Cervical Cancer Identification Based Texture Analysis Using GLCM-KELM on Colposcopy Data

Early identification of cervical cancer is still being carried out intensively by the World Health Organization (WHO). Some programs for early identification of cervical cancer are carried out in several ways such as pap smears, IVA test, and colposcopy. Examinations such as pap smears require laboratories to identify cancer from a network of cervical cells. IVA test is done using acetic acid fluid, while colposcopy is done by identifying the condition of the vulva in the vagina and recorded into colposcopy photo data. From the colposcopy photos can be identified automatically using Computer Aided Diagnosis (CAD), by utilizing image processing and classifying them using artificial intelligence methods. In this research, early identification of cervical cancer based on cancer stage using texture information on colposcopy images looks at pixel neighbor information using the Gray Level Co-occurrence Matrix (GLCM) method and classifies it using the Kernel Extreme Learning Machine (KELM) method which is a development of the method ELM by adding a kernel to the system. The results showed that using a linear kernel resulted in an error of 78.5%, a polynomial kernel of 87.5% and the best accuracy achievement of 95% using a gaussian kernel with the best neighborhood angle was 45°. This shows that the data is more likely to have a Gaussian distribution with the best reading of the GLCM, using diagonal pixel readings.