Machine Recognition in Medical Microscopy

Abstract

Some general aspects of machine recognition are discussed. The advantages and limitations of machine systems are compared with those of trained human observers and the problems involved in introducing machines into classification systems designed for human observers are outlined. The present classification system for the cellular components of blood samples is described, with emphasis on the different types of white cell encountered. Cervical samples and chromosomes are described and discussed from the machine recognition point of view. All proposed machine recognition systems in microscopy must satisfy certain technical criteria; these are listed and discussed briefly. A number of earlier attempts to recognise cells are described; the results obtained and the difficulties encountered are outlined. The work of a number of university, medical and industrial research groups actively engaged in this field is described. The likely developments in this field are discussed from the point of view of improvements which may occur in the hardware of scanning and computing and also the better understanding, integration and exploration of machine systems in medical research and in routine medical diagnostic procedures.