A whole-slide image grading benchmark and tissue classification for cervical cancer precursor lesions with...

Area of Science:

• Pathology
• Digital Pathology
• Oncology

Background:

• Cervical cancer is preventable through screening for precancerous lesions caused by human papillomavirus (HPV).
• Cervical intraepithelial neoplasia (CIN) and squamous intraepithelial lesion (SIL) are standard grading systems.
• Inter-observer variability among pathologists poses a challenge for accurate diagnosis.

Purpose of the Study:

• To create a whole-slide image grading benchmark for cervical cancer precursor lesions.
• To introduce the first publicly available "Uterine Cervical Cancer Database" for research.
• To propose and evaluate a novel morphological feature for improved diagnostic accuracy.

Main Methods:

• Development of a whole-slide image grading benchmark.
• Creation and release of the "Uterine Cervical Cancer Database".
• Proposal of a morphological feature: angle between basal membrane and nuclear major axis.
• Analysis of papillae presence and overlapping cell issues.
• Evaluation of inter-observer variability through pathologist comparisons.

Main Results:

• Establishment of a benchmark dataset for cervical precancer grading.
• Introduction of a novel morphological feature potentially improving diagnostic consistency.
• Discussion of challenges like papillae and overlapping cells in image analysis.
• Quantification of inter-observer variability in diagnosing CIN/SIL.

Conclusions:

• The created benchmark and database facilitate research in digital pathology for cervical cancer.
• The proposed morphological feature shows potential for reducing diagnostic subjectivity.
• Addressing image analysis challenges is crucial for accurate automated grading of cervical lesions.