Towards Automatic Identification of Missing Tissues using a Geometric-Learning Correspondence Model
Identification
DOI:
10.48550/arxiv.2502.11265
Publication Date:
2025-02-16
AUTHORS (2)
ABSTRACT
Missing tissue presents a big challenge for dose mapping, e.g., in the reirradiation setting. We propose pipeline to identify missing on intra-patient structure meshes using previously trained geometric-learning correspondence model. For our application, we relied prediction discrepancies between forward and backward correspondences of input meshes, quantified correspondence-based Inverse Consistency Error (cICE). optimised threshold applied cICE points dataset 35 simulated mandible resections. Our identified threshold, 5.5 mm, produced balanced accuracy score 0.883 training data, an ensemble approach. This plausible results real case where ~25% was removed after surgical intervention. The pipeline, however, failed more extreme ~50% removed. is first time modelling proposed corresponding anatomy.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES ()
CITATIONS ()
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....