Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Apr;12(7):8690-8699.
doi: 10.1002/cam4.5581. Epub 2023 Jan 11.

Application of EfficientNet-B0 and GRU-based deep learning on classifying the colposcopy diagnosis of precancerous cervical lesions

Xiaoyue Chen  1 , Xiaowen Pu  1 , Zhirou Chen  1 , Lanzhen Li  2   3 , Kong-Nan Zhao  4   5 , Haichun Liu  2   3 , Haiyan Zhu  1
Affiliations

Application of EfficientNet-B0 and GRU-based deep learning on classifying the colposcopy diagnosis of precancerous cervical lesions

Xiaoyue Chen et al. Cancer Med. 2023 Apr.

Abstract

Background: Colposcopy is indispensable for the diagnosis of cervical lesions. However, its diagnosis accuracy for high-grade squamous intraepithelial lesion (HSIL) is at about 50%, and the accuracy is largely dependent on the skill and experience of colposcopists. The advancement in computational power made it possible for the application of artificial intelligence (AI) to clinical problems. Here, we explored the feasibility and accuracy of the application of AI on precancerous and cancerous cervical colposcopic image recognition and classification.

Methods: The images were collected from 6002 colposcopy examinations of normal control, low-grade squamous intraepithelial lesion (LSIL), and HSIL. For each patient, the original, Schiller test, and acetic-acid images were all collected. We built a new neural network classification model based on the hybrid algorithm. EfficientNet-b0 was used as the backbone network for the image feature extraction, and GRU(Gate Recurrent Unit)was applied for feature fusion of the three modes examinations (original, acetic acid, and Schiller test).

Results: The connected network classifier achieved an accuracy of 90.61% in distinguishing HSIL from normal and LSIL. Furthermore, the model was applied to "Trichotomy", which reached an accuracy of 91.18% in distinguishing the HSIL, LSIL and normal control at the same time.

Conclusion: Our results revealed that as shown by the high accuracy of AI in the classification of colposcopic images, AI exhibited great potential to be an effective tool for the accurate diagnosis of cervical disease and for early therapeutic intervention in cervical precancer.

Keywords: artificial intelligence (AI); cervical cancer; colposcopy; precancerous cervical lesions.

PubMed Disclaimer

Figures

FIGURE 1
FIGURE 1
The flow chart of this study. The spatial features of the original images were extracted and resized in the model. Images were imputed, and with the data analysis, the final output is the classification results.
FIGURE 2
FIGURE 2
The architecture of boosted‐EfficientNet‐B0. EfficientNet first extracts the image features through its convolutional layers. The attention mechanism is then utilized to reweight the features via increasing the activation of the significant parts. Next, we performed FF on the outputs of several convolutional layers. Subsequently, the images are classified based on those fused features. Details of these methods are described in the Materials and Methods sections.
FIGURE 3
FIGURE 3
Representative examples of the original images. Every patient had three pictures: normal saline, acetic acid, and Lugol's iodine solution.
FIGURE 4
FIGURE 4
ROC curve of the dichotomous classification model (LSIL+ NC vs. HSIL)by Resnet50, AlexNet, and E‐B0 with GRU.
FIGURE 5
FIGURE 5
ROC curve of the three‐classification model (NC Vs. LSIL Vs. HSIL by E‐B0 with GRU.
See this image and copyright information in PMC

References

    1. Brisson M, Drolet M. Global elimination of cervical cancer as a public health problem. Lancet Oncol. 2019;20(3):319‐321. - PubMed
    1. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209‐249. - PubMed
    1. Waxman AG, Chelmow D, Darragh TM, Lawson H, Moscicki AB. Revised terminology for cervical histopathology and its implications for management of high‐grade squamous intraepithelial lesions of the cervix. Obstet Gynecol. 2012;120(6):1465‐1471. - PMC - PubMed
    1. Bokil M, Lim B. Colposcopy: a closer look into its past, present and future. BJOG. 2019;126(4):543. - PubMed
    1. Mayeaux EJ Jr, Novetsky AP, Chelmow D, et al. ASCCP colposcopy standards: colposcopy quality improvement recommendations for the United States. J Low Genit Tract Dis. 2017;21(4):242‐248. - PMC - PubMed