Lim BH, T Poston (tim@jhs.com.sg), E Burdet (e.burdet@ieee.org), M Sarimuthu

ABSTRACT

Hypotheses: 1) Hand-eye coordination in the microsurgery environment is separately teachable, and testable. 2) Its learning can be improved by appropriate feedback, and 3) Feedback direct to the motor system is stronger than visual feedback alone. 

Experimental design: We propose to divide a group of 20 incoming microsurgery students into a test group and a control, each of 10.The control group will experience only training on real physical and anatomical materials.The test group will experience digital training in basic motor skills. They will then move on to the same training protocols as the control group.In developing the virtual environment, we will test student volunteers on whether the target motor skills are learned faster with (1) visual feedback alone, (2) sound, or (3) haptic (force-reflecting) feedback, that diminishes with skill level, to differentiate their training value.The results will determine the system used with the microsurgery students.It is expected that the mean time taken to reach the standard of reliably performing five successive patent anastomoses in succession will be less in the test group, and that in the mean they will score higher on appropriate measures of quality. 

We will base the simulations on the hand-eye oriented Magma programming environment from the Swedish company ReachIn, (http://www.reachin.se), together with the ReachIn Display (RID) that integrates stereo graphics with the haptic PHANToM robot arm, with a handle for simulated microsurgical forceps, etc., to be developed locally.

Deliverables: The study will determine the viability of digital pre-training for microsurgery, improve knowledge of feedback modalities in task learning, and create a prototype microsurgery pre-training system to be distributed to hospitals around the world by an industrial partner.The study results and protoype will be delivered one year from the project start. 

Further development: This system is expected to be valuable, and commercially viable, in itself.We will then go to the skills of actual surgery: suturing and tying, on simulated tissues.Digital simulation will reveal much more about a trainee's technique than do physical or animal models, and thus teach faster.Our environment must add simulated tissues, needles, sutures, and their mechanical interactions, computed in real time. (World wide, no suture simulator currently supports knot-tying.) New mathematical algorithms under design in the Digital Medicine Laboratory will make these simulations fast and realistic, and the equipment and expertise acquired in creating and testing the pre-training system will provide an ideal setting for their implementation.We will deploy them commercially by the same pathway.

(A)SPECIFIC AIMS

To develop a digital environment for the simulation of work under a surgical microscope, to teach fundamental motor skills to microsurgery trainees, to test its training value, and to deploy the resulting training module through an industrial partner.The goals of the teaching programme and software are to advance the understanding of motor-control learning, and to speed and improve the teaching of microsurgery. 

Hypotheses

Microsurgery involves a correspondence between what the eye sees, and what the hand does, very different from the one learned in infancy [3].The surgeon must master it as deeply and automatically as the childhood one.We will test the hypotheses that:

·This mastery is separately teachable, and testable

·Its learning can be improved by appropriate feedback

·Feedback direct to the motor system is stronger than visual feedback alone

(B)SIGNIFICANCE

Among all human uses of the hand, surgery demands the greatest control and dexterity: and of all surgery, microsurgery demands the most.A surgical microscope enlarges by 10 to 15 times what the eye sees: but the hands must still make microscopic motions.Some experimental systems [6,8–10,13] map large movements of the hand to smaller motions of a robot arm, filter tremor [12], or both.However, such environments remain costly and are not completely transparent to the user, due to the pixel-limited camera view and the unnatural mechanical and control limits of robots. While these devices may open new possibilities in surgical applications requiring very high accuracy, e.g. eye surgery, surgeons will prefer to work directly with their hands if they can do so, e.g., in hand surgery.Advances in surgical technique require more and more surgeons to master the skills of direct microsurgery.

Before even addressing specific surgical techniques, the microsurgeon in training is faced with two new demands.One is to make very small, very delicate motions: the other is to coordinate these, and control them, with a view not matched for size or angle.What feels like a rotation around a vertical axis looks like one with an axis at 45o, making control difficult.When a child learns visual control of a fork, a pen, etc., her brain builds a deep matching of what the visual cortex sees and what the motor cortex does. An inch or direction is a visual concept, and also a muscular one. Well established reflexes move the fork from a point X to a point Y seen as one inch away, directly above it.

The microsurgical environment violates this well-learned matching.To move an instrument tip between points that seem inches apart, the muscles must smoothly move less than a millimetre.A turn of 30o remains a turn of 30o, but in a changed direction.

With hands under a microscope, the student can attempt to follow instructions like "move the needle along its curve, don't push it straight," but cannot easily judge success in this.At present the needed hand-eye coordination skills are learned simultaneously with the surgical tasks that depend on them, and in less than satisfactory settings.The discussion in [1] explains why the rat model is less than ideal for the novice microsurgeon, with deficiencies in glove stitching, and a poor parallel to the three-dimensional subtlety of human anatomy.

A digital simulation can test the motion, report, and enable immediate improvement. In contrast, real motions can only be assessed by watching them, or by measurement at a late stage in skill acquisition. Digitally, dexterity can be measured: the motions of the tool in the user's hand and the force exerted are recorded, and errors in motion or force are easily detected. Feedback on basic motor skills should come at this early stage, rather than wait until problems are revealed by leaky or blocked anastomoses. The ultimate aim should be to master basic skills digitally, then specific microsurgical skills in a digital simulation of human anatomy, before going to live patients.Spatial understanding is much better correlated with surgical skill than is physical dexterity itself, and [15] shows in the case of a normal hand-eye relation that such understanding can be improved in a virtual environment, and the improvement transferred to reality.This proposal addresses the first stage, for the microsurgical hand-eye relation. We will build on its success to move on to digital anatomy.

The senses in the feedback loop that teaches motor skills can be:the eyes alone, through the visual display; sound reporting degree of accuracy or error [5]; or touch.Visual feedback involves a delay of 200 to 300 milliseconds (see, e.g., [2]), with the loop passing through many more systems.Learning visual control involves learning to anticipate motion through this substantial delay.Auditory reaction times can be longer ([7] reports times of 700 msec., but this may be different in fine motor control than for whole body navigation).Motor reactions (innate or conditioned) have shorter pathways and reaction times[4].This greater locality may also accelerate learning.

A normal microsurgery environment provides very little feedback direct to the hand, since the forces involved in (for instance) suturing are small.In a physical simulation, such as replacing small arteries with plastic tubes, the same holds.A digital simulation, however, where the objects exist as computer data and are displayed to the eye by computer graphics and to the hand by a robot arm, can be set to display physically unreal forces or auditory tones that let the trainee feel or hear whether the motion is correct.As skill improves, the simulation reduces these signals to the vanishing point.We propose to test whether this reduces the time required to reach a measured standard of dexterity.

The technique of microvascular surgery improves the surgeon's armamentarium in a difficult problem, be it an organ transplant or the paediatric transfer of an undescended testis.The anticipated impact of this proposal on health care is improved learning and maintenance of basic microsurgery skills for any medical or veterinary surgeon whose practice is not primarily microsurgery.Procedures large and small in most surgical specialties can require the optomotor skills that this digital environment will teach.These will lead to improved outcomes, less frequent need for surgical re-exploration and salvage of a failed anastomosis, and a broader repertoire of surgical options.

(C)RESEARCH TEAM

Dr Lim is Chief and Senior consultant in the department of Hand and Reconstructive Surgery of NUH.He regularly conducts microsurgical courses in Singapore for advanced surgical trainees nationally and around the region, and is much concerned with the issues of microsurgical skill acquisition.His specialty is in microreconstruction of the upper extremities, which includes toe-to-hand transfers, free flap resurfacing for large skin defects in the upper and lower limbs, and vascularised joint transfers.He is the author of chapters on these techniques in various authoritative reference texts.

Dr Poston has sixty published papers in areas from archaeology to economics to medical imaging, and two textbooks that have remained in print for twenty-four years.He has a strong background in the mathematics of display and simulation, and in visualisation and the reach-in approach to human-computer interaction.His founding paper on this schema was included in the Industrial Electronics Handbook published by IEEE and CRC, and he has numerous publications in this area.In 1994 he helped found ReachIn (http://www.reachin.se) in Sweden, with the support of the Industrial Design school of Umeaa University and seed money from Ericsson, to develop the technology further and deploy it more widely, though he is not a salaried employee.This company now employs about twenty persons, and is shipping hardware and software environments for 3D interaction to many customers internationally.It is also developing applications for design and for surgical simulation.

Dr Burdet is doing research in Robotics, Biomedical Engineering and Human-machine interaction.He joined the NUS in September 1999 as assistant professor in Mechanical Engineering.Dr. Burdet obtained two Masters, in Mathematics and Physics, and a Ph.D. in Robotics from ETH in Zurich, Switzerland. After his studies he realised several Biomedical Engineering projects in Canada and Japan, for which he obtained funding. During 1998–2000, he published 7 papers in international peer-reviewed journals.

Ms Sarimuthu has a background in various areas of information technology, and is the first researcher of the NUH Hand and Microsurgery Dept to work on computer visualisation problems.She has established a strong sense of the relation of computational issues to the quality and relevance of real-world data, working in the laboratory to create datasets more appropriate for computer visualisation.She will acquire a deeper knowledge of simulation and force feedback, in coding work, while serving an essential function in bringing together the digital activities with the world of the real microsurgical laboratory.

(D)METHODS

Collisions and grasping: mathematical mechanics

In the first stage, in parallel with the physical setup of a general-purpose reach-in system and its programming environment, we will formulate mathematical descriptions of the interactions between a curved needle and a microsurgical forceps, including the magnetic effects which complicate the task for beginners.

Components of skill

The mechanics provides the basis for the interactions between this system and the user: we will define precise protocols for the either-hand skills of

·Rotating an object about a visually displayed axis (i.e., a physically different axis).

·Moving within the depth of field of the microscope

the right-hand skills of

·Grasping the needle at an intended angle

·Inserting a needle along its own curve (not driving it straight)

·Pushing a needle in small grasp/release steps

·Pulling an emergent needle along its own curve 

·Pulling a suture through, within the depth of field of the microscope

and the left-hand skills of

·Moving an object to be sutured.

(Standard surgical tools are designed for right-handers.)

Physical environment for simulation

Users will reach their hands into a virtual environment, rotated and scaled in the way that a surgical microscope does.They will see tool tips and objects moving as they do in the microscope, while holding the handles of standard microsurgical forceps.The microsurgeon feels no force when acting correctly, except for the spring resistance of the forceps: the same spring force will be physically present, and a sensor will report the degree of forceps closing to the simulation software.Force feedback from the robot arm will act as a signal of error, as will sound.

Software structure for simulation

We will specify an object-oriented software structure for simulating the above interactions, providing visual/aural/haptic feedback to the trainee, and recording/reporting success rates to the trainee and the trainer.In this process we will develop a storyboard for the user interface (UI), showing the different scenes that the software will display, and listing the results of every user input action (move the robot arm, squeeze the forceps handle, select a menu item, etc.) to which the software will respond by changing the state of the display, or moving to a different scene.The needles, forceps, selectors, menus, etc., appearing in these scenes determine the objects that must be represented in software, together with their interactions.Functional specifications of these software objects will form part of an overall functional specification of the software, together with the way that the software incorporates the Magma environment for programming haptic reach-in applications, developed and provided by ReachIn.(There are complexities in displaying a virtual object simultaneously through a visual screen, refreshed a few dozen times per second, and through force feedback, refreshed 1,000 times per second if touch is to function usefully.The Magma environment provides a general framework for such things, so that the problems need not be solved in inconsistent, fragmentary ways as they arise in a development project.)

The storyboard, design and implementation will include a user interface to control needle size, stiffness and curvature, adjustable within the needle object class, embedded in a higher-level interface by which the trainer or trainee manages the difficulty of the task.

Training sequence design

We will define skill-level criteria such as a sequence of five successes in rotating a narrow cylinder within a tube (controlling axis of rotation, despite the rotated view), and passing a curved needle though a curved hole of given diameter (testing curved motion control).Failure will constitute touching the side: at the training stage, this will be signalled by acoustic or haptic feedback, but tests will provide only visual feedback, as in actual surgery.The interface will promote the trainee to a higher level of difficulty (for instance, a smaller needle) when a particular skill level is reached at current settings, until the level is appropriate for microsurgery.A separate storyboard will specify an interface by which a trainer without programming expertise can modify the sequences of tests and promotion.

Feedback comparison and evaluation

When this design has been implemented, we will recruit 20 undergraduate volunteers to practice these tasks: 5 with visual feedback only, 5 with aural feedback added, 5 with haptic feedback, and 5 with all these sensory modalities.We will both test the speed at which the trainees attain the target skill levels, and re-test later for retention of the acquired skills.We will match the recruited group with survey results on local surgery trainees for skill and enthusiasm in playing conventional computer games, and examine the impact of these factors on skill acquisition scores, using standard statistical tests.

Assessment of medical value

In the next phase of the work under this proposal, we will use with surgery trainees the feedback scheme for which the previous subjects learned most successfully, with the proviso that haptic feedback (as the more costly option) will be chosen only if success rates are at least 30% better than any other scheme tested.In this phase, we will divide a group of 20 incoming microsurgery students into a test group and a control, each of 10.The control group will experience only training on real physical and anatomical materials.The test group will experience training in basic motor skills in a virtual environment, before moving on to the same training protocols as the control group.

Trainees will be scored on the number of attempts needed to reach the standard of reliably performing five successive patent anastomoses in succession, in a standard-sized rat.

It is expected that the attempt number will be less in the test group, and that in the mean they will score higher on appropriate measures of quality.Speed will be recorded, and we will examine its correlation with this training, but it is not a primary criterion of excellence.

(E)REFERENCES

[1] "Advanced Non-animal Microsurgical Exercises", N. L. Crosby, J. B. Clapson, H. J. Buncke & L. Newlin, Microsurgery 16:655–658, 1995.

[2] "Arm Trajectory Modifications During Reaching Towards Visual Targets", T.Flash and E. J. Henis, Cognitive Neuroscience 3(3): 220–230.

[3] "Vision and action: the control of grasping", M. A. Goodale. ed, Norwood, N.J., Ablex Pub. Corp., 1990.

[4] "Essentials of neural science and behavior"E. R. Kandel, J. H. Schwartz, T. M. Jessell, eds, Appleton & Lange, 1995.

[5] "Listening: An Introduction to the Perception of Auditory Events", S. Handel, Cambridge, MA: MIT Press, 1989.See also bibliography in http://www.icad.org/websiteV2.0/Biblio

[6] "Performance of robotic augmentation in microsurgery-scale motions", R. Kumar, T. Goradia, A. Barnes, P. Jensen, L. Whitcomb, D. Stoianovici, L. Auer, & R Taylor, in Lecture Notes in Computer Science 1679: Medical Image Computing and Computer-Assisted Intervention – MICCAI 99: the Second International Conference, pp. 1108-1115, Springer Verlag, Cambridge, England, September 1999.

[7 ] "Testing sensory and motor hypotheses of inhibition of return", H-W. Kwak, http://w3.knu.ac.kr/publication/ir-95.html

[8] "Robot Assisted Microsurgery": http://robotics.jpl.nasa.gov/tasks/rams/homepage.html

[9] "Telerobot Control for Microsurgery", NASA Tech Briefs, Vol. 21, No. 10 (October 1997), (NPO-19823), p 46.

[10] "Force-Feedback Device for Microsurgery" (NPO-19822), NASA Tech Briefs, Vol. 21, No. 10 (October 1997), p. 86.

[11] "Reaction times as an index of visual conspicuity at night", S. Plainis, I. J. Murray, K. Chauhan, W. N. Charman, http://www.umist.ac.uk/optometry/dept/plainis/rtviv8.pdf

[12] "Adaptive canceling of physiological tremor for improved precision in microsurgery," "C. N. Riviere, R. S. Rader, and N. V. Thakor, IEEE Transactions on Biomedical Engineering, 45(7):839-846, July 1998.

[13] "The heart of microsurgery", J. K. Salisbury, Mechanical Engineering (Dec. 1998):http://www.memagazine.org/backissues/december98/features/microheart/microheart.html

[14] "Auditory Guidance With The Navbelt — A Computerized Travel Aid For The Blind",S. Shoval, J. Borenstein, Yoram Koren, http://www-personal.engin.umich.edu/~johannb/Papers/Paper44/Paper44.html.

[15] "The Interaction of Spatial Ability and Motor Learning in the Transfer of Training from a Simulator to a Real Task", M. R. Tracey and C. E. Latham, Medicine Meets Virtual Reality 2001, D. Westwood, H. M. Hoffman, G. T. Mogel, D. Stredney and R. A. Robb eds, IOS Press 2001, pp 521–527.

back to Etienne BURDET homepage

January 2000, by E. Burdet