IceLuva: A scripting framework for MR image reconstruction based on free software

Abstract

Image reconstruction is an essential part of the magnetic resonance imaging process, and a whole field of research is dedicated to the development of reconstruction algorithms. For this reason, scanner manufacturers provide researchers with programming frameworks that give full control over the whole procedure. The drawback is that these environments are complex, and the code is non-portable and covered by non-disclosure agreements. In this article, a simplified framework based on a free scripting language (Lua) is presented. It is oriented to the development of postprocessing algorithms that are seamlessly integrated with the pipeline of a commercial scanner. The structure privileges simplicity over performance, to be quickly learned and used by researchers and students who might not be acquainted with low-level programming languages. Common postprocessing algorithms (contrast modulation, pixelwise fitting, phase-contrast imaging) could be implemented with 100 logical lines of code or less, using a syntax that is similar to the Matlab programming language. The average performance of the reconstruction was lower with respect to the native implementation, but superior to offline postprocessing on a desktop computer, without the bottleneck of offline data export.

Pulmonary relaxometry with inversion recovery ultra‐fast steady‐state free precession at 1.5T

“…Postprocessing and fitting of the transient ufSSFP data was performed directly on the scanner using IceLuva scripting framework . Images acquired in the diastolic phase were selected for further processing based on the time after ECG triggering.…”

Combination of a multimode antenna and TIAMO for traveling‐wave imaging at 9.4 Tesla

“…A 2D FLASH sequence (TR/TE = 200/4.5 ms, average FA ≈ 8°, FOV = 230 × 230 mm 2 , matrix 192 × 192, slice thickness = 2.5 mm, BW = 300 Hz/pixel, two averages, acquisition time = 2:33 min per orientation) was implemented that acquired each line in k‐space twice, once with each of the two complementary RF shim configurations calculated in situ. The RSS combination of the resulting images sampled with the six “virtual” receive channels (two acquisitions × three receive channels) was performed directly by the scanner software using the freely available reconstruction framework IceLuva .…”

High‐resolution Fourier‐encoded sub‐millisecond echo time musculoskeletal imaging at 3 Tesla and 7 Tesla

“…The minimum asymmetry allowed by the algorithm was 8% (a value of 100% indicates full sampling before the k‐space center). Online calculation of the image difference was performed (i.e., subtraction of the TE 2 image from the TE 1 image) in case of dual contrast acquisitions using a scripting framework .…”