Adaptive sparse recovery of medical images with variational approach – preliminary study for CT stroke

Abstract

Presented research was directed to effective signal recovery problem for computer-aided medical diagnosis. Extracted and visualized information covered in sensed data of imaging systems supports interpretation according to ”second look” procedure. The integrated framework of compressive sensing was used to optimize CT acute stroke diagnosis. Previously studied nonlinear approximation of the sparse signals in adjusted dictionaries was extended with variational approach to extract more precisely the content components. Proposed methodology adjusts optimized fidelity norms and regularizing priors to semantic question of image-based diagnosis. Preliminary experimental study was performed to provide selected proof-of-concept results for designed CT hypodensity extractors.