pith. sign in

arxiv: 1906.11925 · v1 · pith:J3E7AKA6new · submitted 2019-06-27 · ⚛️ physics.comp-ph · cs.CV· cs.GR

HEMELB Acceleration and Visualization for Cerebral Aneurysms

Sahar Soheilian Esfahani , Xiaojun Zhai , Minsi Chen , Abbes Amira , Faycal Bensaali , Julien AbiNahed , Sarada Dakua , Georges Younes
show 2 more authors
Robin A. Richardson Peter V. Coveney
This is my paper

Pith reviewed 2026-05-25 13:38 UTC · model grok-4.3

classification ⚛️ physics.comp-ph cs.CVcs.GR
keywords cerebral aneurysmHemeLBGPU accelerationblood flow simulationlattice Boltzmannfluid dynamicsmedical visualizationclinical deployment
0
0 comments X

The pith

GPU implementation of HemeLB reaches 15 million site updates per second for cerebral aneurysm blood flow.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The authors developed a GPU version of the HemeLB fluid dynamics code together with a visualization platform to compute blood flow in and around cerebral aneurysms. They report that this version attains a peak performance of 15,168,964 site updates per second on cost-efficient hardware. The stated goal is to move the solver from research machines into hospital settings so that neurosurgeons can obtain simulation results quickly enough to influence diagnosis and treatment planning.

Core claim

The central claim is that the proposed GPU implementation of HemeLB together with its visualization tools achieves a maximum of 15,168,964 site updates per second and is therefore capable of speeding up the solver sufficiently for deployment in hospitals and clinical investigations of cerebral aneurysms.

What carries the argument

The GPU-accelerated HemeLB lattice-Boltzmann solver for blood flow in complex vessel geometries, coupled with an interactive visualization platform.

If this is right

  • Simulation results become available fast enough to influence real-time surgical planning.
  • HemeLB can run on affordable hardware already present in many hospitals.
  • Neurosurgeons gain an interactive visualization layer for exploring flow inside patient-specific aneurysms.
  • The same acceleration path can be applied to other lattice-Boltzmann vascular cases beyond aneurysms.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If accuracy is later verified, similar GPU ports could shorten turnaround for other patient-specific CFD problems in medicine.
  • Real-time capability might allow coupling the solver directly to incoming imaging data streams during procedures.
  • The reported performance number supplies a concrete benchmark that later implementations can target or exceed.

Load-bearing premise

The GPU version produces flow results whose accuracy matches the original CPU HemeLB closely enough for clinical decision making.

What would settle it

Side-by-side comparison of velocity fields, pressure, and wall shear stress between the CPU and GPU codes on the same patient geometry, showing differences larger than the tolerance accepted by neurosurgeons.

Figures

Figures reproduced from arXiv: 1906.11925 by Abbes Amira, Faycal Bensaali, Georges Younes, Julien AbiNahed, Minsi Chen, Peter V. Coveney, Robin A. Richardson, Sahar Soheilian Esfahani, Sarada Dakua, Xiaojun Zhai.

Figure 2
Figure 2. Figure 2: Simplified architecture of Exxact Tensor TWS-289059-DPN [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 1
Figure 1. Figure 1: Lattice node of D3Q19 model [7]. HemeLB is implemented with different Boundary Con￾ditions (BC) [7], such as Ladd iolets for velocity inlet BC [15] and Bouzidi-Firdaouss-Lallemand (BFL) for the interpolated wall collision BC [16]. By the use of topology￾aware two-level domain decomposition, HemeLB provides a good workload distribution for parallel implementation. In addition, the improvements made in HemeL… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization platform designed for real-time HemeLB results [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: The three-tier internal architecture of visualization client. [PITH_FULL_IMAGE:figures/full_fig_p003_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Results of HemeLB visualization framework. [PITH_FULL_IMAGE:figures/full_fig_p004_9.png] view at source ↗
Figure 7
Figure 7. Figure 7: STL files used for evaluation of visualized HemeLB. [PITH_FULL_IMAGE:figures/full_fig_p004_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The inlet blood velocity applied in each patient test case. [PITH_FULL_IMAGE:figures/full_fig_p004_8.png] view at source ↗
read the original abstract

A weakness in the wall of a cerebral artery causing a dilation or ballooning of the blood vessel is known as a cerebral aneurysm. Optimal treatment requires fast and accurate diagnosis of the aneurysm. HemeLB is a fluid dynamics solver for complex geometries developed to provide neurosurgeons with information related to the flow of blood in and around aneurysms. On a cost efficient platform, HemeLB could be employed in hospitals to provide surgeons with the simulation results in real-time. In this work, we developed an improved version of HemeLB for GPU implementation and result visualization. A visualization platform for smooth interaction with end users is also presented. Finally, a comprehensive evaluation of this implementation is reported. The results demonstrate that the proposed implementation achieves a maximum performance of 15,168,964 site updates per second, and is capable of speeding up HemeLB for deployment in hospitals and clinical investigations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper presents a GPU-accelerated version of the HemeLB lattice Boltzmann solver for cerebral aneurysm blood-flow simulations, together with a visualization platform. It reports a peak throughput of 15,168,964 site updates per second and asserts that the implementation is suitable for real-time hospital deployment and clinical investigations.

Significance. A verified GPU port that preserves the original solver's numerical fidelity could enable real-time clinical use of HemeLB; the reported performance figure is concrete, but the lack of any accuracy or fidelity data means the clinical-readiness claim cannot be assessed from the manuscript.

major comments (2)
  1. [Abstract] Abstract: the assertion that the implementation 'is capable of speeding up HemeLB for deployment in hospitals and clinical investigations' rests on an untested assumption that the GPU results match the CPU reference to within clinical tolerances; no error norms, field comparisons, or benchmark runs against the original code are supplied.
  2. [Abstract] Abstract: the headline performance number (15,168,964 site updates per second) is given without error bars, without description of the test geometries or boundary conditions used, and without any statement of how the measurement was obtained or verified against the reference CPU implementation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on our manuscript. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion that the implementation 'is capable of speeding up HemeLB for deployment in hospitals and clinical investigations' rests on an untested assumption that the GPU results match the CPU reference to within clinical tolerances; no error norms, field comparisons, or benchmark runs against the original code are supplied.

    Authors: We agree that the abstract claim would be strengthened by explicit validation data. The GPU port implements the identical lattice Boltzmann algorithm and collision operator as the original CPU HemeLB, so numerical fidelity is preserved by construction. To address the referee's concern directly, we will add error norms, point-wise field comparisons, and benchmark runs against the CPU reference (including L2 and L-infinity norms on velocity and pressure) to the revised manuscript. revision: yes

  2. Referee: [Abstract] Abstract: the headline performance number (15,168,964 site updates per second) is given without error bars, without description of the test geometries or boundary conditions used, and without any statement of how the measurement was obtained or verified against the reference CPU implementation.

    Authors: The performance figure was measured on the aneurysm geometries and boundary conditions described in Section 4 of the full manuscript, using the standard site-update counting method already employed in prior HemeLB publications. We acknowledge that these details are not repeated in the abstract. In the revision we will expand the abstract to include a concise statement of the test cases, boundary conditions, and verification procedure. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical performance measurement only

full rationale

The paper reports an implementation of a GPU-accelerated HemeLB solver together with measured wall-clock performance (site updates per second) on specific hardware. No derivation, ansatz, fitted parameter, or uniqueness theorem is presented; the central claims rest on direct benchmarking rather than any chain that reduces to its own inputs by construction. Self-citations to the original HemeLB CPU code are ordinary references to prior independent work and do not carry load-bearing mathematical premises inside this manuscript.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper is a software engineering and performance report; it introduces no new physical parameters, axioms, or postulated entities beyond the standard assumptions already present in the original HemeLB lattice-Boltzmann solver.

axioms (1)
  • domain assumption Lattice Boltzmann method correctly approximates incompressible Navier-Stokes flow in the regimes relevant to cerebral blood flow
    HemeLB is built on LBM; the paper inherits this modeling choice without re-deriving or validating it.

pith-pipeline@v0.9.0 · 5727 in / 1234 out tokens · 25402 ms · 2026-05-25T13:38:46.384468+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    What you should know about cerebral aneurysms,

    R.T. Higashida, “What you should know about cerebral aneurysms,” Pamphlet. American Heart Association Cardio- vascular Council, 2003

    work page 2003

  2. [2]

    Stroke risk factors in a moroccan population: a multicentric prospective study,

    N. Benkirane, A. Bennis, S. Bellakhdar, R. Habbal, S. Aidi, M. El Alaoui Faris, H. El Otmani, B. El Moutawakil, M.A. Rafai, and I. Slassi, “Stroke risk factors in a moroccan population: a multicentric prospective study,” Journal of the Neurological Sciences, vol. 357, pp. e367, October 2015

    work page 2015

  3. [3]

    Development of three-dimensional hollow elastic model for cerebral aneurysm clipping sim- ulation enabling rapid and low cost prototyping,

    T. Mashiko, K. Otani, R. Kawano, T. Konno, N. Kaneko, Y . Ito, and E. Watanabe, “Development of three-dimensional hollow elastic model for cerebral aneurysm clipping sim- ulation enabling rapid and low cost prototyping,” World Neurosurgery, vol. 83, pp. 351–361, March 2015

    work page 2015

  4. [4]

    Brain aneurysm prognosis,

    Verywell health, “Brain aneurysm prognosis,” https://www.verywellhealth.com/brain-aneurysm-prognosis- 3146347, 2018, [Online; accessed 04 Jan. 2019]

    work page 2018

  5. [5]

    International subarachnoid aneurysm trial (isat) of neurosurgical clipping versus endovascular coiling in 2143 patients with ruptured intracranial aneurysms: a randomised trial,

    A. Molyneux, R. Kerr, International Subarachnoid Aneurysm Trial (ISAT) Collaborative Group, et al., “International subarachnoid aneurysm trial (isat) of neurosurgical clipping versus endovascular coiling in 2143 patients with ruptured intracranial aneurysms: a randomised trial,” Journal of stroke and cerebrovascular diseases , vol. 11, no. 6, pp. 304–314,...

    work page 2002

  6. [6]

    Stent application for the treatment of cerebral aneurysms,

    B.M. Kim, D.J. Kim, and D.I. Kim, “Stent application for the treatment of cerebral aneurysms,” Neurointervention, vol. 6, pp. 53–70, August 2011

    work page 2011

  7. [7]

    HemeLB: A high perfor- mance parallel lattice-Boltzmann code for large scale fluid flow in complex geometries,

    M.D. Mazzeo and P.V . Coveney, “HemeLB: A high perfor- mance parallel lattice-Boltzmann code for large scale fluid flow in complex geometries,” Computer Physics Communi- cations, vol. 178, pp. 894914, June 2008

    work page 2008

  8. [8]

    Zynq SoC based acceleration of the lattice-Boltzmann method,

    X. Zhai, A. Amira, F. Bensaali, A. Al-Shibani, A. Al-Nassr, A. El-Sayed, M. Eslami, S. Dakua, and J. Abinahed, “Zynq SoC based acceleration of the lattice-Boltzmann method,” Concurrency and Computation: Practice and Experience , p. e5184, 2019

    work page 2019

  9. [9]

    Analysing and modelling the performance of the HemeLB lattice-Boltzmann simulation environment,

    D. Groen, J. Hetherington, H.B. Carver, R.W. Nash, M.O. Bernabeu, and P.V . Coveney, “Analysing and modelling the performance of the HemeLB lattice-Boltzmann simulation environment,” Journal of Computational Science , vol. 4, pp. 412–422, September 2013

    work page 2013

  10. [10]

    Modelling patient-specific magnetic drug targeting within the intracranial vasculature,

    A. Patronis, R.A. Richardson, S. Schmieschek, B.J.N Wylie, R.W. Nash, and P.V . Coveney, “Modelling patient-specific magnetic drug targeting within the intracranial vasculature,” Frontiers in physiology, vol. 9, pp. 331, April 2018

    work page 2018

  11. [11]

    Validation of patient-specific cerebral blood flow simulation using transcra- nial doppler measurements,

    D. Groen, R.A. Richardson, R. Coy, U.D. Schiller, H. Chan- drashekar, F. Robertson, and P.V . Coveney, “Validation of patient-specific cerebral blood flow simulation using transcra- nial doppler measurements,” Frontiers in Physiology, vol. 9, pp. 721, June 2018

    work page 2018

  12. [12]

    Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures,

    M.A. Suchard, Q. Wang, C. Chan, J. Frelinger, A. Cron, and M. West, “Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures,” Journal of computational and graphical statistics, vol. 19, no. 2, pp. 419–438, January 2010

    work page 2010

  13. [13]

    Real-time automated image segmentation technique for cerebral aneurysm on reconfigurable system-on-chip,

    X. Zhai, M. Eslami, E.S. Hussein, M.S. Filali, S.T. Shalaby, A. Amira, F. Bensaali, S. Dakua, J. Abinahed, A. Al-Ansari, et al., “Real-time automated image segmentation technique for cerebral aneurysm on reconfigurable system-on-chip,” Journal of Computational Science , vol. 27, pp. 35–45, July 2018

    work page 2018

  14. [14]

    System-on-chip solution for patients biometric: A compressive sensing-based approach,

    Hamza Djelouat, Xiaojun Zhai, Mohamed Al Disi, Abbes Amira, and Faycal Bensaali, “System-on-chip solution for patients biometric: A compressive sensing-based approach,” IEEE Sensors Journal, vol. 18, no. 23, pp. 9629–9639, 2018

    work page 2018

  15. [15]

    Numerical simulations of particulate suspen- sions via a discretized Boltzmann equation. part 1. theoretical foundation,

    A.J.C. Ladd, “Numerical simulations of particulate suspen- sions via a discretized Boltzmann equation. part 1. theoretical foundation,” Journal of fluid mechanics , vol. 271, pp. 285– 309, July 1994

    work page 1994

  16. [16]

    Momentum transfer of a lattice-Boltzmann fluid with boundaries,

    M. Bouzidi, M. Firdaouss, and P. Lallemand, “Momentum transfer of a lattice-Boltzmann fluid with boundaries,” Physics of fluids, vol. 13, pp. 3452–3459, November 2001

    work page 2001

  17. [17]

    Exxact Tensor TWS-289059-DPN spec- ification,

    Exxact Corporation, “Exxact Tensor TWS-289059-DPN spec- ification,” https://www.exxactcorp.com/Exxact-TWS-289059- DPN-E289059, 2018, [Online; accessed 25 Dec. 2018]

    work page 2018

  18. [18]

    HemeLB Github repository,

    UCL, “HemeLB Github repository,” https://github.com/UCL/HemeLB, 2018, [Online; accessed 10 Oct. 2018]

    work page 2018

  19. [19]

    Real-time smoke rendering using compensated ray march- ing,

    K. Zhou, Z. Ren, S. Lin, H. Bao, B. Guo, and H.Y . Shum, “Real-time smoke rendering using compensated ray march- ing,” in ACM Transactions on Graphics (TOG). ACM, August 2008, vol. 27, p. 36

    work page 2008