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

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.

The artificial intelligence-assisted cytology diagnostic system in large-scale cervical cancer screening: A population-based cohort study of 0.7 million women

Background

Adequate cytology is limited by insufficient cytologists in a large‐scale cervical cancer screening. We aimed to develop an artificial intelligence (AI)‐assisted cytology system in cervical cancer screening program.

Methods

We conducted a perspective cohort study within a population‐based cervical cancer screening program for 0.7 million women, using a validated AI‐assisted cytology system. For comparison, cytologists examined all slides classified by AI as abnormal and a randomly selected 10% of normal slides. Each woman with slides classified as abnormal by either AI‐assisted or manual reading was diagnosed by colposcopy and biopsy. The outcomes were histologically confirmed cervical intraepithelial neoplasia grade 2 or worse (CIN2+).

Results

Finally, we recruited 703 103 women, of whom 98 549 were independently screened by AI and manual reading. The overall agreement rate between AI and manual reading was 94.7% (95% confidential interval [CI], 94.5%‐94.8%), and kappa was 0.92 (0.91‐0.92). The detection rates of CIN2+ increased with the severity of cytology abnormality performed by both AI and manual reading (P_trend < 0.001). General estimated equations showed that detection of CIN2+ among women with ASC‐H or HSIL by AI were significantly higher than corresponding groups classified by cytologists (for ASC‐H: odds ratio [OR] = 1.22, 95%CI 1.11‐1.34, P < .001; for HSIL: OR = 1.41, 1.28‐1.55, P < .001). AI‐assisted cytology was 5.8% (3.0%‐8.6%) more sensitive for detection of CIN2+ than manual reading with a slight reduction in specificity.

Conclusions

AI‐assisted cytology system could exclude most of normal cytology, and improve sensitivity with clinically equivalent specificity for detection of CIN2+ compared with manual cytology reading. Overall, the results support AI‐based cytology system for the primary cervical cancer screening in large‐scale population.

Trends in Cervical Cytology Screening and Reporting Practices: Results From the College of American Pathologists 2011 PAP Education Supplemental Questionnaire

Context

The College of American Pathologists periodically surveys laboratories to determine changes in cytopathology practices. We report the results of a 2011 gynecologic cytology survey.

Objective

To provide a cross-sectional survey of gynecologic cytology practices in 2010.

Design

In 2011, a survey was sent to 1604 laboratories participating in the College of American Pathologists gynecologic cytology interlaboratory comparison education program and proficiency testing programs requesting data from 2010 on the following topics: terminology/reporting, cytotechnologist workload, quality assurance, reagents, and ancillary testing.

Results

Six hundred and twenty-five laboratories (39%) replied to the survey. The nonstandard use of “low-grade squamous intraepithelial lesion cannot exclude high-grade squamous intraepithelial lesion” is used by most laboratories to report the presence of low-grade squamous intraepithelial lesion with possibility of high-grade squamous intraepithelial lesion. Most laboratories also report the presence or absence of cells from the transformation zone. Most respondents do not limit cytotechnologist screening workload during the work shift. Only about one-third of laboratories (188 of 582; 32%) use image-assisted screening devices. Rapid prescreening as a quality assurance measure is used by only 3.5% (21 of 594) of the laboratories. When used for screening, most laboratories use the imager for retrospective review of slides to detect human locator and interpretive errors. Most laboratories receive both liquid-based cytology samples (mainly ThinPrep, Hologic, Marlborough, Massachusetts) and conventional Papanicolaou tests. Expiration dates of liquid-based cytology test vials are not usually recorded.

Conclusions

The field of gynecologic cytology is evolving rapidly. These survey results offer a snapshot of national gynecologic cytology practices in 2010.

Assessment of manual workload limits in gynecologic cytology: reconciling data from 3 major prospective trials of automated screening devices

Previous prospective studies have shown different results when comparing automated and manual screening of gynecologic cytology. The results of 3 large prospective studies were reviewed and relative sensitivity used as a gold standard. No significant differences could be shown in relative sensitivity between the ThinPrep Imaging System and the FocalPoint GS Imaging System (P > .05). When manual screening was restricted to less than 6 hours per day, 50 or fewer slides per day, and at least 6 minutes per slide (<10 slides/h), the relative sensitivity for automation was significantly lower for atypical squamous cells of undetermined significance and above (ASC+) (0.81; 95% confidence interval [CI], 0.79-0.83) than when manual screening was not restricted (1.07; 95% CI, 1.03-1.10). All 3 sites that screened 10 or more slides per hour manually had a relative sensitivity for automation that was significantly higher for high-grade squamous intraepithelial lesions and above (HSIL+) than for the remaining groups who screened less than 10 slides per hour (1.40 [95% CI, 1.22-1.60] vs 0.97 [95% CI, 0.95-1.00]). These results suggest that location finding of abnormalities (ASC+) may be more strongly associated with time spent screening per day, whereas classification/interpretation skills (HSIL+) may depend on time spent on an individual case. There is no evidence that automated screening devices are more sensitive than manual screening performed at lower well-defined workloads. More restricted workloads (≤41 slides/d, ≤4.5 h/d) for manual screening may perform significantly better than automated screening devices as measured by histologic cervical intraepithelial neoplasia 2 and above.

Increasing cytotechnologist workload above 100 slides per day using the BD FocalPoint GS imaging system negatively affects screening performance

Studies examining the effects of increased workload on the performance of individual cytotechnologists are limited. Using FocalPoint GS, the performance of 3 cytotechnologists was evaluated. The study consisted of 3 phases. In phase I, cytotechnologists were asked to screen at their usual pace. In phase II, cytotechnologists were asked to screen as fast as possible without feeling that the quality of their work was diminished. In phase III, cytotechnologists were asked to screen at least 15% more than their daily workload from phase II. Productivity was increased by decreasing the percentage of cases that underwent full manual review (from 38% to 19%) and by decreasing the time spent on each slide (from 5.5 min to 3.7 min). Overall, the total abnormal rate decreased by 31.9% from phase I to phase III of the study. In addition, the false-negative fraction increased significantly, from 1% to 6.9%. Our results indicated a negative association between increased cytotechnologist daily workload with FocalPoint GS and CT screening performance. Workloads were increased by decreasing the time spent reviewing 10 fields of view and the percentage of cases that underwent full manual review.