Fast Volume Reconstruction From Motion Corrupted Stacks of 2D Slices

Abstract

Capturing an enclosing volume of moving subjects and organs using fast individual image slice acquisition has shown promise in dealing with motion artefacts. Motion between slice acquisitions results in spatial inconsistencies that can be resolved by slice-to-volume reconstruction (SVR) methods to provide high quality 3D image data. Existing algorithms are, however, typically very slow, specialised to specific applications and rely on approximations, which impedes their potential clinical use. In this paper, we present a fast multi-GPU accelerated framework for slice-to-volume reconstruction. It is based on optimised 2D/3D registration, super-resolution with automatic outlier rejection and an additional (optional) intensity bias correction. We introduce a novel and fully automatic procedure for selecting the image stack with least motion to serve as an initial registration target. We evaluate the proposed method using artificial motion corrupted phantom data as well as clinical data, including tracked freehand ultrasound of the liver and fetal Magnetic Resonance Imaging. We achieve speed-up factors greater than 30 compared to a single CPU system and greater than 10 compared to currently available state-of-the-art multi-core CPU methods. We ensure high reconstruction accuracy by exact computation of the point-spread function for every input data point, which has not previously been possible due to computational limitations. Our framework and its implementation is scalable for available computational infrastructures and tests show a speed-up factor of 1.70 for each additional GPU. This paves the way for the online application of image based reconstruction methods during clinical examinations. The source code for the proposed approach is publicly available.

Point-Spread-Function-Aware Slice-to-Volume Registration: Application to Upper Abdominal MRI Super-Resolution

“…The reconstructions were initially quantitatively assessed by evaluating the residual via a normalized cross correlation metric, instead of the 2 -norm, in order to be insensitive to the intensity normalization step used to compensate for the different acquisition protocols. Following this, subjective assessment in a clinical context was made including direct comparison to reconstructions obtained by open-source toolkits successfully employed in the challenging problem of fetal MRI reconstructions (BTKtoolkit [11], version from 6 Jan 2016, and the IRTK-based toolkit [7], version from 11 Jun 2015). Two radiologists, blinded to the reconstruction methods, individually assessed reconstruction side-by-side and in comparison to the original HASTE data.…”

“…In this context, ultra-fast 2D MRI is the method of choice for many applications [7,13,15]. However, a balance has to be struck between a short scanning time to avoid motion artefacts, and the signal-to-noise ratio which must be maintained at an acceptable level.…”

“…Its application ranges from adult studies on the tongue [15] and thorax [13] to fetal applications [7,10]. To our knowledge, Super-Resolution (SR) has not yet been applied to MRCP studies to define upper gastrointestinal tract and biliary anatomy.…”

“…Existing respiratory motion models require the availability of respiratory surrogate data [8] which are currently not available for MRCP studies. Using an SRR approach such as the iterative two-step registration-reconstruction approach used in fetal MRI [7,10], applied to only two stacks, is prone to generate a strongly biased volume and the currently used rigid motion models might not be sufficient. Additionally, current motion correction techniques do not take into account the PSF for registration.…”

Learning clinically useful information from images: Past, present and future

“…Traditionally, fetal MRI is often limited to the acquisition of 2D images of the 3D fetal anatomy to avoid motion artefacts. We have pioneered a number of approaches that use image registration to compensate for fetal motion during the acquisition [Kainz et al (2015b)]. These approaches not only compensate for motion but exploit multiple motion-corrupted acquisitions in order to reconstruct high-resolution, super-resolved 3D volumes of fetal organs such as the brain, lungs and the cardiac vasculature.…”