Prediction Models Applying Convolutional Neural Network based Deep Learning to Cervical Cancer Outcomes
B. ChitraS. S. KumarD. Subbulekshmi1 Department of Electronics and Communication Engineering,Noorul Islam Centre for Higher Education,Kumaracoil,Kanyakumari District,Tamil Nadu,India2 Department of Electronics and Instrumentation Engineering,Noorul Islam Centre for Higher Education,Kumaracoil,Kanyakumari District,Tamil Nadu,India3 School of Electrical Engineering,Vellore Institute of Technology Chennai,Chennai 600127,IndiaB Chitra is awarded the BE degree in electronics and communication engineering in 2005 from Anna University,Chennai,Tamil Nadu,India and ME degree in applied electronics in 2007 from Anna University,Chennai,Tamil Nadu,India. She is a Research Scholar of Noorul Islam Centre for Higher Education,Kumaracoil,Kanyakumari District,Tamil Nadu,India,in the department of Electronics and Communication Engineering. Her area of interest includes medical and image processing and signal processing. She is working as an assistant professor in the department of ECE at Arunachala College of Engineering for woman,Nagercoil,Tamil Nadu,India,and have 14years of total teaching experience. She has published over 7 papers in international journals. E-mail: umar S S is currently working as associate professor in the Department of Electronics and Instrumentation Engineering,Noorul Islam Centre for Higher Education,Kumaracoil. He completed his BE in electronics and instrumentation engineering from Manonmaniam Sunderanar University,ME applied electronics from Anna University and PhD in information and communication from Anna University. He has published several papers in SCI and scopus indexed journals. His areas of interest include medical image processing,satellite image processing,deep learning and multiresolution techniques. E-mail: kumarss@niuniv.com.D Subbulekshmi has received her BE degree in electronics & instrumentation from Manonmaniam Sundaranar University,Tirunelveli,Tamil Nadu in 2001 and her ME degree in process control and instrumentation from Annamalai University,Tamil Nadu,India in 2002. She received a PhD degree in the Department of Electrical and Electronics Engineering from Anna University,Chennai,Tamil Nadu,India in 2013. From 2007 to 2013,she was an assistant professor at PSG College of Technology,Coimbatore. At present,she is a professor in the School of Electrical Engineering,Vellore Institute of Technology,Chennai,Tamil Nadu,India. Corresponding author. E-mail: subbulekshmi.d@vit.ac.in.
DOI: https://doi.org/10.1080/03772063.2024.2305831
IF: 1.8768
2024-01-30
IETE Journal of Research
Abstract:Early identification of Cervical Cancer (CC) helps in reducing the death rate of CC patients. A Pap smear/Pap test, which is also a widely used method, is used to detect CC in its early stages. The automatic detection and categorization of Pap smear images is a difficult task. So, it is important to design and implement a high-efficiency with low-cost screening system. Therefore, this paper compares the efficiency of two deep learning-based cervical cancer screening strategies. The first technique taken for comparison is the Mutation based Atom Search Optimization (MASO) optimized DenseNet 121 architecture. The MASO technique is used to adjust hyperparameters such as learning rate value, batch sizes, and the number of neurons in the dense layer in the DenseNet 121 architecture. The second technique taken is the Sooty Tern optimized CNN-based Long Short-Term Memory (ST-CNN-LSTM) classifier to identify the different types of cervical cancer stages. The areas affected by the cancer are identified using Kernel-Weighted Fuzzy Local Information C-Means clustering (KWFLICM) model. This model removes the nucleus and cytoplasm from the background area of the cell. Normal, mild dysplastic, severe dysplastic, and carcinoma are among the CC classifications offered by these two classifiers. The efficiency of these classifiers is compared in two datasets (Hervel and SIPaKMeD) via different performance metrics such as specificity, precisions, F-score, accuracy, confusion matrix, sensitivity, and recall. The ST-CNN-LSTM offers improved performance in terms of accuracy (99.80%), sensitivity (98.83%), and specificity (99%) when compared to other classifiers.
telecommunications,engineering, electrical & electronic