index measure as a reconstruction quality metric. Furthermore, the impact of CS on medical imaging applicators is also deliberated. The results established the superiority of the weighted-based CVX compared to the traditional as well as the weighted-based other recovery methods.