Biomedical image processing in pathology: a review.

Pathologists make a diagnostic decision by viewing a specimen and measuring various diagnostically important attributes of an isolated object such as size, shape, darkness, colour and texture. This is a complex process. In recent years, computer-aided image processing and analysis systems have played a significant role in quantitative pathology. This paper summarises basic image processing and analysis techniques and reviews related work in pathology and cytology based on computational image processing since 1987. Firstly, we present a general introduction to image enhancement, segmentation, morphometry and visualisation for those medical colleagues who may not have the necessary background in this area. (The mathematical treatment is kept to minimum and appropriate references are cited to satisfy the more mathematically oriented readers. Selected examples are provided to demonstrate the effects of various basic image processing algorithms on a MRI scan. It should be emphasised that the reviewed techniques are generally used as preprocessing steps in analysing microscopic images and powerful algorithms are more sophisticated and problem-specific.) Secondly, we review image cytometric and histometric methods, standards, calibration and applications. Finally, we touch upon three dimensional confocal image processing and analysis, applications of artificial neural networks, and optical disk database management for recording and retrieving a large number of digitised high resolution images. The development of integrated optical microscope and computer, systems is also briefly described.