The VOXEL project has now ended, and in the context of a FET Seminar held in December 2019, we asked Marta to reflect on the idea behind the project, the challenges and the surprises that her team had to face, and the value that this experience has brought to her career.
Marta, what is the problem Voxel looks at, and which is the proposed solution?
Anyone who has undergone a CAT scan knows how long and uncomfortable it can be. The reason one needs to stand still for such a long and loud stretch of time, is that hundreds of X-ray images, from many different viewing angles, are being taken. They are required for accurate computer reconstruction, to synthesize all those views into a 3D digital image.
Computerized Axial Tomography (CAT) has been a life-saving technology, far outweighing its cost of multiple X-ray irradiations. However, beyond the discomfort and time, patients and physicians worry that the ionizing radiation dose received per scan is very high. In VOXEL project we have made a proof of principle alternative to X-ray tomography, enabling lower dose of ionizing radiation, which, because it doesn’t require taking radiographies from multiple views, result in 100’s less irradiations and much quicker scans.
What is the idea behind VOXEL?
Our disruptive technology is called plenoptic imaging, a poorly understood type of camera that has been shown to be able to capture 3D information in a single view. With our FET project we wanted to make the leap from visible light to X-rays. The ultimate goal would be to perform medical imaging. However even in the visible, these cameras have demonstrated their capacity for 3D microscopy, and X-ray microscopy of small animals is an extremely important field where many discoveries are made… but typically require going to a facility called synchrotron with the sample. Our concept would allow tabletop 3D X-ray microscopy. This is actually what made us come up with the idea: we needed a way to capture 3D images of objects (in our case plasmas) that were changing so fast, that we could not afford to turn an x-ray source around the object, taking pictures, before it was gone.
The basic camera consists of a main imaging optic, followed by a large array of small lenses tightly packed in squares or hexagons, and a high-resolution detector. These optical elements are all readily available in the visible range… but not for X-rays.
What was the main challenges you faced?
Optics for X-rays are very challenging to produce, so the project started with a large set of imperfect solutions.
To engineer the camera, then, we had to understand it completely first. A picture taken from visible light, reflecting off most surfaces, is very different from an X-ray image, where everything is (almost) transparent. Our applied mathematicians were critical to point us to the right direction: an image from our camera would be equivalent to a subset of the irradiations from multiple angles required for CAT, with very high resolution. Once this equivalence was established, we could make two numerical models of the camera from different principles (ray tracing and light field propagation), to test the different available X-ray lenses, and design three prototypes in visible, soft X-rays and hard X-rays.
Which surprises occurred throughout the project?
While X-ray optics were being designed, we had activated a backup plan, using a technique called “lensless imaging”. In plenoptic imaging, the depth resolution is limited by the number of sub-images, one per lens in the large array. To see how far we could go and still recover 3D reconstruction, we performed a very simple experiment with the lensless imaging technique, using soft X-rays. Thanks to machine vision algorithms, used for 3D rendering of 3D views (as in the 3D movies) we achieved extremely high resolutions with just one acquisition, and just two sub-images. This type of resolution is only possible with coherent X-rays, though, and for commercial X-ray sources the optics are still needed.
On the other hand, one bad surprise was our inability to turn on the X-ray source in our home-made prototype, because ionizing radiation regulations are very strict in France, and required the design of a complex shielding and a certification from the national authorities. At this point we had full confidence in our models, because we had built an X-ray emulator which gave great results (shown in the main picture above). We took our optics to a synchrotron, DESY, and made the proof of principle there. While at DESY, we tested a series of different optics, to check the validity our model. The tests worked as designed - X-ray plenoptic image has the potential for capturing fast, low dose images with depth information even in a single view.
This last experiment at DESY was made in the last month of the project. What if it hadn’t worked?
The premise of the FET projects is that they are cutting-edge, and therefore potentially subject to failure. However, we had gathered enough evidence to know that it would work, and we were all invested in establishing the final proof. This team included not only the researchers, but the Project Officer and the project reviewers. The review team was instrumental in making the project a success, as they were steering us in the right direction in a very trusting and encouraging way. It was a lucky bet, based on an educated guess.
How did FET contribute to your career and what is the main value of the programme?
I can largely credit being coordinator of VOXEL for my new long-term professorship at Instituto Superior Técnico. It also has allowed me to build my team, hiring gifted researchers. The investment made in the local infrastructure meant we are now equipped with state-of-the-art ultra-fast X-rays. The VOXEL station is almost mandatory visit for people coming to our University, from VIPs to prospective students. The buzz around our work has piqued the interest of many bright, creative students who will continue using the VOXEL station for years to come.
The soft X-ray prototype at Laboratoire d’Optique Appliquée
Background information
FET-Open and FET Proactive are now part of the Enhanced European Innovation Council (EIC) Pilot (specifically the Pathfinder), the new home for deep-tech research and innovation in Horizon 2020, the EU funding programme for research and innovation.