Towards Accurate Breslow Measurements: Mitigating Issues in Histopathological Imaging

Area of Science:

• Dermatology
• Medical Imaging
• Computer Science

Background:

• Breslow thickness is crucial for cutaneous melanoma staging.
• Manual measurement faces challenges due to operator dependency and epidermal morphology.
• Accurate measurement is vital for prognosis and treatment decisions.

Purpose of the Study:

• To develop and evaluate a semi-automated Computer Vision (CV) software for robust Breslow thickness estimation.
• To quantify inter-operator variability in manual Breslow thickness measurements.
• To compare AI-driven measurements with expert histopathologist assessments.

Main Methods:

• Development of a semi-automated CV software for Breslow thickness measurement.
• Seven histopathologists measured Breslow thickness on 40 Whole Slide Images (WSIs) of pT1a melanomas.
• Comparison of manual measurements against AI results, assessing inter-operator agreement and angular variance.

Main Results:

• Significant variability in manual measurement orientation was observed among operators.
• Epidermal irregularity correlated linearly with increased measurement uncertainty (angular variance).
• The AI system showed statistically significant differences with five of seven operators, indicating a tendency for experts to overestimate Breslow thickness.

Conclusions:

• Semi-automated CV software provides a more robust Breslow thickness estimation compared to manual methods.
• AI-based measurement can mitigate operator dependency and potential overestimation in melanoma staging.
• This technology has the potential to improve the accuracy and consistency of melanoma prognostic parameter assessment.