Quantitative analysis of a class of biomedical images by an image processing system.

Abstract Two fundamental processing components typically characterize image processing systems designed for the analysis of biomedical images. The first is a hardware/software facility that processes an image so as to store it in the computer in a digital form. The second component consists of computing software that operates upon the stored digital image to accomplish the processing aspirations of the system. It is the complexity of these stated aspirations together with the nature of the given image class that constrain any processing philosophy and methodology adopted. From such a viewpoint, an image processing system in use at the University of Western Australia is discussed. The class of images in this system is a set of skeletal muscle tissue images in transverse section. The position is taken that the evaluation of image data in the form of such biological scenes requires quantitative analysis of the components of those scenes as an intermediate step. A descriptive approach to scene analysis is asserted and adopted as the appropriate processing methodology. Consequently a simple data description is defined and implemented in order to deliver to subsequent processors, instances of image entities related under the description. Methods for the quantitative investigation of these entities (muscle fibres and muscle fibre nuclei, for the example class) are finally discussed and some initial results presented.