Automatic model for cervical cancer screening based on convolutional neural network: a retrospective, multicohort, multicenter study

Abstract

Background The incidence rates of cervical cancer in developing countries have been steeply increasing while the medical resources for prevention, detection, and treatment are still quite limited. Computer-based deep learning methods can achieve high-accuracy fast cancer screening. Such methods can lead to early diagnosis, effective treatment, and hopefully successful prevention of cervical cancer. In this work, we seek to construct a robust deep convolutional neural network (DCNN) model that can assist pathologists in screening cervical cancer. Methods ThinPrep cytologic test (TCT) images diagnosed by pathologists from many collaborating hospitals in different regions were collected. The images were divided into a training dataset (13,775 images), validation dataset (2301 images), and test dataset (408,030 images from 290 scanned copies) for training and effect evaluation of a faster region convolutional neural network (Faster R-CNN) system. Results The sensitivity and specificity of the proposed cervical cancer screening system was 99.4 and 34.8%, respectively, with an area under the curve (AUC) of 0.67. The model could also distinguish between negative and positive cells. The sensitivity values of the atypical squamous cells of undetermined significance (ASCUS), the low-grade squamous intraepithelial lesion (LSIL), and the high-grade squamous intraepithelial lesions (HSIL) were 89.3, 71.5, and 73.9%, respectively. This system could quickly classify the images and generate a test report in about 3 minutes. Hence, the system can reduce the burden on the pathologists and saves them valuable time to analyze more complex cases. Conclusions In our study, a CNN-based TCT cervical-cancer screening model was established through a retrospective study of multicenter TCT images. This model shows improved speed and accuracy for cervical cancer screening, and helps overcome the shortage of medical resources required for cervical cancer screening.

Detection of Cervical Cancer Cells in Whole Slide Images Us-ing Deformable and Global Context Aware Faster RCNN-FPN

“…The most important findings of our study are as follows: (1). There are currently few network architectures designed specifically for cervical cancer cell detection in Pap smear images [42,[44][45][46]. Based on one of the currently top-performing detectors faster RCCN-FPN [57]…”

“…As discussed above, Images with higher magnification are strongly favorable for the detection of cervical cancerous cells. A recent study [44] based on the Faster RCNN model used TCT images of 400x and achieved AUC=0.67 with IOU threshold=0.5. which is about 12% patches appear too small, the detector's performance may degrade to a point that the cells are miss-classified or missed entirely.…”

“…Accordingly, the precise positions of the cell nuclei are less important because the cells tend to form clumps in the Pap smear slides. Tan et al [44] used also Faster RCNN architecture for cervical cancer cell detection in ThinPrep cytologic test (TCT) images scanned with a seamless slider at up to 400× and achieved AUC of 0.67. More recently, Ma et al [42] proposed an improved Faster RCNN-FPN architecture for cervical cancer detection in cropped patches out of positive Pap smear images.…”

Detection of Cervical Cancer Cells in Whole Slide Images Using Deformable and Global Context Aware Faster RCNN-FPN

“…The object detection methods have also begun to find applications for cervical cancer detection in cervical cytology images. At present, there are few network architectures designed specifically for the specific task of cervical cancer detection [ 41 , 42 , 43 , 44 , 45 , 46 ]. Xu et al [ 45 ] used the generic Faster RCNN for the detection of abnormal cells in cervical smear images scanned at 20× and showed that detection of various abnormal cells was feasible.…”

“…Accordingly, the precise positions of the cell nuclei are less important because the cells tend to form clumps in the Pap smear slides. Tan et al [ 44 ] also used Faster RCNN architecture for cervical cancer cell detection in ThinPrep cytologic test (TCT) images scanned with a seamless slider at up to 400× and achieved AUC of 0.67. More recently, Ma et al [ 42 ] proposed an improved Faster RCNN-FPN architecture for cervical cancer detection in cropped patches out of positive Pap smear images.…”