BiopSym : a simulator for enhanced 
learning of ultrasound-guided prostate 
biopsy 
Stefano Sclaverano1, Grégoire Chevreau1,2, Lucile Vadcard3, Pierre Mozer2, 
Jocelyne Troccaz1 
1TIMC-IMAG Laboratory, UMR 5525 CNRS UJF, Grenoble, France 
2Urology department, La Pitié Salpétrière Hospital, Paris, France 
3LSE, UPMF, Grenoble, France 
Abstract: This paper describes a simulator of ultrasound-guided prostate biopsies 
for cancer diagnosis. When performing biopsy series, the clinician has to move the 
ultrasound probe and to mentally integrate the real-time bi-dimensional images 
into a three-dimensional (3D) representation of the anatomical environment. Such 
a 3D representation is necessary to sample regularly the prostate in order to 
maximize the probability of detecting a cancer if any. To make the training of 
young physicians easier and faster we developed a simulator that combines images 
computed from three-dimensional ultrasound recorded data to haptic feedback. 
The paper presents the first version of this simulator. 
Keywords: ultrasound simulation; biopsy training; prostate.  
1. Problem 
Prostate cancer is the first cancer of men in many countries. When a cancer is suspected 
biopsies are collected in the gland, most often based on ultrasound (US) guidance, for 
further anatomo-pathologic analysis. The US probe is inserted into the rectum of the 
patient and a mechanical guide is attached to the probe for needle insertion (cf. fig.1). 
Based on prostate exploration and image understanding the clinician repeatedly 
positions the probe and collect a sample in order to execute a predefined sampling 
scheme. The clinician has to face several difficulties: (1) the cancer, if any, is generally 
not visible in the US images; (2) the biopsy schemes are 3D whilst US images are 2D; 
(3) the prostate may be moved and deformed by the US probe. Such difficulties limit 
the gesture accuracy and degrade the sensitivity of biopsies. We have developed a 
simulator for US-guided prostate biopsies (1) to enable relevant learning, (2) to make 
quantitative evaluation of operators possible and (3) to allow the quantitative 
comparisons of protocols. 
 
 
Figure 1. Prostate anatomy and US image acquisition for biopsy (left) – US image 
with needle trajectory (right). 
 
In addition to the previous clinical difficulties, the training phase of this 
intervention also poses some problems. A large part of the expertise concerning 
biopsies relies on the coordination between vision (US 2D images), haptic perception 
(holding the probe) and cognition (mental representation of the gland). This 
coordination is necessary to link visual and haptic feedback to anatomical knowledge, 
and take decision concerning samples location. Simple observation of clinicians at 
work is not sufficient to provide residents with all these aspects of knowledge and to 
train their links in situation. In particular, the cognitive processes involved for the 
treatment visual information and for the control of actions (coordination of 2D and 3D) 
are far from being easy to elaborate [1], [2]. 
Simulation is a possible answer to this training difficulty. Simulators may involve 
role plays, virtual systems, or mixed systems including a virtual and a robotic part [3]. 
In any case, we argue that simulation alone is not sufficient to manage relevant training. 
Focusing on training implies to define some pedagogical features of the training system. 
Even in the case of standardized patients1, these latter are aware of the kind of reaction 
and feedback they have to give residents during and at the end of the training session 
[4]. In the following sections we give details on the way we address these issues in the 
case of Biopsym.  
2. Tools and methods 
The simulator includes a haptic device, the Omni Phantom from Sensable Devices Inc., 
which allows the operator to move the virtual US probe with respect to the anatomy of 
the virtual patient (see fig.2). The system generates US images corresponding to the 
position of the probe. A haptic feedback renders both the constraint related to the 
motion of the probe in the rectum and the biopsy gun percussion when a sample is 
collected.  
                                                           
1 Standardized patients are healthy volunteers who participate in a simulated 
examination. Of course, such type of simulation cannot be reasonably used for gestures 
such as biopsies. 
 
Figure 2: Using the Omni Phantom for the manipulation of the virtual US probe 
2.1. Methods 
Based on the motions of the virtual probe transmitted by the operator using the haptic 
device, the system generates corresponding US images. The haptic feedback renders 
both the constraint related to the motion of the probe in the rectum and the biopsy gun 
percussion when a sample is collected.  
Two approaches can be used for US image simulation. One consists in generating 
US images [5] based on the modelling of US wave propagation and interaction with the 
human tissues. Such an approach generally requires complex models and is time 
consuming or images may have a limited realism when the model is simplified for 
computational efficiency. Another approach consists in using databases of recorded US 
exams and producing new images by exploiting this database. Interpolation may be 
necessary [6] when the image position does not correspond to recorded data. We have 
selected this second type of approach based on a 3D database. In this case, the 
generation of an image simply corresponds to slicing the volume in the direction of 
interest.  
One major issue related to US images concerns the tissue deformations that may 
result from the image acquisition itself (probe pressure) or from external actions 
(needle insertions for instance). This issue not yet handled with this version of the 
simulator will be discussed in section 4. However, even without this function, the 
simulator may already help the trainee to navigate in the prostate volume, to visualize 
the prostate appearance and to correlate his/her probe 3D motions to corresponding 2D 
slices. We design targeted exercises for this training objective. 
Biopsym uses open source libraries (see fig.3) and is multi-platform. 
2.2. Database 
3D US volumes have been acquired during real biopsy sessions at La Pitié Salpétrière 
Hospital for more than two years using a GE Voluson 730 system with a 3D endorectal 
US probe. By mid-june 2008, 87 patients were included. The database provides 
information about the prostate size, the patient age, the PSA level, etc. The SQL 
encoded database also includes information about operators using the simulator and 
records simulations of the operator in terms of patients concerned and positions of the 
simulated biopsies. Statistics can be computed from this database. 
 
 
Figure 3: Software environment 
3. Results 
The simulator allows the trainee visualizing 3D US data to explore patients’ anatomy 
using a multi-view rendering (axial, sagittal, coronal, probe-oriented 2D US images 
and 3D representation of the probe and needle with respect to the prostate volume) – 
see fig. 4.1. It also allows to simply visualize the 2D view corresponding to a real 
biopsy session and to exercise biopsies on the virtual patient (see fig. 4.2). This 
simulation includes the haptic feedback which makes the simulation more realistic. 
This haptic feedback can be switched off to make possible training on a PC with 
standard keyboard and mouse interfaces. Three clinicians have tested the simulator and 
they evaluate very positively this first version of the simulator from the viewpoint of 
image realism and haptic feeling. 
 
 
Figure 4.1: 3D visualization of the simulation environment (the probe and the 
plane to be visualized w.r.t. the volume of data and prostate). 
 
  
 
Figure 4.2: Multiple views of a biopsy trajectory computed from the 3D volume 
(left: sagittal – right: coronal – the needle trajectory is visualized in green in both views 
(line on left view and dot n°2 on right view); a previously recorded biopsy is in red in 
the right view (dot n°1) 
 
Statistics computed from trainees’ actions will be used to identify particular 
difficulties: for instance about sampling the apex of the prostate or its boundaries, 
about the length of biopsies (i.e. the actual depth of the needle in the gland), about the 
regular spreading of samples, etc. Those difficulties are linked to targeted sets of 
training exercises. Biopsym will thus offer personalised training paths, according to 
trainees’ performances on the simulator.   
4. Current and future work 
On the scientific side, we plan to increase the image realism by adding real-time 
deformations of the images due to the probe. The generative approach can deal quite 
naturally with those questions provided that a model of deformation is available to 
deform the organs before image generation. Interpolation approaches may integrate 
deformed volumes to produce deformed images [6] or may apply a deformation model 
(mass-spring [7] or Finite-Element [8]) to images produced from a deformation-less 
recorded volume. We are in the process of comparing the advantages and drawbacks of 
the different approaches in the interpolation framework. 
On the clinical side, further work deals with the extensive use of the simulator by 
clinicians with various levels of expertise. On the pedagogical side, the specific 
exercises will be included to help the learner in his/her skill acquisition: e.g. try to 
locate a given region in the 3D US volume or attempt to reach a specific target in the 
gland. Hints for visualisation will be available according to the trainee level (see fig. 
4.1 for example). We will use data on learning curves (those of clinicians on real 
patients and those of trainees on Biopsym) to evaluate the impact of the simulator on 
biopsies performances.  
5. References 
[1] A. Blavier, A.S. Nyssen, (2007). Impact of the da Vinci robotic system on the learning of surgical skills: 
implication and limits for training and safety. Proceedings of Surgetica 2007, Chambery, France 
[2] L. Vadcard, V. Luengo, Interdisciplinary approach for the design of a learning environment, E-Learn 
2005 – World Conference on E-Learning in Corporate, Government, Healthcare, and Higher Education, 
Vancouver, Canada, October 24-28, 2005 
[3] R. Moreau, , M.T. Pham, T. Redarce, O.Dupuis, A new learning method for obstetric gestures using the 
BirthSIM simulator, IEEE International Conference on Robotics and Automation, 10-14 April  2007, 
2279 - 2284 
[4] M Siebeck, F. Fischer, C. Frey, B. Schwald. Standardisierte Patienten für die rektale Untersuchung. 
Poster, Jahrestagung der Gesellschaft für Medizinische Ausbildung, Hannover, 16.-18.11.07   (GMA 
Posterpreis) 
[5] C. Forest, O. Comas, C. Vaysière, L. Soler, J. Marescaux. Ultrasound and needle insertion simulators buit 
on real patient-based data. Proceedings of MMVR’2007 
[6] J.Troccaz, D. Henry, N. Laieb, G. Champleboux, J.L. Bosson, O. Pichot. Simulators for medical training: 
application to vascular ultrasound imaging. Journal of Visualization and Computer Animation, 11:51-
65, Wiley, 2000 
[7] D. d’Aulignac, C. Laugier, J. Troccaz, S. Viera. Towards a realistic echographic simulator. Medical 
Image Analysis, 10:71-81, 2006 
[8] O. Goksel, S.E. Salcudean. Fast B-mode US image simulation of deformed tissue. Proceedings of IEEE 
EMBS Conference, 2007