Computer Assisted Diagnosis in Histopathology

Histology is the study of the microscopic anatomy of cells and tissues of organisms. Histological analysis is performed by examining a thin slice (section) of tissue under a light (optical) or electron microscope (Jungueira & Carneiro, 2005, Kiernan, 2008, Ross et al., 2002, Mills, 2006). After a sequence of technical procedures for tissue preparation (i.e., fixation, dehydration, clearing, infiltration, embedding, sectioning, and staining), histology images can be produced by different imaging techniques (Murphy, 2001), based on which manual or automated analysis can be conducted to detect diseased tissues. In present research, the study of histology images is commonly regarded as the gold standard for the clinical diagnosis of cancers and identification of prognostic and therapeutic targets. In practice, histopathologists visually examine the regularities of cell shapes and tissue distributions and determine cancerous regions and malignancy degree. In addition, histology analysis may be applied to connect the tissue study to gene function analysis at molecular level. For example, in the research of pleiotropy (the association of multiple phenotypes with a single gene), histology analysis provides a potentially powerful tool to study functional genomics by exploring various phenotypic traits. Histopathological study has been conducted for numerous cancer detection and grading applications, including cervix (Keenan et al., 2000, Guillauda et al., 2004, 2005, Price et al., 2003), prostate (Monaco et al., 2009, Naik et al., 2007, Doyle et al., 2007), breast (Doyle et al., 2008, Naik et al., 2008), and lung (Kayser et al., 2002, Schmid et al., 2006, Jütting et al., 1999) cancer grading. A complete review of all these applications is beyond the scope of this chapter. In this chapter we use cervical intraepithelial neoplasia (CIN) classification as an example to illustrate common histopathology image analysis functions. For other cancers, interested readers are referred to the references. Overall, these applications share similar features in the computer techniques that are applied to support clinicians by automatic histology image feature extraction and categorization. Medical image processing and analysis in radiology (e.g. X-ray, ultrasound, Computed Tomography (CT), and Magnetic Resonance image (MRI)) and cytology domains have been major research fields for several decades, and numerous systems (Bankman, 2008, He, 2009, Greenberg, 1984) and software platforms1 2, 3 (Ibanez & Schroeder, 2005) have been developed for these domains. The application of these systems to histology analysis is not straightforward due to the significantly different