Discussion: Automatic scanning for cervical smears
I. M. P. Dawson · Journal of Clinical Pathology · 1969
This leads naturally to the next question, on which Spriggs has already touched, 'If I want full automation what am I going to program my machine to read?' Alternatively, 'Are there differences between a normal and an abnormal cell, and between different patterns of abnormality which a machine can recognize?' We are fond, in discussing the differentiation of, say, malignant squames from inflammatory, of talking about 'subjective impressions' by which we mean the visual comparison of a particular cell with a memory impression of previous cells whose nature we know. My own feeling, with which I know many disagree, is that one's 'subjective impressions' are based on simultaneous assessment of a large number of parameters, such as nuclear size, density of staining, nuclear longest chord, nuclear-cytoplasmic ratio, etc, each of which is capable, at any rate theoretically, of being measured by a machine. Automated recognition of cells should therefore be possible provided that one uses a sufficiently wide multiparameter approach, a view which, if I have understood his paper rightly, Dr Rutovitz would share, at any rate for chromosomes. But if we do this, we must rid ourselves of the idea that the parameters must necessarily all be of visual type or that the method of presentation of material to the automated scanners must necessarily be in a form which can also be used for visual assessment. This many cytologists find hard to accept-but you cannot have it both ways! If the proposition of full automation is accepted one final question remains, 'Is it worth it?' This has to be argued not only in terms of money, because no one will buy a machine costing £1,000,000 to do the work of two technicians, but also in terms of time; no one will buy a machine which takes a week to scan a smear. However, it has also to be argued in terms of interest. The first two can probably be overcome but the third, though I am an advocate of total automation in cervical cytology, still disturbs me. I cannot help being reminded of two lines in T. S. Eliot's 'The Rock'