• Landmark validation
• 7 patients, 7 - 8 fractions
• 23 landmarks per CBCT, two
human observers
• B-spline deformable
registration for landmark
propagation
• Use of ANOVA method to
correct for observer variation
Analysis of variance