No abstract

No abstract

No abstract

Detecting and categorizing Pap smear images automatically has been a difficult task over the past five decades. There is a significant need for the design and implementation of low‐cost, high‐efficiency screening systems. To address these issues, we have proposed a novel approach for cervical cancer classification. At first, each image is pre‐processed and cropped into 2043*1362 pixels. These images are separated into 681*681 pixels with six square patches. Image augmentation techniques such as image enhancement, image flipping, and image rotation are used to reduce the number of parameters necessary for subsequent processes. The cancer‐affected regions are correctly segmented with the help of kernel weighted fuzzy local information c‐means clustering (KWFLICM) model. The KWFLICM model mainly extracts the cytoplasm and nucleus present in the cell from the background region. The features of the cervical image are extracted using a convolution neural network (CNN), and these retrieved features, such as shape, color, and texture, are crucial for classification. Finally, the Sooty Tern Optimization (STO) algorithm with CNN‐based long short‐term memory classifier (CNN‐LSTM) identifies the four types of CC: normal, light dysplastic, severe dysplastic, and carcinoma. The images for the experiment were obtained from Herlev University Hospital in Denmark, and the software used for the implementation was MATLAB. According to the comparative analysis, the proposed CNN‐LSTM with STO algorithm demonstrates better results using various measures such as accuracy, specificity, sensitivity, and F‐score. Depending upon the experimental results, the proposed CNN‐LSTM with STO algorithm provides 99.80% accuracy, 99% specificity, 98.83% sensitivity, and 97.8 F‐score. The proposed methodology offers an improvement of 28.5% and 19.46% when compared with the random forest and ensemble classifier.

No abstract