Biomedical Engineering Reference
In-Depth Information
so-called progression analysis. All experimental statistics indicate very encourag-
ing results.
The organization of this chapter is as follows. The related work with the
level set method is reviewed in Section 2. In Section 3 the overall system and data
acquisition are briefly introduced. Section 4 discusses the theory of front evolution
for boundary estimation of lesions. Section 5 presents a novel CAD system based
on the level set framework, which is proposed to classify breast masses in 2D
continuous strain ultrasound. We then offer our conclusions.
2.
GENERALULTRASOUNDIMAGEANALYSISARCHITECTURE
Image analysis in today's world carries a lot of significance, since it is the key
to supplying information to clinical practitioners. The value of an image analysis
system is enhanced when used for characterization of tissues in breast cancer.
The breast radiologist is increasingly more keen to find if breast lesions are
benign or malignant. This provides them with important feedback and helps in
deciding whether or not to perform a biopsy. However, classification of a breast
lesion is not an easy task, as clinicians do not see the lesion's anatomy during
the decision-making process. They are making an interpretation based on image
data acquired using ultrasound data, as shown in Figure 1. Thus, they urgently
need a reliable tool that can help automatically classify lesions and provide a
course of treatment that will bring comfort to the women. This chapter is all about
developing such a reliable tool. One such tool is a general-purpose image analysis
ultrasound architecture.
As depicted in Figure 1, the key to this scheme is the ultrasound image analysis
system. Such a system must have a preprocessor, whose major role is to remove
system noise during the physics-based reconstruction process. Along with this is
the segmentor, a subsystem in itself that uses a deformable model-based frame-
work for capturing the topology of breast lesions. Note that segmentation of breast
lesions is only useful if we have classified the segmented lesions. Thus, we need
to extract the features of these lesions and feed them to the classifier, which un-
derstands them using their training ability and classifies them into benign and
malignant lesions. This is all about a software-based ultrasound image analysis
system. Finally, the most important and critical component is the validation sys-
tem, which helps to design systems for clinical acceptance. This can all be seen
in Figure 1.
3.
ULTRASOUND DATA ACQUISITION SYSTEM
We now offer an overview of a computer-aided diagnosis system for breast
ultrasound. We then describe data acquisition in 2D and 3D.
