To develop a useful surgical navigation system, accurate determination of bone coordinates and thorough understanding of the knee kinematics are important. In this study, we have verified our algorithm for determination of bone coordinates in a cadaver study using skeletal markers, and at the same time, we also attempted to obtain a better understanding of the knee kinematics.

The research was performed at the Medical Simulation Center of Tzu Chi University. Optical measurement system (Polaris® Vicra®, Northern Digital Inc.) was used, and reflective skeletal markers were placed over the iliac crest, femur shaft, and tibia shaft of the same limb. Two methods were used to determine the hip center; one is by circumduction of the femur, assuming it pivoted at the hip center. The other method was to partially expose the head of femur through anterior hip arthrotomy, and to calculate the centre of head from the surface coordinates obtained with a probe. The coordinate system of femur was established by direct probing the bony landmarks of distal femur through arthrotomy of knee joint, including the medial and lateral epicondyle, and the Whiteside line. The tibial axis was determined by the centre of tibia plateau localised via direct probing, and the centre of ankle joint calculated by the midpoint between bilateral malleoli. Repeated passive flexion and extension of knee joint was performed, and the mechanical axis as well as the rotation axis were calculated during knee motion.

A very small amount of motion was detected from the iliac crest, and all the data were adjusted at first. There was a discrepancy of about 16.7mm between the two methods in finding the hip centre, and the position found by the first method was located more proximally. When comparing the epicondylar axis to the rotation axis of the tibia around knee joint, there was a difference of 2.46 degrees. The total range of motion for the knee joint measured in this study was 0∼144 degrees. The mechanical axis was found changing in an exponential pattern from 0 degrees to undetermined at 90 degrees of flexion, and then returned to zero again. Taking the value of 5 degrees as an acceptable range of error, the calculated mechanical axis exceeded this value when knee flexion angle was between 60∼120 degrees.

The discrepancy between the hip centres calculated from the two methods suggested that the pivoting point of the femur head during hip motion might not be at the center of femur head, and the former location seemed closer to the surface of head at the weight bearing site. Under such circumstances, the mechanical axis obtained through circumduction of the thigh might be 1∼2 degrees different from that obtained through the actual center of femur head. During knee flexion, the mechanical axis also changed gradually, and this could be due to laxity of knee joint, or due to intrinsic valgus/varus alignment. However, the value became unreliable when the knee was at a flexion angle of 60∼120 degrees, and this should be taken into account during navigation surgery.

An intelligent bone cutting tool as well as a navigation system is of high potential to provide great assistance for the surgeons in computer assisted orthopedic surgery. In this paper we designed a coordinated controller for the surgical robot to perform bone cutting more safely, easily and fast compared with being performed by manual bone saw. Coordinated control is in an outer control loop and determines suitable parameters of the inner control loop of the robot. The inner control loop is an admittance controller for the master site and a compliance controller for the slave site. Coordinated control consists of three modes, i.e. automated cutting, cautious cutting and automated prevention depending on bone cutting conditions and human intention. In automated cutting mode, the coordinated control will set larger admittance gain and smaller compliance gain to provide an assistant force to the human for completion of bone cutting. In cautious cutting mode, smaller admittance gain and larger compliance gain will be set and a resistant force will be provided to the operator for micro progress of bone cutting. In emergence mode, the robot will stop the cutter going forward.

Experimental result shows that in automated mode of the proposed coordinated control was able to assist bone cutting at the same time to avoid undesired large cutting force and cutter breakage. The moving speed of cutter slowed down as the cutting forces increased due to the cutter hitting harder bone, thus alleviated sawblade bouncing up and achieved less deviation from designed cutting plane. In cautious cutting mode the cutting forces were magnified to be felt by the operator. The operator was able to perform micro progress of bone cutting with intensive monitoring of the cutting forces. This functionality is especially useful as the cutter approaches the critical area where the surgeon regards as dangerous region. The emergent mode was also successfully triggered by calculating the defined apparent admittance. The apparent admittance is more reliable than using the cutting force only in detection of cutting boundary.

A hand's on robot under coordinated control is demonstrated in conjunction with surgical navigation system in computer assisted orthopedic surgery. This paper experimentally showed that the coordinated control can effective provide assistive and resistant forces to achieve safe and accurate bone cutting in total knee arthroplasty.

Clinical assessment of elbow deformity in children at present is mainly based on physical examination and plain X-ray images, which may be inaccurate if the elbow is not in fully supination; furthermore, the rotational deformity is even harder to be determined by such methods. Morrey suggested that the axis of rotation of the elbow joint can be simplified to a single axis. Based on such assumption, we are proposing a method to assess elbow deformity using rotational axis of the joint, and an optimized calculation algorithm is presented.

The rotation axis of elbow in respective to the upper arm can be obtained from the motion tract of markers placed at the forearm. Cadaver study was done, in which three skeletal motion trackers were placed over both the anterior aspect of humerus, as well as distal ulna. Osteotomy was created at the supracondylar region of humerus through lateral approach, and the bone fragments were stabilized with a set of external skeletal fixator, leaving the soft tissue intact. The amount of deformity was created manually by adjusting the position of the distal fragment via skeletal fixator. Ultrasound 3D motion tracking system from Zebris® was used in this study, and the program was developed under the Matlab environment. Cycles of passive elbow flexion/extension motion were carried out for each set of deformity. The data were initially transformed to humerus coordinate, and since the upper arm was not absolutely stationary, its influence on the measured position of ulna was adjusted. With this adjusted data, a best fit plane that would include most of the ulna positions (M_U) within a minimal distance was obtained. The rotation axis was calculated as the normal vector to this plane, and the carrying angle could subsequently be assessed according to the relationship between this axis and the x-axis on the xy-plane as well as on the xz-plane.

Fresh frozen cadaver study was conducted in the Medical Simulation Center at Tzu-Chi University. After adjustment of the raw data to eliminate the influence of humerus motion, the ulna motion could be narrowed down from a band of 10mm to 3mm, with a significant smaller standard deviation. The rotation axis was obtained by the normal vector to the best fit plane. Two different approaches were attempted to find the plane. In the first method, the plane was obtained via least square method from the adjusted ulna positions, and the second method found the plane via RANSAC method. Calculations were repeated several times for each method, and the results showed a variation of 5 degrees in the first method and about 2 degrees in the second method.

Rotational axis can be used to define the 3-dimensional deformity of elbow joint; however, it is difficult to obtain such axis accurately due to hypermobility and multi-directional motion of the shoulder joint. In this study, we have developed another method to assess the elbow deformity using motion analysis system instead of the conventional image studies, and this may be applicable clinically if the facility becomes more accessible in the future. (This research was supported by the project TCRD-TPE-99-30 granted by the Buddhist Tzu-Chi General Hospital, Taipei Branch).

To close the surgeon in the control loop is able to take advantages of the fertile sensing information and intelligence from human being so that the complex and unpredictable surgical procedure can be better handled properly. However, there is also weakness to be strengthened. For example, the motion control of the human being is not as accurate as required. Assisted equipment for such purpose may be helpful to the procedure. Exerting large forces during the cutting process may exhaust the operator and cause fatigue. The operator may need power assistance during the surgery procedure. The response speed of human being's may not be adequate to take immediate action in certain critical situations. The constraints from the limbs and human's attention usually cause a significant delay to react the critical situations. This may lead to serious damage by failing to response immediately. For example, in knee replacement procedure, the shape of the knee joint has to be prepared by performing bone resection procedure. The surgeon cuts the bone at certain position and orientation with the help of the cutting jig from the surgical planning. The cutter cut the bone by inserting vibration saw through the slot of the cutting block. The surgeon has to stop the cutter right after the bone has been cut through by the saw so that no surrounding soft tissue, blood vessel, and even important nerves, be damaged. Currently the tactile sensing from the hand is heavily relied on by the surgeon. He has also to be experienced and dexterous. If he fails to draw back the cutter after the bone being cut through, damages to the patient may occur. Therefore there is need to develop a cutting tool, which is intelligent, knows where it is, and is able to judge what the cutting situation is, and further assists the operator to stop the cutter in case that the operator fails to do so, or too slow to response. The benefit to the patient as well as the surgeon will be significant. Therefore the purpose of the paper is to develop an algorithm to implement such functionality of auto-detection of bone cutting through for a bone cutting tool when the cutter has cut through the bone. By the developed method, the intelligent cutter can effectively detect the cutting through condition and then adopt an immediate reaction to prevent the cutter from going further and to avoid unexpected damage.

An auto-detection scheme has also been developed for assisting the operator in judging whether the cutter has reached the boundary or not. The auto-detection scheme is based on analyzing the cutting force pattern in conjunction with a bilateral force controller for the hand's on robot. The bilateral force controller consists of two force controllers. The force controller on the master calculates the feedrate of the cutter according to the push force by the operator. Meanwhile, the operator can also feel the cutting force by the force controller on the slave site, which revises the feedrate of the cutter by the measured cutting force. The operator can kinesthetically feel the cutting force from the variation of the feedrate. When the cutter breaks through the boundary of the bone, the cutting force will drop suddenly to almost zero. When this special force pattern occurs, an acknowledge signal of bone been broken through will be activated. The subsequent action will be followed to stop the cutter going further. We implement this concept by defining a “cutting admittance” indicating the resistance encountered by the cutting during bone resection at different cutting feedrate. During the bone resection, the cutting admittance varies from cutting through hard or soft portions of bone. As reaching the cutting boundary, the cutting admittance will suddenly increase. A threshold value will experimentally be determined to indicate cutting through condition. Together with pushing force by the human operator, the criterion for cutting-through detection is defined. Once cutting through is detected, the admittance for cutter movement will be set zero. The cutter stops going further and the operator feel like hitting a virtual wall in front.

In this paper we proposed a human robot cooperative operation method by which the robotic system can intelligently detect where the cutter has cut through the bone. Characterisation of bone cutting procedure was performed. This auto detection scheme was developed by analyzing the information of the motion and cutting force information during the bone cutting process, no medical images are required. The auto-detection bone cut through was able to transfer the experiences of human being to quantitative modeling. The developed model has been tested for its applicability and robustness by saw bones and pig's knee joints. Results have shown the virtual wall generated by the real-time bone detection scheme and active constraint control is very accurate and capable of providing a safety enhance module for computer assisted orthopaedic surgery, in particular, in total knee replacement. By this method the robot system can accomplish a safe and accurate bone cutting with a complement of an imageless navigation system and results in a low cost, but safe and effective surgical robot system.