Classification of Surgeons' Gender from Videos of the Surgical Field

We present a novel approach to classify the gender of the operating surgeon in videos from the microscope (that show only the surgical field) using the SurgGender model and a step-based analysis technique. This method addresses both technical and ethical challenges of implicit gender biases, which impact training and career progression. By focusing on gender classification, our work aims to support equitable, context-aware surgical systems. The SurgGender leverages 3D shifted window-based self-attention to capture key spatialtemporal patterns. Our step-based approach enhances classification by isolating relevant video segments and reducing non-informative noise. Rigorous testing demonstrates that this method improves accuracy and processing efficiency, establishing a foundation for bias-aware surgical video analysis.