Enhancing mitosis quantification and detection in meningiomas with computational digital pathology

Area of Science:

• Neuropathology
• Digital Pathology
• Computational Pathology

Background:

• Accurate mitosis counting is crucial for meningioma grading but suffers from significant inter-observer variability among pathologists.
• Challenges include identifying mitosis hotspots and detecting mitotic figures consistently.

Purpose of the Study:

• To develop and evaluate a computational strategy leveraging digital pathology to enhance the precision and reliability of mitosis assessment in meningiomas.
• To address inter-observer variation and improve diagnostic accuracy for meningioma grading.

Main Methods:

• A depth-first search algorithm was developed to quantify the maximum mitotic count in 10 consecutive high-power fields.
• A collaborative sphere approach was implemented, grouping pathologists to collectively detect mitoses within high-power fields.
• The strategy was validated on meningioma slides, comparing algorithmic counts and group performance against individual pathologist assessments.

Main Results:

• The depth-first search algorithm identified two borderline meningioma cases that were upgraded after analysis, demonstrating improved quantification.
• Groups of three pathologists achieved higher average precision (0.897) and sensitivity (0.699) in mitosis detection compared to individual pathologists (precision: 0.750, sensitivity: 0.667).

Conclusions:

• The proposed computational strategy, combining algorithmic analysis and collaborative review, significantly enhances mitosis assessment accuracy in meningiomas.
• This approach has the potential to be integrated with artificial intelligence workflows for rapid, robust mitosis evaluation, ultimately benefiting patient management.