ABSTRACT
In the following we will discuss two particular applications of image registration and machine learning in medical imaging: (a) segmentation and (b) biomarker discovery and classification.
2 MACHINE LEARNING FOR SEGMENTATION
The amount of data produced by imaging increasingly exceeds the capacity for expert visual
1 INTRODUCTION
For many clinical applications the analysis of medical images represents an important aspect in decision making in the context of diagnosis, treatment planning and therapy. Different imaging modalities often provide complementary anatomical information about the underlying tissues such as the X-ray attenuation coefficients from X-ray computed tomography (CT), and proton density or proton relaxation times from magnetic resonance (MR) imaging. Medical images allow clinicians to gather information about thesize, shape and spatial relationship between anatomical structures and any pathology, if present. In addition to CT and MR, other imaging modalities provide functional information such as the blood flow or glucose metabolism from positron emission tomography (PET) or single-photon emission tomography (SPECT), and permit clinicians to study the relationship between anatomy and physiology. Finally, histological images provide another important source of information which depicts structures at a microscopiclevel of resolution.
