Inclusion/Flaw Detection in Ultrasound Imaging Through Optimization of Random Transmitted Wave

Sébastien Ménigot (1) , Maroun Geryes (1) , Jean Marc Girault (1)

1 iBraiN - Imaging, Brain & Neuropsychiatry

Nowadays, ultrasound imaging is become an essential tool for diagnosis in medicine and in industry. Developments in ultrasound imaging have provided more sensitive imaging system to increase the contrast and the resolution. The main improvement has made possible by tak-ing into account the wave harmonic generation during the propagation in tissues. By extract-ing harmonic components, a higher contrast can be obtained. Several has been developed through post-processing first. However, to improve the resolution while maintaining the ad-vantage of the harmonic imaging, encoding imaging has been developed. One of the most used encoding techniques is the pulse inversion imaging. However, the imaging methods are usually non-optimal, since there is no optimization process included into the imaging system. In most cases, the solution adopted by manufacturers consists of providing empirically pre-set transmit frequencies, even if it is obvious that the medium to be explored should be taken into account during the optimization process. In a previous method, to overcome this problem without assumption on the optimal wave, we have modified a current imaging system by in-cluding feedback from output to input. To resolve the waveform optimization, we have pro-posed using the transmitted stochastic signal chosen by a genetic algorithm where the cost function was the contrast. However the contrast computation required some a priori knowledge about the position of the inclusion in tissue. In this study, to avoid this assumption on the medium, a medium with an inclusion was compared to a reference medium without defect. To distinguish the two media, we characterized energy for each medium. These energy features enabled to create an Euclidean distance between the two media. Our aim was to find automatically the stochastic command that maximized this Euclidean distance. The advantage of the method was that no a priori information was required in order to find the optimal stochastic command and the position of the inclusion. In simulation, by using the optimal stochastic command in pulse inversion imaging, the algorithm converged after 3000 iterations. The distance could be multiplied by 4 in comparison with the distance obtained with an usual excitation at the central frequency of the transducer.

Domaines: Traitement du signal et de l'image [eess.SP]; Acoustique [physics.class-ph]; Imagerie

Sébastien Ménigot, Maroun Geryes, Jean Marc Girault. Inclusion/Flaw Detection in Ultrasound Imaging Through Optimization of Random Transmitted Wave. Acoustics 2013, Nov 2013, New Delhi, India. ⟨hal-01279954⟩