Automated Cervical Digitized Histology Whole-Slide Image Analysis Toolbox

Abstract

Background: Cervical intraepithelial neoplasia (CIN) is regarded as a potential precancerous state of the uterine cervix. Timely and appropriate early treatment of CIN can help reduce cervical cancer mortality. Accurate estimation of CIN grade correlated with human papillomavirus type, which is the primary cause of the disease, helps determine the patient's risk for developing the disease. Colposcopy is used to select women for biopsy. Expert pathologists examine the biopsied cervical epithelial tissue under a microscope. The examination can take a long time and is prone to error and often results in high inter-and intra-observer variability in outcomes.

Methodology: We propose a novel image analysis toolbox that can automate CIN diagnosis using whole slide image (digitized biopsies) of cervical tissue samples. The toolbox is built as a four-step deep learning model that detects the epithelium regions, segments the detected epithelial portions, analyzes local vertical segment regions, and finally classifies each epithelium block with localized attention. We propose an epithelium detection network in this study and make use of our earlier research on epithelium segmentation and CIN classification to complete the design of the end-to-end CIN diagnosis toolbox.

Results: The results show that automated epithelium detection and segmentation for CIN classification yields comparable results to manually segmented epithelium CIN classification.

Conclusion: This highlights the potential as a tool for automated digitized histology slide image analysis to assist expert pathologists.

Keywords: Cervical cancer; cervical intraepithelial neoplasia; classification; convolutional neural networks; detection; digital pathology; histology; segmentation; whole slide image.

Figure 1

Graphical overview of the proposed toolbox

Figure 2

Overview of the proposed toolbox

Figure 3

Steps for region of interest extraction. (a) Finding the contour on the edge of the tissue sample, (b) piece-wise curve for drawing tangents, (c) rectangular boxes drawn with reference to tangents, and (d) region of interest boxes on the original masked image

Figure 4

Mapping of high-resolution region of interest (right) to its low-resolution image (left)

Figure 5

Filtering of epithelium region of interests with the results from the epithelium detection network

Figure 6

(a) Epithelium segmentation mask overlaid as a contour on the epithelium region of interest. (b) Vertical segments generation through the localization process

Figure 7

A cervical intraepithelial neoplasia 3 grade epithelial image with (a) localized vertical segments, and (b) their contribution towards image-level cervical intraepithelial neoplasia classification represented as probability distribution over the segments (attentional weights)

Figure 8

Examples of epithelium detection results. Correctly classified (top row) and misclassified (bottom row) epithelium region of interests