Search

Cervical cancer has influenced life of women worldwide as the fourth most severe cancer. Early screening, detection and treatment of cervical cancer notably increase the life quality and reduce the death rate of patients. Therefore, automatic diagnosis of cervical cancer could bridge the gap between testing needs and capabilities. Cervical cell detection plays an important role in cancer screening, the intent of this study is to classify the cervical cells through deep learning models, which helps to monitor the patients’ health. SSD (Single Shot MultiBox Detector) network is integrated with the positive and negative features to address the shortcomings of insufficient sensitivity to small objects. Besides, center loss function is added to better address situations that intra-class differences are greater than inter-class differences. A dataset containing 1462 benchmarked cervical cells was utilized. 80% (1167) are used for training and the remaining 20% (295) are allocated for testing. Proposed optimized SSD network achieved the accuracy of 90.8% and mAP (mean Average Precision) of 81.53%, which is 7.54% and 4.92% higher than YOLO (You Only Look Once) and classical SSD, respectively. The addition of complementary features improves the network sensitivity and the overall accuracy. It is also concluded that the proposed SSD network could be applied in cell classification for the early automatic detection of cervical cancer.

Cervical cancer if detected early has an upward of 89% survival rate. The leading tool in identifying cervical cancer in its infancy is the Papanicolaou (Pap smear) test, which since its introduction dropped cervical cancer related deaths by 60%. The Pap smear test or Liquid-based Cytology (LBC) is a time-consuming procedure that requires a pathologist to manually identify cervical cells that may be in the middle of the processes that indicate cervical cancer. Unfortunately, due to the expenses related to conducting the Pap smear test many women are blocked from access to it and this leads to over 4000 women dying annually from cervical cancer in the United States alone. The aim of this research is to automate the methods used by pathologists in conducting the Pap smear or LBC. We show that using machine vision, ensemble learning and deep learning methods a significant portion of the Pap smear can be done automated. We set out to extract cells and cell clusters and classify those samples based on the Bethesda System for reporting cervical cytology. Achieving an accuracy of 90.4% and 91.6% for the ensemble learning and deep learning methods respectively when evaluated with a five-fold cross-validation demonstrates promise in the creation of an automated Pap smear screening test.

In recent years, with the increasing incidence of cervical cancer, it is a tedious and time-consuming task with unsatisfying accuracy to manually recognize the cells. Machine recognition can be a good solution, but it suffers from the difficulty of obtaining precise edges of cells, which directly influence the final recognition accuracy. To improve the recognition accuracy and shorten the time used for cell recognition, an AGVF-Snake (Adaptive Gradient Vector Flow-Snake) model for the extraction of cell edges has been proposed in this paper. Firstly, the cell is initially located by the improved Canny algorithm. Then, the adaptive initial contour model and gradient vector model are used to obtain accurate cell edges. Finally, the PSO–SVM (Particle Swarm Optimization-Support Vector Machine) classifier is selected to recognize the cervical cells. Herlev dataset is used to verify the AGVF-Snake algorithm; the accuracy of two and seven classifications are recorded. Six other classification methods are introduced for comparison. According to the experimental results, the accuracy of two and seven classifications can achieve up to 99%, which are better than other six methods. The experiment results show that the proposed algorithm has obvious recognition advantages, and thus provides an effective methodological framework for the diagnosis of cervical cancer diseases.

To apply topological information to solve a cervical histopathology image clustering (CHIC) problem, a graph based unsupervised learning (GBUL) approach is proposed in this paper. First, the GBUL method applies color features and k-means clustering to carry out a first-stage “coarse” clustering. Then, a skeletonization based node generation (SBNG) approach is introduced to approximate the distribution of cervical cell nuclei. Thirdly, based on the SBNG nodes, multiple graphs are constructed. Next, graph features are extracted based on the constructed graphs. Finally, k-means clustering is used again for the second-stage clustering. In the experiment, a practical Hematoxylin–eosin staining cervical histopathology image dataset with 40 whole-slide imaging images is tested, obtaining a promising CHIC result and showing a huge potential in the cancer risk prediction field.

Considering the problem of traditional cervical cancer detection method that brings high false negative rate (FNR) and high false positive rate (FPR), a new abnormal cervical cells detection method of multi-spectral Pap smear is proposed in this thesis, on the basis of multi-spectral microscopic imaging technology and computer automotive recognition technology. At first, image in a specific wave band is segmented according to the relationship between intensity and spectrum of each pixel. Then, multi-spectral features of each pixel are extracted making use of improved cosine correlation analysis (CCA) algorithm. Combined with the characteristic of each cell’s area, final definition is made. Experiments have proved the new approach could identify abnormal cells efficiently as well as lower FNR and FPR.

COMPUTER programmes which process data obtained by scanning 35-mm films of individual cervical cells are being tested in an attempt to obtain a quantitative definition of the differences between normal and malignant cells. The scanning instrument is of an electro-optical nature and has a digital output1,2. It scans the photographic film in an array of points with each point yielding a number (between 0 and 128) the value of which depends on the emulsion density. The scan resolution can be varied from 4 lines each with 4 points to 256 lines each with 512 points. A typical single line scan (with 128 points) across a cervical cell nucleus is seen in Fig. 1. Normally these data are punched out automatically on paper tape, suitable for input to a Ferranti Sirius computer.