Automated vision system for skeletal age assessment using knowledgebased techniques

摘要

This paper presents a knowledge-based automated vision system tosegment bones in a child's hand radiograph image, and to determinegrowth progress using decision theoretic approaches. A hierarchicalknowledge-based localisation scheme is used to localise bones in thehand radiograph image. Bone contour detection is then implemented withfurther knowledge represented by active shape models (ASM). Hence a setof parameters is generated to describe the bone contour shape. The boneimage is parameterised to describe its texture which is correlated togrowth age. Regression and Bayesian methods are then used to model thecharacteristics of the most correlated shape parameters to the growthage as well as texture parameters in a training set. The models arefinally applied to test images to estimate their bone ages. The Bayesianmethods result in an 8.93% average relative error