Comparing Manual and Automated Landmarking Accuracy in Geometric Morphometrics: A Study on Cattle Skulls and Distal Phalanges

Abstract

Geometric morphometric studies often rely on accurate landmark placement to analyse shape and size variations. Manual and automated landmarking methods are widely used, but their performance can vary depending on the anatomical structure and complexity. This study compares manual and automated landmarking accuracy in two anatomical samples: the skull and the distal phalanx. Fifteen Holstein cattle skulls and 15 distal phalanges were analysed. Two landmark configurations were used: 10 and 20 landmarks for the skull and 5 and 10 landmarks for the distal phalanx. Both manual and automated landmarking were performed using Slicer software. Procrustes distance and centroid size were calculated to assess shape and size differences. ANOVA was applied to evaluate statistical differences, and PCA was conducted to visualise shape variations. Procrustes distance revealed significant differences between manual and automated landmarking for most configurations, particularly in the skull, highlighting the variability introduced by automated methods. No significant differences were observed for centroid size, indicating consistency in size measurements. Automated landmarking showed increased variability in capturing shape, especially in complex structures and higher landmark densities. Automated landmarking provides efficiency but introduces significant shape variability, particularly in complex anatomical structures. Our findings highlight the superior accuracy of manual landmarking, particularly for capturing subtle anatomical features and complex structures where automated methods face challenges. Although more time-consuming, manual landmarking minimises variability and preserves crucial morphological details, making it essential for precise analysis, especially in the presence of thresholding artefacts in 3D models. Future research should explore multiple software platforms and refine automated algorithms to improve performance in morphometric analyses.