Modelization and multi modality simulation of medical data

I currently work for VIP project about modelization and multi-modality simulation of medical data.

Photoacoustic imaging of biological tissues

Photoacoustic, optoacoustic (light) or thermoacoustic (RF) effect refers to generation of acoustic waves from an object illuminated by a pulsed or modulated electromagnetic radiation. This physical phenomenon was discovered in 1880 byr Graham Bell who demonstrated emission of sound by illuminating an object with sunlight.

Researches about this phenomenon were not very intensive until the 70's, due to the lack of appropriate light sources. New researches then focused on using PA effect to estimate gaz concentration. After, several applications appeared, notably in chemistry, physics, engineering and medicine. Recently, more and more works for biomedical applications were presented.

Important interest for photoacoustic imaging is due to the fact that it presents advantages of optical and ultrasound imaging. Indeed, it is based on the high contrast coming from optical absorption and allows a larger penetration depth coming from lower attenuation of ultrasound waves in tissues.

During my first post-doctoral year, I was intesrested in this new medical imaging modality, interfacing optical and ultrasound imaging. The main objective was to propose a system allowing imaging of biological tissues. After a large review of existing works, we performed several exprimentations with femtosecond and nanosecond laser. Below is presented an illustration of one of our experimentation.

Motion estimation in ultrasound imaging

PhD Title:"Spatiotemporal oriented filtering of image sequences: application to blood flow motion estimation in ultrasound imaging"

Background : Medical ultrasonic imaging is a real-time modality that allows the tracking of anatomical structures along the time. By using Doppler mode, it is possible to estimate blood flow velocities. However, Doppler techniques suffers from a number of limitations:
- it estimates only the apparent velocity in the imaging plane
- it is necessary to know the angle between the echographic probe and the flow
- the use of a narrow-band emission reduces the spatial resolution

In order to reduce these limitations, we developed at CREATIS alternative methods to estimate blood flow velocity. During my PhD thesis, I was particularly interested in to spatiotemporal approaches considering temporal sequences as 3D data volumes (2D+t).

Models and methods : A temporal sequence of 2D images can be seen as a 2D+t spatiotemporal volume. A sequence in translation leaves a trace in this volume as we can see it below.

         

The vessel represented into the imaging plane (x,y) is delimited by the white edges and consisted of moving particles. The plane (x,t) contains an oriented texture and its orientation is related to the velocity of the particles.
Velocity is local and consequently we have to estimate orientation in each pixel of the sequence. In [1], we proposed to estimate orientation with a bank of spatiotemporal oriented filters. A simplified bank of oriented filters represented in the Fourier domain is presented above.


Results :
Our method was validated with a large set of sequences, simulated and experimental. We can see below results obtained with our approach as a dense field of velocity vectors and a parabolic velocities profile.

         


Below is a high-frequency ultrasound imaging sequence of a fluid within a phantom. Calibrated velocity is v=0.8mm/s.