Biomedical Engineering Reference
In-Depth Information
1.
INTRODUCTION
The rapid development and proliferation of medical imaging technologies is
revolutionizing medicine. Physicians and scientists noninvasively gather poten-
tially life-saving anatomical information using the images obtained from these
imaging devices. The need for identification and interaction with anatomical tis-
sues by physiologists has led to an immense research effort into a wide range of
medical imaging modalities. The intent of medical image analysis is manifold,
ranging from interpretation, analysis, and visualization to a means for surgical
planning and simulation, postoperative progression of the disease, and intraoper-
ative navigation. For example, ascertaining the detailed shape and organization
of the aortic arch in the abdomen for an aneurysm operation enables a surgeon
preoperatively to plan an optimal stent design and other characteristics for the
aorta.
Each of the imaging modalities captures a unique tissue property. Magnetic
resonance imaging (MRI) uses the heterogeneous magnetic property of tissue to
generate the image [1]. The response to an applied magnetic field is distinctive
for each tissue and is reflected in the image. Doppler ultrasound, on the other
hand, relies on the acoustic scattering property of each tissue [2]. X-ray and
computed tomography (CT) imaging [3] are based on absorption. Functional
imagingmodalities like positron emission tomography (PET) and fMRI (functional
MRI) highlight metabolic activities in the region of interest [4-6].
Although modern imaging devices provide exceptional views of internal ana-
tomy as well as functional images, accurate quantification and analysis of the re-
gion of interest still remains a major challenge. Physicians manually segment and
analyze the images, which is highly time consuming and prone to inter-observer
variability. Accurate, reproducible quantification of medical images is required
in order to support biomedical investigations and clinical activities. As imaging
devices are moving toward higher-resolution images and the field of view (FOV)
is increasing, the size of datasets is exploding. Manual analysis is becoming
more challenging and nearly impossible. Thus, the need for computer-aided auto-
mated and semi-automated algorithms for segmenting and analyzing medical data
is gaining importance.
The variability of anatomic shapes makes it difficult to construct a unique and
compact geometric model for representation of an anatomic region. Furthermore,
many factors contribute to degradation of image quality, which makes the process
of segmentation even more challenging. Although the nature of artifacts may vary
with imaging modality and the tissue concerned, their effect on image quality
is nevertheless detrimental. Figures 1 and 2 the illustrate effects of two different
types of artifacts in the process of image acquisition. In Figure 1 the inhomogeneity
factor makes the middle region of the image darker compared to the top and lower
side. In this case the inhomogeneity factor is a slowly varying intensity gradient
