Towards the automated prescreening of cervical smears

The authors present a summary of the results of 4 preliminary investigations into techniques for the automated prescreening of cervical smears based on a database of expertly classified cervical cell images: (1) A comparison of spatial domain features used to classify normal/abnormal cervical cell images; (2) Automatic segmentation of the nucleus in cervical cells; (3) Automatic classification of cervical cells, using back propagation neural networks, based on features extracted from the cell's frequency spectra; (4) Automatic classification of cervical cells, using competitive neural networks, based on the cell's frequency spectra.<
>