VolRec: 4D real-time volumetric reconstruction of OCT data

摘要

Abstract This work presents the design, analysis and development of VolRec, a volumetric 4D reconstruction system for OCT (optical coherence tomography) data. VolRec is able to obtain and process more than 25 high-resolution OCT data volumes per second. This way, VolRec allows OCT systems to perform real-time measurements that could help doctors and physicians in many different cases, e.g., during surgical interventions or during exams to restless patients, such as kids. Moreover, VolRec allows to adapt to telemedicine techniques such as telesurgery or image-guided medical robots. To achieve real-time performance, it is necessary to address the limitations of current OCT systems, which are restricted by both the acquisition and processing of massive amounts of data. To overcome those limitations, this work proposes an approach based on a 4D reconstruction. Our proposal allows to generate high-resolution volumes from acquired low-resolution ones. Hence, increasing the acquisition speed while reducing the acquisition resolution without sacrificing image quality. To this end, we use parallel programming mechanisms such as OpenMP (for CPUs) and CUDA to exploit the computing capabilities of modern GPUs. Our real-time volumetric reconstruction algorithm efficiently achieves a very high performance, reaching a processing rate of 18 GigaVoxels/s and about 72 $$\times$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mo>×</mml:mo> </mml:math> speedup over a parallel CPU-based algorithm using a GPU, by efficiently exploiting the vast amount of data-level parallelism inherent to OCT data volumes.