Article: Supervised classification approach for cervical cancer detection using Pap smear images Journal: International Journal of Medical Engineering and Informatics (IJMEI) 2022 Vol.14 No.4 pp.358 - 368 Abstract: Cervical cancer is found in women and is the global life threatening problem. Papanicolaou test is the well-known technique used for diagnosing the cancer at the early stage. However, the pathological screening is manual, tedious and time consuming process. Therefore, the proposed method employs adaptive fuzzy k means clustering to segment the cell containing nucleus and cytoplasm from the unwanted background from the pathological Pap smear image. Thereafter, the 40 features are extracted from the segmented images based on the shape, size, intensity, orientation, colour, energy and entropy of nucleus and cytoplasm individually. Finally, the performance of supervised classification approach utilising multilayer perceptron with three kernels and support vector machine with five different kernels as the classifiers to predict the cancerous cells is at par with the existing techniques. The classifier is trained and tested on benchmark database with 280 Pap smear images. The performance of these two classifiers are evaluated and it is found that the MLP classifier with <i>hyperbolic tangent</i> activation function outperforms SVM classifier in all the performance criteria, with classification accuracy of 97.14%, sensitivity of 98%, specificity of 95% and positive predictive value of 98%. Inderscience Publishers - linking academia, business and industry through research

Title: Supervised classification approach for cervical cancer detection using Pap smear images

Authors: Pallavi V. Mulmule; Rajendra D. Kanphade

Addresses: Department of E and TC, D.Y. Patil Institute of Technology, Pimpri, Pune, 411018, India ' JSPM's Jayawantrao Sawant College of Engineering, Survey No. 58, Indrayani Nagar, Handewadi Road, Hadapsar, Pune-411028, Maharashtra, India

Abstract: Cervical cancer is found in women and is the global life threatening problem. Papanicolaou test is the well-known technique used for diagnosing the cancer at the early stage. However, the pathological screening is manual, tedious and time consuming process. Therefore, the proposed method employs adaptive fuzzy k means clustering to segment the cell containing nucleus and cytoplasm from the unwanted background from the pathological Pap smear image. Thereafter, the 40 features are extracted from the segmented images based on the shape, size, intensity, orientation, colour, energy and entropy of nucleus and cytoplasm individually. Finally, the performance of supervised classification approach utilising multilayer perceptron with three kernels and support vector machine with five different kernels as the classifiers to predict the cancerous cells is at par with the existing techniques. The classifier is trained and tested on benchmark database with 280 Pap smear images. The performance of these two classifiers are evaluated and it is found that the MLP classifier with hyperbolic tangent activation function outperforms SVM classifier in all the performance criteria, with classification accuracy of 97.14%, sensitivity of 98%, specificity of 95% and positive predictive value of 98%.

Keywords: cervical cancer; Pap smear stain; pathological images; radial basis function; RBF; multi-layer perceptron; neural network.

DOI: 10.1504/IJMEI.2022.123930

International Journal of Medical Engineering and Informatics, 2022 Vol.14 No.4, pp.358 - 368

Received: 18 Apr 2020
Accepted: 11 Oct 2020
Published online: 05 Jul 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Supervised classification approach for cervical cancer detection using Pap smear images

Keep up-to-date