Comparative Analysis Between Training Methods for a Fuzzy Inference System Potentially Applicable to the Assessment of the Health Status of the Spine in Military Personnel

Abstract

The Colombian Army personnel are exposed to long training and work periods, which involve abrupt, sudden and unexpected body movements together with carrying excessive loads on their backs. That leads to the development of spine disorders on the short, medium and long term. It also affects the soldier’s quality of life, and has a great economic impact on their families, the Army and the State due to absenteeism, incapacity for work and specialized medical attention. In traumatology, orthopedics and physiatry, pelvic parameters are used to check the sagittal balance and to evaluate the erect position and the spinal function. These parameters also are important in the prevention, diagnosis and treatment of diseases such as scoliosis, spondylolisthesis, herniated discs and others. This work presents and compares various training methods for a computer-assisted system able to assess and classify spinal health status among military personnel. Different methods were used: diffuse inference system designed from experience, supervised learning obtained from a previous study by the ‘Backpropagation’ algorithm, computational evolution by the simple genetic algorithm and automatic learning algorithms and their combinations in classifiers sets with neuronal networks and fuzzy classifier from other previous researches on this topic. The system was trained with a collection of data taken from the UCI´s automatic learning repository, which includes sagittal balance parameters data (pelvis incidence or PI, pelvis inclination or PT, lumbar lordosis angle or LL, sacrum slope or SS). The system was validated by the evaluation of its computational performance and by diagnostic utility measurements. Results show that the simple genetic algorithm brings the highest performance, is the best solution to the problem, and is an excellent tool to help health professionals to leave the zone of diagnostic uncertainty.