Supercomputing Technologies Research Group SC@FIT

Hardware topics

• Architecture of parallel systems
• Performance and scaling models of parallel systems
• HPC systems based on GPUs and other accelerators
• Low power systems, power-performance optimization

Software topics

• Domain decomposition for spectral methods
• Automatic load balancing on heterogeneous systems
• Multi-scale and multi-physics model coupling
• Asynchronous input/output in HPC applications
• System for managing and monitoring supercomputers
• Modern programming languages in HPC (Python, OpenACC, Thrust)

Applications

• Simulation of ultrasound propagation in human body
• Acceleration of photoacoustic imaging
• Design of ultrasound sensors
• High performance scientific applications

Currently, the group is primarily focused on the CITRUS project, which aims at precise ultrasound neurostimulation. As part of this project, they are developing techniques to create accurate surgical plans based on the patient’s physiological data, and they perform the necessary computations on IT4Innovations supercomputers. Additionally, the group is working on real-time control of the ultrasound beam with the capability to fine-tune its focus on specific targets.

The group has also contributed to and continues collaborating with the PAMMOTH consortium, which focuses on photoacoustic imaging of breast tissue. Within this project, we handle data collection, real-time quality assessment of captured data, and subsequent processing on a supercomputer. For these purposes, we utilize a massively parallel implementation of the k-Wave toolbox based on local decomposition of spectral methods, with our proprietary k-Dispatch system managing the automation and monitoring of the calculations.

The group is also engaged in accelerating the k-Wave toolbox through advanced numerical methods, including adaptive decomposition and reduced Fourier transform. In addition, it develops techniques for accelerating scientific algorithms on graphic workstations and explores possibilities for incorporating artificial intelligence to generate ultrasound planning previews, particularly for neurostimulation applications.

 

• Prof Hlavička Award for the best presentation at PhD conference  PAD 2019 - Filip Vaverka
• Excellent Poster Presentation Award - Signal Frequency Content and Appearance of Tumors in Photoacoustic Breast Tomography: A Simulation by Maura Dantuma, Felix Lucka, Bradley Treeby, Jiri Jaros, Ben Cox, and Srirang Manohar -  Ultrasonics 2018, Caparica, Portugal.
• Research Poster Award - GPU-accelerated Simulation of Elastic Wave Propagation by  Kristian Kadlubiak, Jiří Jaroš, and Bradley Treeby - HPC applications, ISC 2018, Frankfurt, Germany.
• PRACE HPC Ambassador Award, Marta Čudová, Summer Of HPC, 2016.
• Best paper award, BREITENBACHER Dominik, HOMOLIAK Ivan, JAROŠ Jiří and HANÁČEK Petr. Impact of Optimization and Parallelism on Factorization Speed of SIQS. In: Proceedings of The 20th World Multi-Conference on Systemics, Cybernetics and Informatics. Orlando: The International Institute of Informatics and Systemics, 2016, pp. 55-62. ISBN 978-1-941763-41-4.

Software

• k-Wave toolbox - 20,000+ registered users

Publications

• Modeling nonlinear ultrasound propagation in heterogeneous media with power law absorption using a k-space pseudospectral method - Journal paper, JASA (110 citations)
• Parallel Genetic Algorithm on the CUDA Architecture - Conference paper, EvoStar (50 citations)
• Parallel Genetic Algorithm Solving 0/1 Knapsack Problem Running on the GPU - Conference paper, Mendel (15 citations)

Conferences

• Supercomputing - every year attendance at the best conference in the field, poster 2017, poster 2018
• ISC HPC - every year attendance at the best European conference in the field, poster 2018

PRACE Summer of HPC

• Martin Stodůlka, ULux-University of Luxembourg, Lux, 2021
• Filip Kukliš, SURFsara, Amsterdam, NL, 2018
• Petr Stehlík , CINECA, Bologna, It, 2017
• Marta Čudová , EPCC, Edinburgh, UK, 2016
• Ondřej Vysocký , EPCC, Edinburgh, UK, 2015
• Vojtech Nikl, CINECA, Bologna, It, 2014

Computing resources

• IT4Innovation, CZ, 17 milion core hours
• National Computing Infrastructure, AU, 1 million core hours
• PRACE, EU, 1 million core hours
• Swiss National Supercomputing Centre, CH, 24 thousand core hours

Collaboration offer

The SC@FIT research group offers collaboration in the design and development of applications demanding high compute power, large operational memory and big data. We have deep experience in the R&D of scientific codes for shared memory systems (high-end servers), distributed systems (clusters) and hybrid systems with processors and graphics processing units.

The group also offers the application of the k-Wave toolbox for the nonlinear simulation of ultrasound wave propagation in medical and industrial problems.

Current Collaboration

• Biomedical Ultrasound Group, University College London, UK - development of the simulation toolbox k-Wave for modelling ultrasound propagation in human body
• Institute of Biomedical Engineering, Ludwig Cancer Research, University of Oxford, UK - Kidney cancer treatment planning
• Turku High-Intensity Focused Ultrasound (HIFU) Research Centre, Turku, Finland - Ultrasound Prostate cancer treatment
• IT4Innovations národní superpočítačové centrum, VŠB - Technical University Ostrava, CZ - Research and development on the Anselm and Salomon supercomputer, collaboration in the READEX project
• The University of Twente, Twente, NL - acceleration of the photoacoustic imaging of breast (PAMMTOH project)
• University of Bern, Bern, CH - acceleration of the photoacoustic imaging of breast (PAMMTOH project)
• Medisch Spectrum Twente, Twente, NL - acceleration of the photoacoustic imaging of breast (PAMMTOH project)
• Imasonic SAS, Voray sur l'Ognon, FR -acceleration of the photoacoustic imaging of breast (PAMMTOH project)
• EKSPLA, Vilnius, Lt - acceleration of the photoacoustic imaging of breast (PAMMTOH project)
• Continental, Ostrava - Research and development of  ultrasound sensors
• Faculty of Civil Engineering, Brno University of Technology, CZ - Simulation of fraction tests in quasi-brittle materials

Currently solved BP and DP theses:

• Using the noarr library to implement efficient algorithms
• Automatic tests for k-Dispatch system
• Integration of Amazon Compute Cloud computing resources into the k-Dispatch system
• Design and implementation of a new version of the k-Dispatch REST API
• Development of a web interface for monitoring compute clusters using the Slurm REST API
• Acceleration of acoustic propagator on GPU
• Acceleration of a backprojection algorithm to simulate ultrasound propagation
• Supercomputer dashboard with monitoring of current workload
• Library for storing ultrasound simulation data

• k-Wave toolbox
• PAMMOTH
• CITRUS project

Brno University of Technology
Faculty of Information Technology
Department of Computer Systems
Bozetechova 2, Room L334
612 00 Brno, Czech Republic

Phone: +420 54114-1207
E-mail: jarosjir@fit.vutbr.cz

Brno University of Technology
Faculty of Information Technology
Department of Computer Systems
Bozetechova 2
612 00 Brno, Czech Republic

Phone: +420 54114-1207
E-mail: jarosjir@fit.vutbr.cz