Digital processing of medical images: application in synthetic cardiac datasets using the CRISP_DM methodology

Abstract

In this work an adaptation of the Cross IndustryStandard Process for Data Mining (CRISP-DM) methodology,in the context of digital medical imageprocessing is proposed. Specifically, synthetic images reportedin the literature are used as numerical phantoms.Construction of the synthetic images was inspired by a detailedanalysis of some of the imperfections found in thereal multilayer cardiac computed tomography images. Ofall the imperfections considered, only Poisson noise wasselected and incorporated into a synthetic database. Anexample is presented in which images contaminated withPoisson noise are processed and then subject to two classicaldigital smoothing techniques, identified as Gaussianfilter and anisotropic diffusion filter. Additionally, the peakof the signal-to-noise ratio (PSNR) is considered as a metricto analyze the performance of these filters.