Canonical representations for tissues and textures
Judith M. S. Prewitt · 2005
Histologic evidence revealed by the light microscope has remained a cornerstone of medical diagnosis, and presents a significant visual in formation processing burden in clinical medicine. Computer technology can obviate the personal equation of traditional visual microscopy, and may lead to medical decision-making systems having a reliability, consistency and accuracy that can earn for them the rubric, "intelligent machine." This paper lays theoretical foundations for computer-based automation of histology. Graph-theoretic and syntactic models for tissues and other natural textures, and statistical models which complement them, provide canonical representations for analysis, discrimination, and synthesis of histologic patterns. Examples are selected from epithelium and muscle.