The effect of automated landmark identification on morphometric analyses

Area of Science:

• Morphological analysis
• Bioinformatics
• Comparative anatomy

Background:

• Manual landmark identification in morphometrics is time-consuming.
• Automated methods offer potential efficiency gains for large datasets.
• Quantifying intra- and inter-strain variation requires robust landmarking.

Purpose of the Study:

• Compare manual vs. automated landmarking for mouse skulls.
• Assess impact on morphometric analysis of large samples (n=1205, 62 genotypes).
• Evaluate a non-linear image registration-based automated method.

Main Methods:

• Compared manual and automated landmark placement on 1205 adult mouse skulls.
• Utilized non-linear image registration for automated landmarking.
• Tested genotype-specific registration pre-processing for accuracy improvement.

Main Results:

• Automated landmark placement differed significantly from manual but correlated in shape covariation.
• Genotype-specific registration did not substantially improve accuracy.
• Lower accuracy at landmarks with poor image registration alignment.
• Automated methods reduced estimated skull shape variance, partly due to lack of observer error.

Conclusions:

• Automated landmarking significantly reduces time compared to manual methods.
• The method retains power to identify shape differences between mouse genotypes.
• Image registration errors can be mitigated by specimen preparation and acquisition improvements.
• Automated landmarking is a viable alternative for large-scale morphometric studies.