Determination of Significant Features to Precancerous Cervical Classification

Feature selection is a process used in the automatic learning that consists in selecting an optimal subset of features of a database to reduce its dimensionality, remove noise and improve performance of a learning algorithm. That is, improve the learning speed, precision prediction (measured by the hit rate) and comprehensibility of the results obtained. The aim of this paper is to apply such techniques of dimensionality reduction on processed image features extracted through textural analysis. The processed images were obtained through the colposcope as part of routine gynaecological examinations. From a practical point of view the authors try to extract patterns from the processed images to classify the existing cervix lesions with diagnostic purposes. The resulting attributes of the image processing were analysed using supervised classification techniques of data mining.