Identification and reconstruction of bullets from multiple X-rays
AFRIGRAPH '06 Proceedings of the 4th international conference on Computer graphics, virtual reality, visualisation and interaction in Africa
ASM driven snakes in rheumatoid arthritis assessment
SCIA'03 Proceedings of the 13th Scandinavian conference on Image analysis
| Hi-index | 0.00 |
We introduce and test two methods for segmenting bones from wrist-hand radiographs. The segmentation method is needed as a component of an image processing system for defining the bone age of a child on the basis of X-ray image of hand and wrist and the bone segmentation is a basic operation of the recognition process. For this purpose we use two novel segmentation methods: Active Contour Model (Snake) and Active Shape Models (ASM). We discuss the applicability of the methods using a set of typical test cases from clinical practice. It turns out that the ASM method should be preferred when segmenting the fingers and the metacarpal bones. This is because their shape can be predicted with sufficient accuracy. The snake method is preferred when segmenting the wrist and epiphysis of bones due to their greater variation in shape. Both methods need some initial approximation of the contour and ASM utilise also the patterns of a training set. The recognition results were good for both segmentation methods but the ASM was not as sensitive to the initial approximations as the snake method. On the other hand, the former method gave not as accurate segmentations as the Snake. We therefore propose a composite method, which starts the segmentation with ASM and then continues with ACM.