Active Shape Modeling of Medical Images for Vertebral Fracture Computer Assisted Assessment System

Abstract

This paper describes the development of a computerized method for detection of vertebral fractures by measuring the shape and appearance of vertebrae on cervical x-ray radiographs. The goal of our work is to assist radiologists' image interpretation and thus allow the early diagnosis of osteoporosis. To achieve this goal, we propose a software system known as AVFAS, short for Automatic Vertebral Fracture Assessment System. Our proposed system consists of three modules which we named as the Medical Image Training & Verification System (MITVS), Medical Image Measurement & Decision System (MIMDS) and Medical Image Registration System (MIRS). The MITVS deals with the extraction of feature vectors from example images, training the extracted feature vectors and verifying the similarity between the extracted feature vectors based on the Active Shape Model (ASM) and stores the result as text file. The MIMDS module provides the interface for image pre-processing procedures which includes region of interest (ROI) selection, image enhancement, similarity matching between features extracted from the query visual and those stored in the database. Finally, the MIRS module provides method to combine patient data via text file and image queries. The generated text data will include the patient data acquired from the Second National Health and Nutrition Examinations Survey (NHANES II) as well as the result of the vertebra image analysis and recognition process.