Development of Intelligent Algorithm to Find Osteoporosis Susceptibility from Orthopantomogram

Abstract

Working of a person, animal or plant is affected by a disease or illness preventing the body or mind from its normal working. A member of a population who is at risk of becoming infected by a disease is a susceptible individual. Finding disease susceptibility and generating an alert in advance, is valuable for an individual. newlineThe present research project Development of Intelligent Algorithm to Find Osteoporosis Susceptibility from Orthopantomogram aims to analyze osteoporosis disease indications using soft computing algorithms. newlineOrthopantomogram (OPG) are dental X-ray images. OPG gives complete view of mouth including teeth and bones of both the jaws. Dental surgeons usually recommend OPG images to diagnose problems like bone abnormalities, fracture in upper or lower jaw bone, locate impacted teeth, cysts, infections or tumors. The OPG images also give a fare orthodontic assessment and present condition of earlier dental implant. newlineAs per the National Institutes of Health, signs found in OPG could be related to systemic disease with gums disease called as periodontal disease. The research work presented in this report focuses on a systemic disease: Osteoporosis. newlineMedical professionals working in dentistry from several countries have carried out a survey which pointed out that osteoporotic condition could be inferred from detailed and experienced study of the OPG images. The OPG images of various age groups were studied to find signs like sharpness of mandible cortex mandible width. In these surveys, these observations were analyzed visually by experienced medical practitioners and the findings were compared with respective Dual Energy X-ray Absorptiometry results. newlineThe algorithms developed in the present research work makes use of Image Processing (IP) techniques. It also uses Artificial Neural Network (ANN) as a soft computing learning tool. Algorithms developed during this research innovatively uses IP processes to achieve automatic segmentation of a mandible bone from lower jaw observed in digital OPG X-ray image. Gray Level Co-occurrence Matrix is utilized to perform statistical texture (consistency) analysis of segmented mandible. Structural features like connected components and line counts at various angles are considered to generate structural feature vector from segmented mandible. Combination of these statistical and structural features is used as input to ANN. Unsupervised learning algorithms, Self Organizing Map (SOM) and competitive ANN are used to cluster the data in to two clusters as normal (healthy) and osteoporotic. Feed forward ANN is also used to classify input vector in to two classes as normal and osteoporotic. The results of both the learning algorithms are compared. The results are verified from professors of dental college. newlineAs there is constant increase in number of osteoporosis patients and nature of osteoporosis is irreversible, prevention of osteoporosis is important than its cure. Automatic detection of osteoporotic condition from OPG image is an area of research which is presented in this report. This research will help patients to locate osteoporosis disease condition. newline newline newline