This item is available under a Creative Commons License for non-commercial use only
Biological cell analysis has, and is still, an important aspect in medical research and clinical diagnosis. Although cytologists routinely undertake a diagnosis using optical mi-croscopy, human factors make this routine unreliable especially when it involves many consecutive tasks that are monotonous, time consuming and focus on pattern matching tasks where the patterns concerned are not always entirely clear and/or do not necessarily belong to a well defined class. Raman Spectroscopy provides the potential to generate a fundamental representation on the status of cellular conditions through the characteristics of a Raman spectrum generated by the back-scatter of a laser pulse incident on the cell nucleus. However, this approach requires the nucleus of the cell to be accurately targeted from a complex of many hundreds of such cells within a conventional optical field of view as defined by the resolving properties of a microscope. This requires specialist digital image processing methods to be developed and in this paper we discuss a new approach to the processes of object detection, recognition and classification for target detection in cytology using Raman Spectroscopy. In particular, we report on a system designed for the inspection of slides used in a cervical cancer screening system known generally as a ‘Pap-smear’ test. After providing a short introduction to the pattern recognition in general, we present a unique procedure for automating the targeting process based on an analysis of the principal issues associated with object recognition which include the basic model used and segmentation algorithms derived from the model.
Blackledge, J., Dubovitskiy, D., Lyng, F.:Targeting Cell Nuclei for the Automation of Raman Spectroscopy in Cytology. ISAST Transactions on Computers and Intelligent Systems, Vol.4, issue 1, 2012, pg.42-51.