More than only cross-correlation to align a raw tilt-series without fiducials, 3D reconstruction with iterative algorithms (SART, SIRT, OS-SART) using multi-GPUs, intuitive 3D/2D visualisation: DigiECT positions itself as a revolutionary alternative compared to existing commercial and free software.
New-generation electron tomography software commercially available
Shortly after its commercial release in May 2010, Digisens is pleased to announce the growing success of its novel electron tomography software DigiECT, which is the result of a two-year fruitful collaboration with Dr. Sergio Marco, Research Director Inserm U759, Institut Curie, Orsay, France. Moreover, DigiECT has been recently adopted by IPCMS-CNRS, a french nanotechnology institute based in Strasbourg, along with two other well-renowned research centers.
Using a TEM (Transmission Electron Microscope), electron tomography is an imaging technique for advanced 3D nanocharacterization in Life Sciences as well as Materials Sciences.
Requiring only a raw tilt-series at input, DigiECT is fully compliant with all TEM brands, all image acquisition modes (BF-TEM, BFS/TEM, DF-S/TEM or EFTEM) and schemes (linear, Saxton or user-defined). Its core philosophy relies in allowing reduced 3D reconstruction times, targeted to better image quality using iterative algorithms – best suited to cope with the missing wedge – while minimising manual interventions to save time and prevent human erros.
“We were using free software for 3D reconstruction and it was really time-consuming, even on the GPU. The alignment step has often to be corrected manually. With our recent investment in DigiECT, way faster 3D reconstruction speeds will bring us more time to achieve high-quality results” states Dr. Ovidiu Ersen, responsible of the electron tomography facility at IPCMS-CNRS.
Marker-less alignment, multi-GPUs 3D reconstruction and ergonomic visualisation
The primary step consists in automatically aligning the images without markers such as gold beads then numerically determining the location of the tilt axis in 3D (x, y, z). “This process is more than a rudimentary cross-correlation step, in order to avoid any drifts which are likely to occur along the stack: we basically track virtual markers” explains Dr. Jean-Pierre Bruandet, Digisens’ Chief Technical Officer.
As everyone knows, perfect alignment is mandarory prior to the subsequent 3D reconstruction phase. As for this, unique advantages of DigiECT’s 3D reconstruction module are numerous.
With Digisens’ exclusive multi-GPUs technology, DigiECT is first able to make use – at the same time – of several GPU boards (Graphical Processing Units) mounted in parallel, to split and resolve the 3D reconstruction process with unparalleled speeds. For instance, a 2048 x 2048 x 512 voxels volume (8.0 GB) is resolved in 31.2 seconds per iteration (SIRT) – including I/O – using 8 GPUs (Nvidia C2050). “With the recent advent of high-resolution cameras of 16 megapixels and beyond, we now encounter bottlenecks during the 3D reconstruction process. DigiECT overcomes this issue, offering incredible processing speeds” explains Dr. Marco.
Second, users can reach optimal convergence of the 3D reconstruction since they can adjust if desired the convergence speed, a major parameter even more important than the number of iterations when using iterative algorithms. Third, DigiECT proposes to 3D reconstruct a volume using OS-SART algorithm (Ordered-Subset Simultaneous Algebraic Reconstruction Technique) to get the best possible balance between the resolution of SART and noise robustness of SIRT.
Following both alignment and 3D reconstruction stages, end-users benefit from an intuitive, user-friendly visualisation tool, featuring uppermost quality volume and isosurface rendering, high-resolution oriented cut planes extraction, advanced measurements, certified surface extraction and seamless data import or export in numerous file formats.