Journal Logo

Review Article

A Review of Computer-Based Simulators for Ultrasound Training

Blum, Tobias Dipl.-Inf.; Rieger, Andreas MD; Navab, Nassir PhD; Friess, Helmut MD; Martignoni, Marc MD

Author Information
Simulation in Healthcare: The Journal of the Society for Simulation in Healthcare: April 2013 - Volume 8 - Issue 2 - p 98-108
doi: 10.1097/SIH.0b013e31827ac273
  • Free


In the last decades, ultrasound has become indispensable for a wide range of diagnostic, therapeutic, and surgical applications. Ultrasound has several advantages over other imaging modalities: ultrasound does not involve ionizing radiation, the devices are inexpensive, and portable systems can be used bedside, during interventions or in ambulances and helicopters. In emergency medicine, ultrasound has become a valuable tool for providing a first diagnosis, as it requires less preparation time compared with other modalities.

Although the use of ultrasound has many advantages, there are some fundamental drawbacks. The image quality suffers from low dynamics, spatial resolution, and signal-to-noise ratio. For novices, it is difficult to interpret ultrasound images owing to different imaging artifacts, some of them depending on the viewing direction. Although radiologist can interpret every computed tomographic (CT) or magnetic resonance (MR) volume owing to standardized cut planes and viewing modes, ultrasound images depend on the position of the probe. The use of ultrasound relies heavily on the examiner. Furthermore, interobserver and intraobserver repeatability is low, which has been shown for thyroid volumetry1,2 and prostate volumetry.3 Even for routine procedures such as trauma ultrasound, sensitivity is low.4 The requirement for intensive training is widely accepted. The recommended number of examinations a novice should perform under guidance range from 20 for sentinel node biopsy5 and 25 for fetal echocardiography6 to 300 for critical care7 and 480 for echocardiography.8 However, even such high numbers of guided examinations may not suffice to use ultrasound confidently. Hertzberg et al9 showed that even after 200 recommended cases, radiologists still had problems depicting anatomic landmarks and performing examinations. Even more surprising, a study by Moore et al10 showed that in emergency medicine, few hospitals even follow the recommended number of supervised procedures.

During the last 15 years, several computer-based ultrasound simulators have been developed. In this article, we will classify simulators according to (1) the method that is used to simulate the ultrasound image; (2) the user interface, which consists of the input device, haptic simulation and output device; and (3) the medical application domain. We discuss training concepts that can be realized using computer-based ultrasound simulators and their advantages over traditional training. Finally, we summarize existing evidence on the learning effect and discuss commercial systems.


Today, physical phantoms are primarily used for ultrasound training. These phantoms mimic relevant physical properties of real tissue and are used with real ultrasound probes. Phantoms are made of different materials such as foam, gelatin, or fiber,11–13 and many different phantoms of whole human bodies and single organs exist.

Computer-based simulators do not use a real ultrasound probe but rather simulate the ultrasound image in the computer. Because many different methods for computer-based ultrasound simulation exist, we will first classify and discuss existing research systems, before turning our attention to training concepts that can be realized. Table 1 summarizes key aspects of the most relevant research systems.

Summary of Key Aspects of the Most Relevant Research Systems

Simulation of Ultrasound Images

The methods to simulate ultrasound images can be categorized into interpolative, generative image-based, and generative model-based methods. A brief summary of these methods is provided in Table 2.

Summary of the Most Important Aspects of Different Methods to Simulate Ultrasound Images

Interpolative Simulation

The most common method to simulate 2-dimensional (2D) ultrasound images is interpolation from 3-dimensional (3D) volumes that have been acquired from real patients.16,17,19,20,25,28–30,38,56–59 An example of an interpolative simulation can be seen in Figure 1C. Using state-of-the-art hardware, it is simple to implement interpolative methods running in real time on the graphics processing unit. Images are very realistic when they are simulated from a position similar to the position that was used to acquire the 3D volume. The main drawback is that as soon as the user changes views, view-dependent effects are not simulated correctly. There are different solutions to this problem. It is possible to acquire several 3D volumes from different viewpoints and switch among them depending on the position of the probe. Although this method can provide multiple viewpoints, it is not practical for small changes of the position because a high number of volumes would have to be acquired. View-dependent effects such as gain, depth gain compensation, focus, and shadow can be added to the ultrasound image after interpolation.39,56 However, simulation of all view-dependent effects is not possible because no underlying physical model is used. In particular, when a 3D ultrasound volume already contains artifacts and shadows, it is difficult to remove them and replace them by new artifacts. For procedures such as cardiac ultrasound, motion is highly important. For prenatal heart diagnostic, Wüstemann et al60 acquired 4-dimensional (4D) ultrasound volumes and interpolated 2D slices from the 4D volume running in a loop.

Ultrasound images simulated with different methods. A, Simulated IVUS image based on a computer model.53 B, Image of the liver and a needle generated from a segmented CT volume and texture database.45 C, 2D slice showing a fetal aortic arch interpolated from 3D ultrasound volume.60

Generative Image-Based Simulation

For non–real-time applications such as transducer design, generative methods have been developed that simulate wave propagation on 3D volumes such as CT or MR volumes.61 Because these methods do not run in real time, different methods using simplified physical models have been proposed, usually only considering ray-based propagation of sound instead of wave propagation. The image quality depends on the physical models and on the tissue properties embodied in the 3D volumes. In CT, a segmentation into the air, bone, and soft tissue can be obtained using intensity thresholds and basic effects such as absorption and reflection can be simulated, as done by Hostettler et al40 and Vidal et al.42 A correlation between Hounsfield units and the acoustic impedance was assumed by Reichl et al62 and Shams et al49 who used this assumption to simulate absorption, reflection, and transmission. Imani et al63 and Bommersheim et al52 simulated ultrasound from the segmented Visible Human Dataset (VHD).64 They assigned tissue characteristics such as acoustic impedance, absorption, and scattering coefficients to the segmented structures. Bürger et al51 used segmented CT and MR images and manually assigned tissue properties. An alternative method to ray-based simulation is the use of textures. Zhu and colleagues45,46 manually segmented CT images and assigned labels to each voxel. They simulate ultrasound images by texturing a 2D slice with textures obtained from real ultrasound images. An image of this is shown in Figure 1B.

Generative Model-Based Simulation

Although large numbers of CT and MR images are available, image-based simulation has some limitations. In particular, for ultrasound of small and moving anatomy such as the heart, CT does not provide enough information. One way to overcome this is modeling of the anatomy. Sun and McKenzie65 have built a model of the heart that models the movement of the valves. The ultrasound image is created by extracting a 2D slice from the model and texturizing it. For simulation of intravascular ultrasound (IVUS), Abkai et al53 used functional descriptions of a flexible tissue model of the vessel system. An exemplary image of this method can be seen in Figure 1A. For echocardiography, the pumping motion of the heart, operation of the valves, and blood stream were modeled.31 Bürger et al51 used 3D CT or MR volumes and animated the heart by forward free-form deformation, and for simulation of the abdominal area, Ni et al39 modeled respiratory motion. One challenge for model-based simulation is offering different cases. Most models are based on images of one patient and therefore can only provide simulation of this patient. Based on an MR image, Köhn and colleagues54,55 constructed a mathematical model of the heart. The impact of different pathologies on the structure of the heart was modeled. This allows simulating different pathologies or combinations of pathologies. Although this method is very appealing for generating cases for different pathologies, a lot of work has to go into modeling and confirming that the model is correct.

User Interface

In this section, we will discuss all issues related to user interfaces and user interaction. This includes not only the methods for tracking the position of the ultrasound probe but also the simulation of haptics and the output device. Some of the systems that are discussed can be seen in Figure 2.

Different user interface setups of ultrasound simulators. A, Physical probe and patient phantom.66 B, Haptic device with stereoscopic overlay.67 C, Transesophageal simulator with phantom.68

Input Devices

Many systems use physical patient and probe phantoms and track their position. The most common choice for tracking is the use of electromagnetic (EM) tracking.28–30,47,57,65 In particular, for applications where the ultrasound transducer is inserted into the patient, EM systems have the advantage that they do not require a free line of sight to the tracked objects, as optical tracking systems do.

Markov-Vetter et al59 used optical tracking to estimate the position of a phantom of a newborn and an ultrasound probe. One advantage of optical tracking systems compared with EM systems is that the tracking volume is larger. This is important, in particular, when a system requires tracking of additional devices such as a head-mounted display (HMD) for augmented reality (AR) visualization.50

An alternative to using patient and probe phantoms are haptic force-feedback devices, which are used in many systems.14,21,23,25,38,41,42,51,62 Haptic devices have a limited working range and require more abstraction from the user because no patient phantom is used. Haptic devices allow measuring the force that is applied. This can be used to simulate the deformation of the image. In some applications such as diagnosis of thrombosis or carotid stenosis, it is important to apply a certain amount of pressure. For such applications, it is crucial to estimate the exact force that is applied to allow realistic simulation and provide feedback to a trainee.

With the aim of building a low-cost simulator for developing countries, ap Cenydd et al44 used controllers from the Nintendo Wii console, which contain accelerometers and a simple infrared tracking system, to control probe and patient position. Some systems do not mimic a real ultrasound probe but use keyboard and mouse to navigate a virtual probe.16,17,69 Although this has the huge drawback that training of hand-eye coordination is not possible, such systems are less expensive, and students can use them on their own PC or even over the Internet. For instance, Kempny and Piórkowski69 showed a simulator for transesophageal echocardiography (TEE) that is accessible through the Internet. Table 3 provides some examples of Web sites offering interactive content related to ultrasound simulation.

Web Sites with Interactive Content Related to Ultrasound Training (Last Accessed: June 5, 2012)

Haptic Simulation and Image Deformation

A crucial but very complex issue for realistic simulation is to provide realistic haptic feedback when using force-feedback devices. Another related issue is the deformation of the ultrasound image as a result of pressure applied to the patient.

Proxy-based methods have been used by Vidal et al42 to provide haptic feedback. Although these methods are fast to compute, image deformation is usually not simulated. Common methods for simulating tissue deformation are the mass spring model and the finite element method. Alterovitz et al14,15 and Goksel and Salcudean24 used the finite element method to simulate soft tissue deformation owing to the force applied through the probe. D’Aulignac et al21 used mass spring models with deformation parameters obtained experimentally from a patient using force sensors.

Another way to simulate image deformation for specific applications is to record the deformations during a real ultrasound procedure. Troccaz et al22 presented such a method, where ultrasound images with and without pressure have been obtained, and the images are interpolated between both volumes during runtime.

Several systems use haptic force-feedback devices to simulate the insertion of a needle. These systems need to simulate the forces between the needle and the tissue as well as needle bending. Alterovitz et al14,15 proposed a method taking into account the cutting at the needle tip, membrane puncture, and friction. Ni et al38 simulated needle insertion, taking into account prepuncture forces, friction forces, and cutting forces. Magee et al45 simulated the bending of the needle, and Vidal et al42 used a model that combines a proxy-based approach when the needle is outside the patient body and a method using the CT intensities and in vitro measurements when the needle is inside the body.

Output Devices

Most systems display the 2D ultrasound slice on a standard monitor. Systems not using a physical phantom have to visualize relative poses of the probe and the patient. This can be done by showing a 3D scene on a monitor. The drawback of using standard monitors is the reduced immersion because the user has to look at the virtual probe, which is not colocated in space with the user’s hand. To solve this problem, Vidal et al43,67 used a semitransparent mirror and shutter glasses to augment a visualization of the patient into the workspace of the haptic devices. This system allows for AR overlay of bones and other key organs onto the phantom and can be seen in Figure 2B. Berlage et al31 used a stereo monitor to show a virtual representation of the heart and the ultrasound plane. Guirlinger70 presented an AR simulator for prenatal diagnosis augmenting the geometry of the ultrasound plane into the image of a webcam. An AR simulator using a video see-through HMD has been shown by Blum et al.50 It can show bones and organs augmented onto a patient phantom.

Medical Applications

Many simulators were developed for cardiology. Ultrasound simulators have been used in transthoracic echocardiography (TTE) to get a better understanding of the heart anatomy and to practice standard slices by Berlage et al,31 Weidenbach et al,32,33,37 and Sun and McKenzie.65

For some applications, the transducer is inserted into the patient. Although the insertion of the probe is usually not invasive, training on patients is problematic because it is very uncomfortable for the patient. One example is TEE, where the examiner cannot see the head of the transducer. Weidenbach et al,30 Bose et al,68 and Kempny and Piórkowski69 presented simulators for TEE.

Another challenging field for training on real patients is prenatal diagnosis. Many findings are rare, making it difficult to practice them on real patients. In addition, using patients for training, particularly for anomalies, is stressful for the patient. Forest et al41 developed a simulator for obstetrics, and Maul et al71 showed a simulator for prenatal diagnosis. Later they extended the same system to allow simulation of a beating heart.60 Simulators for neonatal cranial ultrasound were developed by Markov-Vetter et al59 and by Arkhurst et al.17,19

Simulators for gynecology were developed for ultrasound mammography by Marquardt et al,72 and a system for simulation of transvaginal ultrasound diagnosis has been developed by Heer et al.28 Simulators for radiology include systems that simulate IVUS by Abkai et al53 and deep venous thrombosis examination by Troccaz and colleagues.21,22 Terkamp et al57 showed a simulator that covers different pathologies in the abdominal area. For trauma surgery, Blum et al50 proposed a system for the focused abdominal sonography for trauma (FAST) protocol. Furthermore, Bommersheim et al52 developed a system for training longitudinal endoscopic ultrasound of the gastrointestinal tract.

Lastly, ultrasound simulators are important, in particular, for invasive procedures such as ultrasound-guided needle insertion where training on humans is problematic. An example by Alterovitz et al14 is radioactive seed implantation for prostate brachytherapy. Simulation of ultrasound-guided needle insertion has been shown for biopsies38,42,47 and hepatic biopsies.41 Simulators for transrectal ultrasound (TRUS)–guided biopsy of the prostate have been proposed by Sclaverano et al20 and Goksel and Salcudean.23 Another invasive procedure, which has been simulated by Forest et al,41 is radiofrequency thermal ablation.


After having discussed the different technologies and medical applications, we will now examine how computer-based simulators can be used for training. To understand the advantages computer-based ultrasound simulators can offer, we will first discuss the need for a mental model. This discussion is based on the work by Trochim34 and Berlage.36 Building up a mental model of complex relations and situations is a central part of learning. Because models direct our actions, only adequate models lead to adequate actions. We have a mental model that controls our actions, and we can observe the outcome of our actions. Therefore, we are able to verify our mental model, reject, or correct it. Although for easy tasks, trial and error might be enough to build up a correct mental model, complex tasks such as the use of ultrasound require a transfer of knowledge from an expert to the trainee. The most common method to transfer knowledge is writing and reading books. However, it is difficult to use books to transfer a mental model including knowledge about anatomy, spatial relations, and ultrasound physics. The main advantage of ultrasound simulators is that they provide new ways for a student to build up a mental model. Details about how simulators can facilitate this process are discussed later.

Virtual Scenes

One of the main problems for beginners is the mental mapping between the 2D ultrasound plane and 3D anatomic structures. Many novices mix transducer rotation with angular movements and fail to combine several consecutive images during an ultrasound sweep into a 3D structure.31 In echocardiography, the most difficult aspects to learn are the relationship between 2D images and 3D heart anatomy and the adjustment of standard planes.32 In particular, this is true for procedures that do not use a hand-held probe, such as endoscopic ultrasound or IVUS, because there is a lack of context and a very small field of view.

Virtual scenes visualize these relations in 3D. Some examples are shown in Figure 3. To transfer the mental model of an expert to a trainee, the expert might decide that less important information is completely hidden and more important information is shown in greater detail. Many simulators show the ultrasound scan plane coregistered with additional 3D data such as the skin of the patient, bones, or other relevant organs. Some systems show the surface of the patient/phantom16,29,56; others show more detailed models of the anatomy23,48 and allow changing visualization parameters.33,38 Augmented reality has been used to allow visualizing structures directly onto the physical phantom50 as can be seen in Figure 3C. Arkhurst et al17 and Bose et al68 segmented and labeled anatomy and enhanced the virtual scene by textual information. Later Arkhurst19 used a more advanced model where segmented organs were connected to a knowledge base.

Different visualization methods that have been used in ultrasound simulators. A, Model of the heart with coregistered ultrasound slice.68 B, Ultrasound image and coregistered CT/MR data from Visible Human.27 C, AR visualization of an ultrasound plane and a physical phantom.50

Coregistered Images

The mental model when using ultrasound includes a mapping between the real anatomy and their appearance in the ultrasound image. In ultrasound, it is very difficult to build up a mental model about the appearance of structures in the image. One way to aid in building up a mental model is showing an ultrasound slice and the corresponding slice from another modality. Ehricke16 showed ultrasound slices and corresponding CT/MR/photographic slices from the VHD. Coregistered MR slices have been used by Arkhurst et al17,19 and CT slices by Blum et al.50 Tahmasebi et al26,27 used a haptic feedback device to move the ultrasound probe, while showing corresponding slices from the VHD, which can be seen in Figure 3B. Coregistered images could also be presented without using a simulator, printed or on a screen. However, it has been shown that simultaneous training of cognitive and motor skills enables faster learning compared with training of only 1 aspect.73


The previous training concepts virtual scenes and coregistered images can help correct mental models. However, they cannot provide feedback on more complex actions such as how to reach the correct view of a structure or whether a student carried out a procedure correctly. A literature review by Issenberg et al74 synthesized existing evidence from 109 studies regarding the features of medical simulators that lead to most effective learning. The most important feature of a simulator is to provide feedback, which conforms to the concept of a mental model that has to be updated. One way to provide such feedback is to let a novice perform a task and provide record and replay capabilities.31,48 This can be used to identify and discuss errors. Feedback can also be provided by showing how an expert performs the same procedure. This feature has been implemented by Aiger and Cohen-Or.56 Blum et al50 showed a system that records the performance of an expert and a trainee and performs a temporal synchronization between both. This is used to provide a synchronized replay, which can be used by trainees to compare their actions to the expert. Trochim34 built a system that recognizes whether a trainee carries out a procedure correctly and provides feedback. Hidden Markov models are used to model the temporal sequence of actions, and fuzzy rules are used to detect single actions and compare them to reference actions. Color-coding of structures provides feedback on how well a view taken by the trainee captures important structures.35

Different Difficulty Levels

Another aspect that has been found critical for medical simulators is the ability to provide a range of difficulty levels74 for users with different skills. An example of this has been shown in a system by Weidenbach et al33 where different visualization aides can be used in an echocardiographic training scenario. The system can show the relative position of the probe and the anatomy, and it can visualize additional outlines of the target organs on the scan plane to aid beginners. For more advanced users, these aides can be switched off. Similarly, Forest et al41 offers different difficulty levels, where a transparency mode is used for less experienced users to improve understanding of spatial relations. An educational game for teaching ultrasound skills, offering different stages with varying difficulty, has been presented by Chan et al.75

Case Databases

One reason for using computer-based simulators is the lack of appropriate cases in traditional training. A study by Costantino et al76 showed that the number of scans during the residency year is more important than the number of didactic hours. When using simulators, it is possible to make a wide range of cases available to students, including rare cases. For instance, there are fetal abnormalities that only occur in one of 200,000 to 400,000 pregnancies.77 Even if a patient with a rare abnormality would be available, training of a high number of students is delicate for prenatal abnormalities because the mother is subject to high mental stress. Owing to the lack of training cases, detection rates are low. The reported values range from 14.2%78 to 44.5%.79 Another application where appropriate cases are not easily available for training is the diagnosis of patients with trauma, where ultrasound is, for example, used for diagnosis of internal bleedings.

The UltraSim system56 provides different cases, including patient history, patient medical information, and diagnostic data. A case database has also been built by Ehricke.16 The Sonotrainer system has a range of cases, each consisting of a 3D ultrasound volume, an assignment, and answers for the assignment.58 A parametric representation of heart and pathologies, which allows generating different cases, has been proposed by Reis et al.55

Standardization of Training

Today, there is a low level of standardization in ultrasound training. Ultrasound simulators could be used in a highly standardized learning curriculum, by providing a defined set of cases. For standardized assessment of students, simulators have advantages because objective criteria for evaluation can be defined. Monsky et al80 used a simulator for the training of residents. To assess them before taking overnight calls, trainees had to perform measurements, which were compared with measurements of experts on the same patient data.

Invasive Procedures

Procedures such as transesophageal ultrasound, transvaginal ultrasound, or TRUS can be uncomfortable and painful for the patient. This is even more problematic for some ultrasound-guided procedures that are invasive and where training on healthy patients can potentially be harmful for the patient. Examples for this are needle biopsies or radioactive seed implantation. For these applications, training using simulators is of great value to improve patient safety and comfort.


The number of evaluations of computer-based ultrasound simulators is still low. Different aspects of simulators have been looked at. The evaluations are based on different population types, medical applications, and simulation techniques. There are currently not enough studies to perform a full metareview and to draw definite conclusions. Nevertheless, it is interesting to study the first results. Later, an informal summary of previous studies is provided.

Different aspects of a simulator have to be evaluated. Face validity judges the degree of resemblance between simulator and reality. Content validity shows how appropriate a system is for teaching and to what extent it covers the real activity. Construct validity is the ability of a simulator to discriminate among subjects of different ability or experience. Multiple studies have investigated content and face validity by carrying out questionnaires with experts and trainees. Content validity was rated high to very high, and face validity was rated medium to high. More details are provided later.

The face validity for the realism of the simulated ultrasound images30,37,48,58,71 and for the realism of the entire ultrasound procedures32,37,42,48,57 has been rated medium to high in different questionnaires. Additional questionnaires on face validity for the handling of the probe showed high ratings for using physical phantoms30 but low ratings for using haptic feedback devices.42,48 Face validity can also be shown by letting a user perform the same procedure on a simulator and on real patients and comparing the results. Two such studies showed medium28 to high57 correlations between real and simulated diagnoses. High specificity and sensitivity for performing a diagnosis on a simulator using ultrasound from real patients has been reported by Maul et al.58,71 High–to–very high ratings were given in different questionnaires regarding whether simulators are useful for training.28,30,37,42,48,58,71 In particular, training of spatial relations, transducer steering, and standard planes have been rated very high.30,32 Questions regarding content validity of haptic feedback simulation have only been rated medium.42 Different studies reported medium-to-high levels of construct validity.37,39,45

There are few studies investigating how well simulators perform compared with traditional training. Maul et al71 showed that additional simulator training improves performance compared with only performing standard training. Two studies reported that students receiving simulator training instead of a standard lecture using images and drawings showed better results.33,81 In a pretest/posttest, it was shown that simulator-based training leads to comparable results as training on real patients,82 and another study reported that no significant difference from noncomputer phantoms could be found.83 Although these results are promising, further studies are required to establish well-founded evidence about the effect of simulator-based ultrasound training. In particular, comparisons of long-term training effects between traditional and computer-based training are lacking.


Among the systems that have been discussed before the UltraSim system by MedSim,56 the Schallware system58,60,66,71,72 (Figs. 1C and 2A) and the HeartWorks system68 (Figs. 2C and 3A) are commercially available. In addition, there are other commercial systems that have not been described in the scientific literature. A summary of commercial systems is shown in Table 4.

Summary of Key Aspects of Commercial Systems

The Vimedix and the HeartWorks systems are made for teaching of TTE and TEE. Both use model-based generative image simulation. This is because imaging of the heart requires a very detailed simulation, which can be achieved by using computer models. The UltraSim, ScanTrainer, and Schallware systems use interpolative methods. One system, the SonoMom, shows prerecorded 2D ultrasound images, depending on the position of the probe in the simulator. This has the disadvantage that only the position, but not the orientation, of the probe is considered. However, this solution does not require expensive hardware to track the orientation of the probe. Although many research systems use image-based generative simulation, none of the commercial systems uses this technique. One possible reason for this is that these methods require very powerful hardware, and most of the related research has only been published within the last years.

For the user interface, all commercial systems to date use standard monitors and no system uses technologies such as stereo monitors or AR. Only the ScanTrainer uses a haptic device to control the probe. All other systems use phantoms of patients and ultrasound probes. Most commercial systems target applications where training on healthy patients is difficult. Both the Vimedix and the HeartWorks systems provide training for TTE and TEE. Training for obstetrics can be done using the SonoMom or the ScanTrainer systems. Both the Schallware and the UltraSim systems provide different modules for a range of medical applications.

Unlike most research systems, most commercial systems provide multiple cases for training, often with different difficulty levels. The training concept of virtual scenes is used by several systems. Both systems for training of TTE and TEE are based on computer models and use these models for visualizing the relative position of the probe and the ultrasound plane with respect to the anatomy.


Although long-term training effects have not been investigated yet, the first studies on computer-based ultrasound simulation are promising, and it is already used in regular training programs.58 However, there are still many technological, educational, and organizational issues that have to be addressed.

To see whether computer-based ultrasound simulators will be helpful, it is important to compare them with traditional simulation using physical phantoms. Regarding image quality, computer-based simulators do not provide as realistic images as physical phantoms. Although interpolative simulation methods can provide very realistic images, they cannot do this from every viewpoint. Generative methods do not provide very realistic images because they use simplified physical models. Another aspect where traditional systems are better than computer-based systems is haptics and deformation, where current computer-based simulation methods are not very realistic. Furthermore, haptic devices only simulate interaction between probe and patient. Feeling anatomic landmarks with the hands is not possible. Only few systems simulate the deformation of the ultrasound image, which is very important for some applications. Additional research is required to improve the image simulation, haptic simulation, and deformation. Improved haptic simulation is important to simulate invasive procedures where simulators would be of great benefit.

Although there are some drawbacks, computer-based simulators offer many advantages over traditional systems. One important advantage, which is provided by most commercial systems, is different cases. A physical phantom can only represent one specific anatomy or pathology, whereas a computer-based simulator can provide many different cases and can be extended by a software update. Another advantage, which is implemented in several commercial systems, is the use of virtual scenes.

Looking at the evaluations, drawbacks, and advantages, the use of computer-based simulation using state-of-the-art methods cannot be recommended for all applications. For areas where ultrasound can be easily practiced on patients, the use of computer-based simulators does not seem reasonable. Owing to the low realism of the haptic simulation, traditional simulators should be preferred for applications where haptics are crucial. Nevertheless, we believe that many of the problems the current system have will be addressed within the next few years. In areas where it is difficult to train on patients and where it is important to see many different pathologies, the use of computer-based simulators is already a valuable alternative. Also for applications such as echocardiography, where it is very difficult to understand the spatial relations, current computer-based simulators have big advantages over traditional simulation.

We believe that long-term, one major advantage of computer-based simulators is that they will alleviate the time burden of teaching novices under the guidance of expert doctors. When using traditional simulators, an experienced medical doctor must be present to provide feedback. Virtual scenes help a student build a mental model. Advanced training systems can analyze how a student performs a procedure and provide feedback to the student. For simulators of minimally invasive surgery, research has already been done on automatically analyzing the performance of a student to do skills assessment and to provide feedback.84 Investigating the use of similar methods will be very important to make full use of the possibilities of computer-based ultrasound training and should be one of the main future research directions. Advanced methods to provide feedback and metrics to evaluate the skills of a student have a big potential because they allow autonomous and competency based training.

Most systems that have been presented up to now have been developed by medical and computer science researchers. To improve training using simulators, it will be important to involve researchers from pedagogy. To better understand how helpful the use of advanced visualization techniques, automatic skills evaluation, expert systems, and methods to automatically provide feedback are, the theoretical foundations of teaching should be taken into account when designing future training systems.

Although solving those educational and technological problems, organizational and financial issues have to be considered as well. It should be investigated whether simulators are only efficient when used under supervision of an experienced sonographer or whether it would allow students to perform training without supervision. We believe that simulators cannot replace a teacher, but they can reduce the time that experts have to be present. Moreover, the prices of simulators have to be considered. Systems using haptic devices and tracking systems cost more than a physical phantom, whereas systems without additional hardware running on a standard PC can be very inexpensive. Expensive systems might only be reasonable for skills laboratories that are used frequently or when they are integrated into training courses that are offered at multiple hospitals. On the other hand, systems that do not require additional hardware could even be offered for free. Furthermore, combinations should be taken into account. Skills laboratories could offer access to a simulator, whereas students could recapitulate and refresh their knowledge on a software version of the same simulator running on their own PC. Introducing such novel teaching methods that are only possible using computer-based simulation could help reduce costs, all the while increasing the quality of training.


1. Andermann P, Schlögl S, Mäder U, Luster M, Lassmann M, Reiners C. Intra-and interobserver variability of thyroid volume measurements in healthy adults by 2D versus 3D ultrasound. Nuklearmedizin 2007; 46: 1–7.
2. Brauer V, Eder P, Miehle K, Wiesner T, Hasenclever H, Paschke R. Interobserver variation for ultrasound determination of thyroid nodule volumes. Thyroid 2005; 15: 1169–1175.
3. Tong S, Cardinal H, McLoughlin R, Downey D, Fenster A. Intra- and inter-observer variability and reliability of prostate volume measurement via two-dimensional and three-dimensional ultrasound imaging. Ultrasound Med Biology 1998; 24: 673–681.
4. Stengel D, Bauwens K, Sehouli J, et al.. Systematic review and meta-analysis of emergency ultrasonography for blunt abdominal trauma. Br J Surg 2001; 88: 901–912.
5. Tafra L. The learning curve and sentinel node biopsy. Am J Surg 2001; 182: 347–350.
6. Ones M, Creager M. ACC/AHA clinical competence statement on echocardiography. J Am Coll Cardiol 2003; 41: 687–708.
7. Neri L, Storti E, Lichtenstein D. Toward an ultrasound curriculum for critical care medicine. Crit Care Med 2007; 35: S290–S304.
8. Ehler D, Carney D. Guidelines for cardiac sonographer education: recommendations of the American Society of Echocardiography Sonographer Training and Education Committee. J Am Soc Echocardiogr 2001; 14: 77–84.
9. Hertzberg B, Kliewer M, Bowie J, et al.. Physician training requirements in sonography: how many cases are needed for competence? AJR Am J Roentgenol 2000; 174: 1221–1227.
10. Moore C, Gregg S, Lambert M. Performance, training, quality assurance, and reimbursement of emergency physician-performed ultrasonography at academic medical centers. J Ultrasound Med 2004; 23: 459–466.
11. Bude R, Adler R. An easily made, low-cost, tissue-like ultrasound phantom material. J Clin Ultrasound 1995; 23: 271–273.
12. Smith J, Bergmann M, Gildersleeve R, Allen R. A simple model for learning stereotactic skills in ultrasound-guided amniocentesis. Obstet Gynecol 1998; 92: 303–305.
13. Liu Y, Glass N, Power R. New teaching model for practicing ultrasound-guided regional anesthesia techniques: no perishable food products! Anesth Analg 2010; 110: 1233–1235.
14. Alterovitz R, Pouliot J, Taschereau R, Hsu I, Goldberg K. Simulating needle insertion and radioactive seed implantation for prostate brachytherapy. Stud Health Technol Inform 2003; 94: 19–25.
15. Alterovitz R, Goldberg K, Pouliot J, Taschereau R, Hsu I. Needle Insertion and Radioactive Seed Implantation in Human Tissues: Simulation and Sensitivity Analysis. IEEE International Conference on Robotics and Automation (ICRA). 2003; 1793–1799.
16. Ehricke H. SONOSim3D: a multimedia system for sonography simulation and education with an extensible case database. Eur J Ultrasound 1998; 7: 225–230.
17. Arkhurst W, Pommert A, Richter E, et al.. A virtual reality training system for pediatric sonography. Comput Assist Radiol Surg 2001; 1230: 483–487.
18. Hacker S, Tiede U, Burmester E, Leineweber T, Höhne K. Ein virtuelles Trainingssystem für endoskopische Longitudinal-Ultraschalluntersuchungen. In: Bildverarbeitung für die Medizin. 2002: 149–152.
    19. Arkhurst W. Ein interaktiver Atlas für die Sonographie und Anatomie des Säuglingsgehirns [PhD thesis]. Hamburg, Germany: University Hamburg; 2005.
    20. Sclaverano S, Chevreaua G, Vadcardc L, Mozerb P, Troccaz J. BiopSym: a simulator for enhanced learning of ultrasound-guided prostate biopsy. Stud Health Technol Inform 2009; 142: 301–306.
    21. D’Aulignac D, Laugier C, Troccaz J, Vieira S. Towards a realistic echographic simulator. Med Image Anal 2005; 10: 71–81.
    22. Troccaz J, Henry D, Laieb N, Champleboux G, Bosson J, Pichot O. Simulators for medical training: application to vascular ultrasound imaging. J Vis Comput Anim 2000; 11: 51–65.
    23. Goksel O, Salcudean S. Haptic simulator for prostate brachytherapy with simulated ultrasound. Int Symp Biomed Simulation 2010; 150–159.
    24. Goksel O, Salcudean S. B-mode ultrasound image simulation in deformable 3-D medium. IEEE Trans Med Imaging 2009; 28: 1657–1669.
    25. Abolmaesumi P, Hashtrudi-Zaad K, Thompson D, Tahmasebi A. A haptic-based system for medical image examination. Conf Proc IEEE Eng Med Biol Soc 2004; 3: 1853–1856.
    26. Tahmasebi A, Abolmaesumi P, Hashtrudi-Zaad K. A haptic-based ultrasound examination/training system (HUTES). Int Conf Robot Automation 2007; 3130–3131.
    27. Tahmasebi A, Abolmaesumi P, Thompson D, Hashtrudi-Zaad K. A framework for the design of a novel haptic-based medical diagnostic simulator. IEEE Trans Inf Technol Biomed 2008; 12: 658–666.
    28. Heer I, Middendorf K, Müller-Egloff S, Dugas M, Strauss A. Ultrasound training: the virtual patient. Ultrasound Obstet Gynecol 2004; 24: 440–444.
    29. Stallkamp J, Wapler M. UltraTrainer—a training system for medical ultrasound examination. Stud Health Technol Inform 1998; 50: 298–301.
    30. Weidenbach M, Drachsler H, Wild F, et al.. EchoComTEE—a simulator for transoesophageal echocardiography. Anaesthesia 2007; 62: 347–353.
    31. Berlage T, Fox T, Grunst G, Quast K. Supporting ultrasound diagnosis using an animated 3D model of the heart. Int Conf Multimedia Comput Syst 1996: 34–39.
    32. Weidenbach M, Wild F, Scheer K, et al.. Computer-based training in two-dimensional echocardiography using an echocardiography simulator. J Am Soc Echocardiogr 2005; 18: 362–366.
    33. Weidenbach M, Wick C, Pieper S, et al.. Augmented reality simulator for training in two-dimensional echocardiography. Comput Biomed Res 2000; 33: 11–22.
    34. Trochim S. Situiertes Lernen in Augmented-Reality-basierten Trainingssystemen am Beispiel der Echokardiographie [PhD thesis]. Bielefeld, Germany: University Bielefeld; 2002.
    35. Weidenbach M, Trochim S, Kreutter S, Richter C, Berlage T, Grunst G. Intelligent training system integrated in an echocardiography simulator. Comput Biol Med 2004; 34: 407–425.
    36. Berlage T. Improving Human Skills in Echocardiography Through Scene-Based Enabling Systems. Technical Report. German National Research Center for Information Technology; 2008.
    37. Weidenbach M, Razek V, Wild F, et al.. Simulation of congenital heart defects: a novel way of training in echocardiography. Heart 2009; 95: 636–641.
    38. Ni D, Chan W, Qin J, et al.. An ultrasound-guided organ biopsy simulation with 6DOF haptic feedback. Med Image Comput Comput Assist Interv 2008; 11: 551–559.
    39. Ni D, Chan W, Qin J, et al.. A virtual reality simulator for ultrasound guided organ biopsy training. IEEE Comput Graph Appl 2009; 31: 36–48.
    40. Hostettler A, Forest C, Forgione A, Soler L, Marescaux J. Real-time ultrasonography simulator based on 3D CT-scan images. Stud Health Technol Inform 2005; 111: 191–193.
    41. Forest C, Comas O, Vaysiere C, Soler L, Marescaux J. Ultrasound and needle insertion simulators built on real patient-based data. Stud Health Technol Inform 2007; 125: 136–139.
    42. Vidal F, Healey A, Gould D, John N. Simulation of ultrasound guided needle puncture using patient specific data with 3D textures and volume haptics. Comput Anim Virtual Worlds 2008; 19: 111–127.
    43. Vidal F, Chalmers N, Gould D, Healey A, John N. Developing a needle guidance virtual environment with patient-specific data and force feedback. Comput Assist Radiol Surg 2005; 1281: 418–423.
    44. ap Cenydd L, Vidal F, John N, Gould D, Joekes E, Littler P. Cost effective ultrasound imaging training mentor for use in developing countries. Stud Health Technol Inform 2009; 142: 49–54.
    45. Magee D, Zhu Y, Ratnalingam R, Gardner P, Kessel D. An augmented reality simulator for ultrasound guided needle placement training. Med Biol Eng Comput 2007; 45: 957–967.
    46. Zhu Y, Magee D, Ratnalingam R, Kessel D. A virtual ultrasound imaging system for the simulation of ultrasound-guided needle insertion procedures. Med Image Underst Anal 2006; 61–65.
    47. Magee D, Kessel D. A computer based simulator for ultrasound guided needle insertion procedures. Int Conf Vis Info Eng 2005: 301–308.
    48. Zhu Y, Magee D, Ratnalingam R, Kessel D. A training system for ultrasound-guided needle insertion procedures. Med Image Comput Comput Assist Interv 2007; 10: 566–574.
    49. Shams R, Hartley R, Navab N. Real-time simulation of medical ultrasound from CT images. Med Image Comput Comput Assist Interv 2008; 11: 734–741.
    50. Blum T, Heining S, Kutter O, Navab N. Advanced training methods using an augmented reality ultrasound simulator. Int Symp Mixed Augmented Reality 2009; 177–178.
    51. Bürger B, Abkai C, Hesser J. Simulation of dynamic ultrasound based on CT models for medical education. Stud Health Technol Inform 2008; 132: 56–61.
    52. Bommersheim S, Tiede U, Burmester E, Riemer M, Handels H. Simulation von Ultraschallbildern für ein virtuelles Trainingssystemfür endoskopische Longitudinal-Ultraschalluntersuchungen. Bildverarbeitung für die Medizin. 2005; 450–454.
    53. Abkai C, Becherer N, Hesser J, Männer R. Real-time simulator for intravascular ultrasound (IVUS). SPIE Med Imaging Proc SPIE 2007; 6513: 1–10.
    54. Köhn S, Van Lengen R, Reis G, et al.. VES: virtual echocardiography system. In: Visualization, Imaging, and Image Processing. Palma de Mallorca, Spain: IASTED International Conference. 2004; 465–471.
    55. Reis G, Lappé B, Köhn S, Weber C, Bertram M, Hagen H. Towards a virtual echocardiographic tutoring system. Vis Med Life Sciences 2006; 2: 99–119.
    56. Aiger D, Cohen-Or D. Real-time ultrasound imaging simulation. Real-Time Imaging 1998; 4: 263–274.
    57. Terkamp C, Kirchner G, Wedemeyer J, et al.. Simulation of abdomen sonography. Evaluation of a new ultrasound simulator. Ultraschall Med 2003; 24: 239–244.
    58. Maul H, Scharf A, Sohn C. Was kann der Sonotrainer-Ultraschallsimulator? Der Gynäkologe 2006; 39: 870–877.
    59. Markov-Vetter D, Muehl J, Schmalstieg D, Sorantin E, Riccabona M. 3D augmented reality simulator for neonatale cranial sonography. Comput Assist Radiol Surg 2001; 1230: 483–487.
    60. Wüstemann M, Jehle G, Schwerdtfeger R, Mühlhaus K. Fetale echokardiographie am Virtueller Herzultraschall erstmals möglich: Fetale echokardiographie am ultraschallsimulator. Frauenarzt 2008; 49: 46–49.
    61. Jensen J. Field: a program for simulating ultrasound systems. Nordicbaltic Conf Biomed Imaging 1996: 351–353.
    62. Reichl T, Passenger J, Acosta O, Salvado O. Ultrasound goes GPU: real-time simulation using CUDA. SPIE Med Imaging Proc SPIE 2008; 7261: 726116.
    63. Imani F, Reinig K, Spitzer V, Chen Y, Lee C. Simulation of medical ultrasound images using Visible Human Dataset. Visible Human Conference 2002.
    64. Spitzer V, Ackerman M, Scherzinger A, Whitlock D. The visible human male: a technical report. J Am Med Inform Asssoc 1996; 3: 118–130.
    65. Sun B, McKenzie F. Medical student evaluation using virtual pathology echocardiography (VPE) for augmented standardized patients. Stud Health Technol Inform 2008; 132: 508–510.
    66. Staboulidou I, Wüstemann M, Vaske B, Elsässer M, Hillemanns P, Scharf A. Quality assured ultrasound simulator training for the detection of fetal malformations. Acta Obstet Gynecol Scand 2010; 89: 350–354.
    67. Vidal FP, Villard PF, Holbrey R, et al.. Developing an immersive ultrasound guided needle puncture simulator. Stud Health Technol Inform 2009; 142: 398–400.
    68. Bose R, Matyal R, Panzica P, et al.. Transesophageal echocardiography simulator: a new learning tool. J Cardiothorac Vasc Anesth 2009; 23: 544–548.
    69. Kempny A, Piórkowski A. CT2TEE—a novel, internet-based simulator of transoesophageal echocardiography in congenital heart disease. Kardiol Pol 2010; 68: 374–379.
    70. Guirlinger S. Simulation to augment standardized patients in obetetric ultrasound training [PhD thesis]. Norfolk, VA: Old Dominion University, 2007.
    71. Maul H, Scharf A, Baier P, et al.. Ultrasound simulators: experience with the SonoTrainer and comparative review of other training systems. Ultrasound Obstet Gynecol 2004; 24: 581–585.
    72. Marquardt R, Scharf A, Stoiber B. Sonotrainer-Kurse jetzt auch zur Mammasonographie. Frauenarzt 2007; 48: 151–152.
    73. Kahol K, Vankipuram M, Smith M. Cognitive simulators for medical education and training. J Biomed Inform 2009; 42: 593–604.
    74. Issenberg S, Mcgaghie W, Petrusa E, Gordon D, Scalese R. Features and uses of high-fidelity medical simulations that lead to effective learning: a BEME systematic review. Med Teach 2005; 27: 10–28.
    75. Chan W, Ni D, Pang W, et al.. Learning ultrasound-guided needle insertion skills through an edutainment game. Trans o Edutainment IV 2010; 200–214.
    76. Costantino T, Satz W, Stahmer S, Dean A. Predictors of success in emergency medicine ultrasound education. Acad Emerg Med 2003; 10: 180–183.
    77. Lee D, Cottrell J, Sanders R, Meyers C, Wulfsberg E, Sun C. OEIS complex (omphalocele-exstrophy-imperforate anus-spinal defects) in monozygotic twins. Am J Med Genet 1999; 84: 29–33.
    78. Lys F, De Wals P, Borlee-Grimee I, Billiet A, Vincotte-Mols M, Levi S. Evaluation of routine ultrasound examination for the prenatal diagnosis of malformation. Eur J Obstet Gynecol Reprod Biol 1989; 30: 101–109.
    79. Levi S, Schaap J, De Hava P, Coulon R, Defoort P. End-result of routine ultrasound screening for congenital anomalies: The Belgian Multicentric Study 1984-92. Ultrasound Obstet Gynecol 1995; 5: 366–371.
    80. Monsky W, Levine D, Mehta T, et al.. Using a sonographic simulator to assess residents before overnight call. AJR Am J Roentgenol 2002; 178: 35–39.
    81. Bose R, Matyal R, Warraich H, et al.. Utility of a transesophageal echocardiographic simulator as a teaching tool. J Cardiothorac Vasc Anesth 2001; 25: 212–215.
    82. Knudson M, Sisley A. Training residents using simulation technology: experience with ultrasound for trauma. J Trauma 2000; 48: 659–665.
    83. Salen P, O’Connor R, Passarello B, et al.. Fast education: a comparison of teaching models for trauma sonography. J Emerg Med 2001; 20: 421–425.
    84. Reiley C, Lin H, Yuh D, Hager G. Review of methods for objective surgical skill evaluation. Surg Endosc 2011; 25: 356–366.

    Ultrasound simulation; Training; Review

    © 2013 Society for Simulation in Healthcare