Automated reporting of cervical biopsies using artificial intelligence.

Area of Science:

• Digital pathology
• Artificial intelligence in oncology
• Cervical cancer diagnostics

Background:

• Early-stage invasive cervical cancer has a 5-year survival rate of 92%.
• Awareness of cervical cancer signs and symptoms is crucial for early detection and successful treatment.
• Pathological diagnosis of cervical biopsies is critical for accurate cancer reporting.

Purpose of the Study:

• To develop and evaluate an Artificial Intelligence (AI) algorithm for automated reporting of digital cervical biopsies.
• To enhance the efficiency of pathological diagnosis by identifying malignant and high-grade lesions.
• To achieve high sensitivity for detecting malignant cervical cancer cases.

Main Methods:

• An AI algorithm was trained and validated on 1738 cervical whole slide images (WSIs).
• The algorithm was evaluated on an independent test set of 811 WSIs.
• The AI's performance was assessed for classifying slides and identifying areas of interest for pathologists.

Main Results:

• The AI algorithm achieved 93.4% malignant sensitivity in classifying slides on the independent test set.
• The algorithm processes whole slide images (WSIs) in approximately 1.5 minutes using an NVIDIA Tesla V100 GPU.
• The system can process various WSI formats (TIFF, iSyntax, CZI) and is extensible to others.

Conclusions:

• The developed AI algorithm demonstrates high sensitivity for detecting malignant cervical cancer in digital biopsies.
• This AI tool has the potential to significantly improve the efficiency and speed of pathological diagnosis.
• Automated analysis of cervical WSIs by AI can aid pathologists in triaging cases and reducing reporting time.