Investigation of Image Registration between 3D Multi-Modal MRI and CT Scans in an Orthotopic Breast Cancer Model

Abstract

Our study aims to construct a 3D multi-modal dataset of magnetic resonance imaging (MRI) and computed tomography (CT) from an orthotopic breast cancer model using two cell lines, and evaluate which combination of MR contrast images and CT images is most suitable for MR-CT registration. For the normal model, BALB/C and Nude mice were used, while for the breast cancer model, 40, 24, and 15 subjects were obtained for T2-weighted and CT images from the Normal model and cancer models (4T1 and MDA-MB-231), respectively. T1-weighted and PD-weighted images were obtained from 39, 24, and 14 subjects, respectively. Prior to registration, initial alignment based on bone structure was performed, followed by interpolation to adjust matrix sizes, and masking to remove unnecessary information such as cradles and markers used during experiments. MR images (T1W, T2W, PDW) were set as fixed images, and CT as moving images for intensity-based image registration. Registered CT images were compared with original MR images using evaluation metrics including Normalized Mutual Information (NMI) and Normalized Cross-Correlation (NCC). Statistical analysis using ANOVA was conducted to analyze relationships between various MR and CT image combinations. The results statistically demonstrate that the choice of MRI modality does not significantly impact the registration quality of CT images. This study provides a comprehensive 3D longitudinal MR-CT multi-modal paired dataset from an orthotopic breast cancer model, contributing insights into optimal selection criteria for MR-CT image registration.