One synthetic shoulder and one cadaver specimen were tested on the new biomechanical shoulder simulator (Figure 1). The new shoulder simulator contains six active pneumatic muscles (DMSP, Festo, Esslingen, Germany) which are connected via UHMWPE-ropes and ball bearing pulleys to the corresponding muscle tendons. Thus the three parts of the deltoid muscle and the rotator-cuff muscles [supraspinatus, infraspinatus + teres minor (combined), subscapularis] can be actively controlled. The advantages of the used pneumatic muscles are the high force density and the inherent compliance, which is needed to fit into the elastic properties of cadaver specimen. Furthermore, two passive muscles are realized with springs (pectoralis major muscle combined with latissimus dorsi muscle and biceps brachii muscle).

Figure 1:

Schematic of the biomechanical shoulder simulator (simplified to 1 artificial muscle).

Adverse the pneumatic muscles have a highly nonlinear behaviour which makes control difficult. Therefore, a nonlinear adaptive controller for force and length was developed on the basis of Zeng and Wan [7]. With an additional compensation for the nonlinear characteristics of the used pneumatic valves (VPWP Festo, Esslingen, Germany), a precise control over a wide range of muscle forces and lengths for a changing and unknown control path, namely the specimen, is expected.

An innovative and completely new approach which is realized in the new developed shoulder simulator and control algorithm, respectively, and which is contrary to the shoulder testing apparatus found in literature, is that motion of the shoulder can be controlled by muscle lengths rather than forces. This is necessary to create controlled and free shoulder-movements but depends on detailed information on movement and specimen specific muscle lengths over time. This information is acquired by a so called “Teach-In” process, where the operator moves the humerus on a desired trajectory, while the muscles are force controlled. During this movement the muscles follow the forced movement, and the control system records the required muscle lengths for the realisation of the specific trajectory. After this procedure the system can use the measured, over the trajectory varying muscle lengths, to replay the shoulder movement without the operator’s guidance.

The shoulder simulator setup is equipped with measurement devices for muscle lengths (WS10SG, ASM GmbH, Moosinning, Germany) and forces (KM30z, ME-Messsysteme, Henningsdorf, Germany), a 6D-force torque sensor (ATI, Apex, USA) for the joint reaction forces and moments, and an optical tracking system (Polaris Spectra, NDI, Ontario, Canada) to record the arm motion. A real-time control system (MicroAutoBoxII, dSPACE, Paderborn, Germany) was used for data recording, control and communication with all included devices.

To evaluate the control concept and to analyse the achievable range of motion (ROM) of different movements, an experiment with a synthetic shoulder joint was initially realized. A simplified shoulder model, made from synthetic bones and elastic rubber bands representing the capsular ligament system (glenohumeral ligament & coracohumeral ligament), is attached to the shoulder simulator. This was done by fixating the inferior part of the scapula in polyurethane (PUR) foam and then mounting it to the motion simulator. Scapulae were mounted in the setup with the neutral plane of the glenoid tilted 10° superiorly, the scapula tilted 20° anteriorly, and the plane of the scapula parallel to the middle deltoid line of action [8]. The muscle and tendon attachments and pulling directions are estimated based on anatomical landmarks and the joint surfaces of the synthetic shoulder are lubricated with petrolatum.

In a second step the experiments are repeated with a human cadaver specimen (Figure 2). One fresh-frozen, upper extremity was obtained from a female donor (age: 85 years). The cadaver was free of disease, and the joint capsule was not vented. The arm was thawed at approximately 20°C for 24 h and kept hydrated with normal saline during preparation. The scapula was exposed from the medial border to the suprascapular notch before being embedded with PUR foam in a rectangular block. Then the fixed shoulder was placed into the motion simulator. The distal end of the humerus is likewise attached to a cylinder where the rigid body of the optical tracking system and the spring representing the biceps brachii muscle is fixed. The artificial muscles are connected by chirurgical suture using fibre wire cords.

Figure 2:

Human cadaver shoulder in the Aachen shoulder simulator.

The desired movement of the arm was then performed manually by the operator while the artificial muscles are on force control and the variable muscle lengths are recorded. After this “Teach-In” procedure the movement of the arm was performed by the active length controlled muscles. That way both synthetic and cadaver specimen were articulated in different characteristic motions (abduction/adduction, internal/external rotation with adducted arm, anteversion/retroversion) through a physiologic ROM.