A Deep Neural Network for Cervical Cell Classification Based on Cytology Images

Abstract

Cervical cancer is one of the most common cancers among women. Fortunately, cervical cancer is treatable if it is diagnosed timely and administered appropriately. The death rate of cervical cancer has been greatly reduced since Pap smear test was applied. However, Pap smear test is a time-consuming and error-prone process. Moreover, classifying cervical cells into different categories is clinically meaningful but also challenging in the field of cervical cancer detection. To address these concerns, computer-aided diagnosis systems with deep learning need to be designed to automatically analyze cervical cytology images. In this study, we construct a deep convolutional neural network with feature representations learned via multiple kernels with different sizes to automatically classify cervical cytology images, named DeepCELL. Firstly, we design three different basic modules of DeepCELL to capture feature information via multiple kernels with different sizes. Then, we stack several such basic modules to form the cervical cell classification model. Finally, we perform a series of experiments to evaluate the proposed method on two cervical cytology datasets: Herlev and SIPaKMeD. Our method achieves the accuracy of 95.628%, precision of 95.685%, recall of 95.647% and F-score of 95.636% on SIPaKMeD dataset, which are the highest among all competing methods. Similarly, our method also achieves satisfactory result on Herlev dataset. In summary, extensive experimental results demonstrate that our proposed method has a promising performance in cervical cell image classification.INDEX TERMS Cell image classification, cervical cell detection, deep learning, neural networks.

CerviFormer: A pap smear‐based cervical cancer classification method using cross‐attention and latent transformer

“…Howeverx, this method introduced high‐dimensional feature vectors, resulting in increased computational complexity 65 ; utilized a combination of global significant value, texture statistical features, and time series features to attain a lower accuracy of 88.47% 66 ; achieved an accuracy of 94.51% on the Herlev dataset by integrating cervical cell morphology with appearance information using GoogleNet. Despite the promising initial results, the task of fine‐grained cervical cell classification through DL remains highly challenging for achieving high precision in diagnosis 67 . developed a deep CNN that utilized feature representations learned from multiple kernels with varying sizes for classification purposes.…”

“…Despite the promising initial results, the task of fine-grained cervical cell classification through DL remains highly challenging for achieving high precision in diagnosis. 67 developed a deep CNN that utilized feature representations learned from multiple kernels with varying sizes for classification purposes. However, the model exhibited a greater proficiency in classifying abnormal cells compared to normal cells on the Herlev dataset, leading to an accuracy of 92.72%.…”

High Precision Cervical Precancerous Lesion Classification Method Based on ConvNeXt

“…Mainstream screening for cervical cancer relies on several typical diagnostic methods: HPV concentration testing [6][7][8], colposcopy and biopsy [40][41][42], and cytology or PAP smear testing [43][44][45]. Among these, HPV detection refers to the use of changes in HPV concentration in patients to diagnose cervical cancer.…”

Enhancing Computer-Aided Cervical Cancer Detection Using a Novel Fuzzy Rank-Based Fusion

“…Kundu et al [11] employed two pre-trained CNN as feature extractors and applied a Genetic Algorithm (GA) for feature selection followed by SVM for classification. The authors in [12], introduced DeepCELL, specifically designed to classify cervical cytology images by learning feature representations through several kernels of different sizes, contributing to its effective image classification capabilities. Vision Transformer (ViT)-based approaches have also shown state-of-the-art (SOTA) results in the medical image classification problems [13], [14], [15], [16], [17].…”

“…In this proposed work, we employed three pre-trained CNNs and fine-tune them to classify the input image into one of the classes. The selection of these pre-trained models is based on a comprehensive literature survey [12], [38], [26] and initial experiments. In the initial experiments, we employed several deep-learning models for the base classifier prediction and considered those models that show higher performance and consistency on multiple executions.…”