Assessing the Accuracy of NRR with and without Ground Truth

Generalised Overlap Measures

Overview

Non-rigid Registration (NRR)

Motivation for Assessment

Two New Approaches

Generalised Overlap

Overlap Measures

Binary Overlap Measures

Alternate Form

Interpolated Label Images

Generalised Overlap

Label Weighting

Model-Based Assessment

Model-based Framework

Building an Appearance Model

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Training and Synthetic Images

Model Quality

Measuring Inter-Image Distance

Shuffle Distance

Varying Shuffle Radius

Experimental Evaluation

Experimental Design

Brain Data

Perturbation Framework

Examples of Perturbed Images

Results – Overlap

Results – Model-Based

Results – Comparison

Results – Sensitivities

Conclusions

Acknowledgements