Personalized Respiratory Motion Modeling Incorporating Longitudinal Data through Two-stage Transfer Learning.

Abstract

Purpose: This study aims to develop an accurate image registration framework for personalized respiratory motion modeling.

Methods: The proposed framework incorporates longitudinal data through a two-stage transfer learning approach. In the first stage, transfer learning is employed on longitudinal data collected from the same device. In the second stage, a personalized model is constructed using the transfer learning approach, reusing the model from the first stage. A novel cross-error function is introduced to guide the customized adaptation stage.

Results: The experiments demonstrate the effectiveness of the proposed framework in respiratory motion modeling. Integrating longitudinal data allows for improved accuracy for personalized respiratory motion modeling.

Conclusion: The study presents a novel approach that incorporates longitudinal data into a two-stage transfer learning process for personalized respiratory motion modeling. The framework demonstrates improved accuracy. The results highlight the potential of leveraging longitudinal data to provide personalized image registration solutions.