Review Article

A Systematic Review of Real-Time Medical Simulations with Soft-Tissue Deformation: Computational Approaches, Interaction Devices, System Architectures, and Clinical Validations

Table 6

Classification of developed modelling methods for soft tissue deformations in real time: meshfree-based techniques.

ReferenceApproachModelling methodsSoft-tissue typesTissue BehaviorsComputation time/speedGeometry discretizationHardware configurations

Nedel and Thalmann [38]MDMass-spring system method (MSM)The muscleLinear elasticity16 FPS (model deformation)
84 FPS (model deformation)
82 mass points
17 mass points
SGI Impact workstation, MIPS R10000 CPU

Monserrat et al. [4]MDBoundary element method (BEM)The general cube meshLinear elasticity15 Hz (model deformation)<150 N
Surface triangle elements
R-4400 CPU, 64 MB RAM

Goto and Lee [39]MDStatistical analysis method (SAM)-muscleThe human faceNI1 minute (facial feature detection)NIPentium II, 333 MHz CPU

Bonamico et al. [40]MDMesh geometry VRML-like representation (VRML) & radial basis function (RBF)-muscleThe human faceLinear elasticity475 ms (facial deformation)
1,430 ms (facial deformation)
1,253 V-2,444 F
4,152 V-8,126 F
Pentium II 450 MHz CPU, 128 MB RAM

Brown et al. [2]MDMass-spring system method (MSM)The blood vesselNonlinear viscoelasticity24 FPS (system iteration)
6 FPS (system iteration)
216 N-1,440 E
Surface triangle elements
8,000 N-66,120 E
Surface triangle elements
Sun Ultra 60 Workstation 450 MHz CPU, 1 GB RAM

Sorkine et al. [41]MDLaplacian surface deformation (LSD)The face modelLinear elasticity0.07 s (model solving)10,000 V
Surface triangle elements
2.0 GHz Pentium IV CPU

Chandrasiri et al. [42]MDPersonal facial expression space method (PEES)-muscleThe human faceLinear elasticity12 FPS (facial animation)NI1 GHz Athlon CPU

Mollemans et al. [43]MDMass tensor method (MTM)The cube
The human face
Linear elasticityFrom 24.57 s to 2.3 sFrom 53,3380 N to 10,368 N
Tetrahedral mesh
CPU

Chen et al. [44]MDMass-spring system method (MSM) combined with quasistatic algorithmThe human brainLinear elasticity48 Hz–3,000 Hz (haptic feedback)8,000 N
Surface triangle elements
SGI Prism Server 4 GPU, 8 CPU, 32 GB RAM

López-Cano et al. [45]MIMass-spring system method (MSM)The human inguinal regionLinear elasticity73 FPS (system iteration)4,891 V
Surface triangle elements
GPU NVIDIA 6,800, Pentium IV 3.0 GHz CPU, 1 GB RAM

Lim and De [46]MDPoint collocation-based finite spheres (PCMFS)The human liverNonlinear elasticity1 ms (model deformation)1,186 polygons
Polygon elements
Pentium IV 2 GHz CPU, NVIDIA Quadro4 XGL

Murai et al. [5]MDInverse dynamic computation (IDC)The human musclesLinear elasticity16 ms (muscle tension estimation)
15 FPS (model rendering)
274 musclesIntel Xeon 3.33 GHz CPU, 3.25 GB RAM, NVIDIA Quadro FX3700 GPU

Basafa and Farahmand [47]MDMass-spring-damper method (MSD)The cube modelNonlinear viscoelasticity5 ms (model deformation)
150 Hz (haptic feedback)
30 Hz (model rendering)
96 N-270 E
Tetrahedral mesh
500 N
Tetrahedral mesh
3.2 GHz Core Duo CPU, 1 GB RAM

Wang et al. [48]MDLaplacian surface deformation (LSD)The human noseLinear elasticityNINIWindows 2000 or Windows XP, 512 MB RAM or 250 MB

Ho et al. [6]MDMass-spring system method (MSM)The human eardrumLinear elasticity1 kHz (haptic feedback)
30 Hz (model rendering)
917 E
Surface triangle elements
Intel Core2 Q6600 CPU, NVIDIA GeForce 9,600

Wan et al. [49]MDRadial basic function (RBF) & geodesic distance-muscleThe human faceLinear elasticity0.0316 s (one system frame computing)5,272 V-10,330 F
Surface triangle elements
Intel Core 2 Duo E7200 2.53 GHz CPU, 2 GB RAM

Le et al. [50]MDThin-shell deformation method (TSD)-muscleThe human faceLinear elasticity73.8 FPM (facial animation)
164.3 FPM (facial animation)
40 markers
100 markers
Intel Xeon 2.4 GHz 16-Core CPU, NVIDIA Tesla C1060 240-Core GPU

Zhang et al. [51]MDElastic-plus-muscle-distribution-based (E+MD)The facial musclesLinear elasticityNININI

Weng et al. [52]MDFacial motion regression algorithm (FMR)-muscleThe human faceNI>200 FPS (graphic rendering on PC)
30 FPS (graphic rendering on mobile devices)
75 facial markersCore i7 3.5 CPU
Intel Atom 2.0 GHz CPU

Goulette and Chen [53]MDHyperelastic mass link method for FEM (HEML-FEM)The cube modelViscohyperelasticity4.02 ms (one model computation iteration)
21.24 ms (one model computation iteration)
4,430 E–1,128 N
Tetrahedral mesh 21,436 E–5,591 N
Tetrahedral mesh
Core 2 Duo 2.40 GHz CPU, 3.45 GB RAM

Zhang et al. [54]MDThe time-saving volume-energy conserved ChainMail method (TSVE-Chainmail)The cube modelNonlinear elasticity30 Hz (model rendering)NICore i7-4700 3.4 GHz CPU

Woodward et al. [55]MDRadial basis function mapping approach (RBF)-muscleThe human faceLinear elasticity2 minutes (system initializing)
Up to 30 Hz (facial feature detection)
NINI

Zhou et al. [56]MDMarquardt radial basis meshless method (MRM)The general cube modelNonlinear elasticity0.1509 s (model deformation)121 nodes
Tetrahedral mesh
Core i7-4790 3.60 GHz CPU, 8 GB RAM, Intel HD Graphics 4600 (64 MB)