Development of a multi-sensor and artificial intelligence based dexterity assessment system

Dexterity tests are needed in the diagnosis and treatment stages of diseases such as carpal tunnel syndrome, cerebral palsy, Parkinson's, arthritis, in the control of professional competencies in the employment of personnel in jobs requiring manual dexterity and in ergonomics studies. Various methods have been developed to determine motor skills and they consist of many stages and tasks. These tests are carried out under the supervision of an expert. This leads to patients being given appointments days later, especially in overpopulated countries. In addition, various errors may occur due to the human factor in the measurements. The aim of this thesis is to automatically evaluate upper limb motor skills with high accuracy using multiple sensors and artificial intelligence algorithms, to present the results to experts, and to reduce the workload and measurement errors. Upper limb motor skills consist of many components such as joint structure, muscle strength, sensation, mobility and coordination. Measurement and determination of motor skills require detailed testing of these components. In this thesis, hand motor skill performances of individuals were determined automatically with multiple sensors and artificial intelligence algorithms. In the performance determination phase, muscle contraction levels and movement classification of the individuals were made and hold-release movement times during the test were determined. The angular velocity, linear acceleration and velocity change values of the limb were measured and recorded, and the total test time and the insertion and rotation times of each of the disks involved in the test were calculated by image processing. All these data were presented to the experts through the user interface. In this way, the workload of the experts was reduced, the number of daily patient admissions was increased, and objective and accurate measurements were made. In addition to the test completion time, which is used as an evaluation parameter in existing test methods, additional metrics were also presented to the experts. The performance of the system was tested with 20 healthy participants. As a result of these tests, the total test time, disc hold and release times, placement and rotation times of each disc on the test setup, and limb dynamic parameters were obtained and evaluated. As a result, it was demonstrated that the system can perform hand dexterity assessment automatically.