Detection of cervix tumor using an intelligent system accompanied with PNN classification approach

• Published in: Signal, image and video processing Vol. 17; no. 7; pp. 3873 - 3881
• Main Authors: Vijila Rani, K., Eugine Prince, M., Sujatha Therese, P., Josephin Shermila, P., Anna Devi, E.
• Format: Journal Article
• Language: English
• Published: London Springer London 01.10.2023; Springer Nature B.V
• Subjects: Adaptive algorithms; Algorithms; Cancer; Cervix; Classification; Computer Imaging; Computer Science; Datasets; Edge detection; Image analysis; Image contrast; Image enhancement; Image processing; Image Processing and Computer Vision; Image segmentation; Imaging techniques; Medical imaging; Multimedia Information Systems; Neural networks; Original Paper; Particle swarm optimization; Pattern Recognition and Graphics; Signal,Image and Speech Processing; Tumors; Vision; Asymmetrical triangular function with moving average fuzzy filter; Gray-level co-occurrence matrix (GLCM); Segmentation; Structural features based on tumor region; Classification; Combinational feature; Cervix cancer; Feature extraction; FOS; PNN classifier
• ISSN: 1863-1703; 1863-1711
• DOI: 10.1007/s11760-023-02616-w

In the medical field, image analysis is used to identify tumors. Image segmentation is a critical aspect of image processing. Pap-smear imaging techniques are one of the tools to diagnose the cervix cancer and to detect and identify the malignant and benign tissue in the human body. Here, in this research article, the experimental analysis works the input images are taken from the public Herlev dataset. The proposed work for cervix cancer detection and classification was carried out in four phases in which the first phase was the selection of suitable algorithm for denoising and contrast enhancement the Pap smear image. First, the image is preprocesses using the asymmetrical triangular function with a moving average fuzzy filter. The next step, canny edge detection and a boundary monitoring algorithm are used to isolate the cervix. The adaptive concave-hull algorithm is used to fix the cervical boundary. Then Otsu thresholding based Fractional-Order Darwinian Particle Swarm Optimization concept is used for segmentation of tumor contour. The combination function is then used to derive features such as gray-level co-occurrence matrix (GLCM), FOS, and structural features dependent on the tumor region for the segmented tumor region. Finally, the features are extracted, and then the image is classified as a benign or malignant cervix using the PNN (probabilistic neural network) classifier function. The sensitivity, specificity, and accuracy of the various evaluation standards are used to validate the proposed procedure. At the end of the work, this approach is well correlated with other state-of-the-art approaches on the same image open-access dataset.