Skip to Main content Skip to Navigation

Recalage et analyse d’un couple d’images : application aux mammographies

Abstract : In the scientific world, signal analysis and especially image analysis is a very active area, due to the variety of existing applications, with issues such as file compression, video surveillance or medical image analysis. This last area is particularly active. The number of existing devices and the number of pictures taken, cause the production of a large amount of information to be processed by practitioners. They can now be assisted by computers.In this thesis, the problem addressed is the development of a computer diagnostic aided system based on conjoint analysis, and therefore on the comparison of medical images. This approach allows to look for evolutions or aberrant tissues in a given set, rather than attempting to characterize, with a strong a priori, the type of fabric sought.This problem allows to apprehend an aspect of the analysis of medical file performed by experts which is the study of a case through the comparison of evolutions.This task is not easy to automate. The human eye performs quasi-automatically treatments that we need to replicate.Before comparing some region on the two images, we need to determine where this area is located on both pictures. Any automated comparison of signals requires a registration phase, an alignment of components present on the pictures, so that they occupy the same space on the two images. Although the characteristics of the processed images allow the development of a smart registration, the projection of a 3D reality onto a 2D image causes differences due to the orientation of the tissues observed, and will not allow to analyze a pair of shots with a simple difference between images. Different structuring of the pictures and different deformation fields are developed here to efficiently address the registration problem.After having minimized the differences on the pictures, the analysis of tissues evolution is not performed at pixels level, but the tissues themselves, as will an expert. To process the images in this logic, they will be reinterpreted, not as pixels of different brightness, but as patterns representative of the entire image, enabling a new decomposition of the pictures. The advantage of such a representation is that it allows to highlight another aspect of the signal, and analyze under a new perspective the information necessary to the diagnosis aid.This thesis has been carried out in the LIPADE laboratory of University Paris Descartes (SIP team, specialized in image analysis) and in collaboration with the Society Fenics (designer of diagnosis aid stations in the analysis of mammograms) under a Cifre convention. The convergence of the research fields of those teams led to the development of this document.
Document type :
Complete list of metadata

Cited literature [92 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, March 8, 2013 - 11:37:32 AM
Last modification on : Saturday, June 19, 2021 - 3:49:28 AM
Long-term archiving on: : Sunday, June 9, 2013 - 6:15:16 AM


Version validated by the jury (STAR)


  • HAL Id : tel-00798271, version 1


Arnaud Boucher. Recalage et analyse d’un couple d’images : application aux mammographies. Ordinateur et société [cs.CY]. Université René Descartes - Paris V, 2013. Français. ⟨NNT : 2013PA05S001⟩. ⟨tel-00798271⟩



Record views


Files downloads