We utilized two complementary datasets: a clinical CT dataset used to train the landmark detection model, and an RGB-D dataset of cadaveric surgeries used exclusively for validation. Throughout the ...