Investigation of measurement errors in microwave imaging system for brain stroke monitoring - Laboratoire Jacques-Louis Lions Accéder directement au contenu
Communication Dans Un Congrès Année : 2017

Investigation of measurement errors in microwave imaging system for brain stroke monitoring

Résumé

Interest in Microwave imaging for medical applications has grown rapidly over the past decade. In this paper, we are dealing with the detection and monitoring of cerebrovascular accidents. Brain strokes are of two types: ischemic (about 85%) and hemorrhagic (15%). They are both sensitive to the variation of the complex permittivity in the microwave domain, with an increase of 10% in case of a hemorrhagic stroke and a decrease of 10% in case of an ischemic one. Taking into account the possibility of building a compact and movable/portable device, the Microwave Imaging System (MImS) turns out to be a good candidate for this application. X-ray CT and MRI are actually "gold standards" but they cannot be used for treatment monitoring. In addition, repeated X-ray scans cannot be performed, as X-ray radiation itself is not safe. A bedside stroke monitoring during treatment is another potential advantage of using a MImS in hospital ICUs (Intensive Care Units). Nevertheless, the delivery of tomographic images with a MImS is not straightforward. The problem is twofold: - First, we have to discriminate the nature of the stroke. For this, we must develop algorithms of quantitative imaging since we are looking for positive or negative variations in the permittivity values. Not to be neglected is the sensitivity of the algorithm to the noise occurring during the measurements. For this, a model of noise coming from measurements must be developed. Additionally, the brain is highly heterogeneous, as well in the values of the complex permittivity, as in the dimensions of the zones that compose it. This is a real problem for the EM-modeling that will be used as direct solver in the inversion algorithm. - Second, we must monitor the treatment. At this stage a qualitative imaging might be sufficient because we primary need observe the changes of its contours over time. The paper focuses the modeling of measurements errors and the two main issues to be addressed by a MImS for brain stroke detection and monitoring: - The detection and stroke’s discrimination between ischemic and hemorrhagic. - The monitoring of the stroke. A simple method is proposed for the monitoring that turns out to be robust to the noise. We have succeeded in the reconstruction of the contour of a hemorrhagic stroke over the time under an SNR of 9 dB. The SNR value is calculated with an error model derived from experimental data. This altogether makes both the model and the method realistic with respect to the application. An important part of the work beyond these results lies in the ability to have a powerful EM-solver. This part will not be developed in details but it is worth mentioning that a Finite Element Domain Specific Language (DSL) FreeFem++ solver for the EM modeling has been used with specific developments for the application. In addition, high parallel computing with Domain Decomposition is implemented to handle the simulations on a supercomputer for speeding up the process. Simulations of the whole brain are obtained within tens of seconds.
Fichier non déposé

Dates et versions

hal-01700670 , version 1 (05-02-2018)

Identifiants

  • HAL Id : hal-01700670 , version 1

Citer

Christian Pichot, Ibtissam El Kanfoud, Iannis Aliferis, Claire Migliaccio, Victorita Dolean, et al.. Investigation of measurement errors in microwave imaging system for brain stroke monitoring. Progress In Electromagnetics Research Symposium (PIERS), May 2017, Saint Petersburg France. pp.1-4. ⟨hal-01700670⟩
434 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More