Juli Yamashita

Surgical Skill Evaluation by Force Data for Endoscopic Sinus Surgery Training System

In most surgical training systems, task completion time and error ratio are common metrics of surgical skill. To avoid applying unnecessary and injurious force to the tissue, surgeons must know for themselves how much force they are exerting as they handle surgical tools. Our goal is to develop an endoscopic sinus surgery (ESS) training system that quantitatively evaluates the trainees surgical skills. In this paper, we present an ESS training system with force sensors for surgical skill evaluation. Our experiment revealed that the integral of the force data can also be one of the useful metrics of surgical skill.

A direct deformation method

A new free form deforming interface named direct deformation method (DDM) is presented. DDM allows designers to touch and deform free formed surfaces directly, at any point of the surfaces, even if they are represented only by control points and knot vectors. Users do not need to know such parameters that represent surfaces. The deforming actions of a user are used to calculate new parameters that define the deformed shape of the object. The shape is then reconstructed using the new parameters, allowing the user to visualize his or her deformation. This technique is easy to apply and useful for various form representations.<<ETX>>

Development of 3D-input device for virtual surface manipulation

This paper reports the virtual surface manipulation system with force feedback. The developed system is a Cartesian manipulator with stiffness, toughness, and linearity. The proposed manipulator allows the operator spatially manipulates the virtual space with 6DOF, using a force sensor. The control system of the manipulator realizes smooth movement using velocity control. As an application of the system, the deforming operation of virtual surface is shown in an experiment. This paper describes the mechanism and the control method of the manipulator with adaptive damper using network architecture learning through backpropagation technique.<<ETX>>

The chicken egg and skull model of endoscopic endonasal transsphenoidal surgery improves trainee drilling skills.

BackgroundWe verified the effectiveness of training in endoscopic endonasal transsphenoidal surgery (eETSS) techniques using chicken eggs and a skull model. MethodsWe verified the area of eggshell removed by drilling when five residents and four experts used the chicken eggs and a skull model.ResultsWhen residents performed drilling on 10 eggs, a mean (± standard deviation [SD]) area of 31.2 ± 17.5 mm2 was removed from the first egg, and 104.8 ± 3.3 mm2 from the tenth and final egg, representing an increase in area and a decrease in SD. The experts performed the same drilling operation on a single egg, and removed a mean area of 257± 31.7 mm2. These results demonstrated that skills improved as a result of this training, and suggested that this method was also capable of overcoming the initial individual differences in the amount of force applied and ability. An obvious difference between residents and experts was seen in the area removed (p = 0.00011); however, this was attributed to differences in endoscopic manipulation, rather than drilling skill.ConclusionOur findings suggest that this training method could be adequate for acquiring eETSS techniques. Although experts showed superior endoscopic manipulation, residents may also be able to acquire adequate endoscopic skills through further training, and our training method appears to offer an effective means of improving eETSS techniques.

Development of 3D-input device using adaptive control

This paper introduces the 3D-input device with adaptable control for human characteristics. For the problem of low response of the human motor system and differences of impedance between man and machine, the human friendly device is proposed with adaptive and predictive control using neural network architecture and machine impedance control according to the impedance of a human hand. The proposed device is evaluated using a comparison between the tracing resistance of virtual surface and real surface (ice, steel, and wood).

Feature based haptic rendering: architecture, protocol, and application

3D I/O devices, long awaited in Computer Aided Design (CAD) systems, are finally available as force feedback devices such as PHANToM (SensAble Technologies, Inc.). This paper presents Feature Based Haptic Rendering (FBR), which provides haptic feedback of a more accurate shape required by the haptic man-machine interface for surface CAD than intermediate representation (IR) [1][2]. Fig. 1 shows the FBR architecture. Virtual surface modelling process is connected to a haptic rendering loop using a single asynchronous TCP/IP socket stream. As [2] states, decoupling haptic rendering process is indispensable to obtaining high haptic fidelity that requires device control latency of ~kHz, since shape modelling, e.g. NURBS knot insertion, may take time. IR is data exchanged between decoupled processes, which actually is a plane in [1] and to which [2] added a few more primitives. Because only received IR is used in haptic rendering, the haptic process does not need a geometry database. The database coherency problem that arises with such architecture, in that each process has its own geometry database copy, can thus be avoided. FBR generalizes the idea of IR to a feature, which is the unit of communication and can be categorized into two groups -data and commands. The data feature includes local haptic properties of virtual objects, such as shape (Fig. 2), texture, stiffness, and position. Command feature is used to control other processes, e. g., switching interpolation strategies of received shape features in the haptic rendering loop. FBR provides a communication protocol to exchange features. A feature is split into one or more fixed-length packets from which the receiver reconstructs the original feature. The haptic renderer has feedback force calculation algorithms and interpolation strategies for features. FBR can be used by any point-contact type force feedback device as IR and can be easily extended to multiple-contact type devices by introducing an object momentum feature. Haptic volume rendering is also easy: shape features can be constructed as local isosurfaces by the Marching Cube algorithm [3], for example. Physically correct haptic rendering of a thin, elastic object is then possible, which is important for surgical simulation but is difficult with the voxel density field algorithm [4]. A virtual surface modeler with force feedback has been implemented with FBR, whose modelling and graphic rendering run on a SGI Indigo2 Elan (200 MHz R4400) and the haptic display is PHANToM (6 DOF input and 3 DOF force output) on a PC (200 MHz Pentium Pro). The user can touch, trace, and deform a B-Spline and a triangular surface directly [5] with virtual tools catching haptic feedback (Fig. 3). Rich shape features display the surface shape accurately. A 5-10 kHz device control latency has been achieved with 20-30 Hz communication and graphics refresh rate. Quantitative and qualitative evaluation of FBR using psychophysical methods are in work.

Self reflection can substitute eye contact

In video communication systems, gaze is an important topic and is widely being studied. Unlike other studies, displaying users reflections on the video screen we stopped imitating the gazes during face-to-face conversation but tries to solve the problem by providing an environment that enables other expressions to substitute for the roles of gazes during conversation. The relative positioning of their self reflections among other reflections (RPAR) serves as gaze and helps smooth communication, which was experimentally verified.

Perceptual performances of force rendering methods

In virtual reality techniques, the force feedback method becomes important especially when manipulating shapes interactively. This paper describes two methods for presenting haptic stimuli, and compares their perceptual effects. One of the methods involves generating undulations by spatially varying the positions where reaction forces are generated while keeping their directions constant, and the other method involves spatially varying the directions of reaction forces while keeping the positions constant. The subjective intensities of undulations generated by the two methods were compared. The results indicated that the subjective intensities depend on their spatial frequencies. Then, two stimuli were superimposed 180 degrees out of phase to check if the two stimuli generated by different methods offset each other. No offset was observed regardless of the spatial frequencies of the stimuli.