Deep TPS-PSO: Hybrid Deep Feature Extraction and Global Optimization for Precise 3D MRI Registration

• Published in: IEEE open journal of the Computer Society Vol. 6; pp. 1090 - 1099
• Main Authors: Ramasamy, Gayathri, Singh, Tripty, Yuan, Xiaohui, Naik, Ganesh R
• Format: Journal Article
• Language: English
• Published: IEEE 2025
• Subjects: 3D MRI registration; ablation study; Accuracy; brats 2022; deep learning; Deformable models; Deformation; dice similarity coefficient; Feature extraction; Image registration; Magnetic resonance imaging; mutual information; non-linear deformation; Optimization; Particle swarm optimization; ROC analysis; Splines (mathematics); thin plate spline; Three-dimensional displays; voxel-based registration
• ISSN: 2644-1268; 2644-1268
• DOI: 10.1109/OJCS.2025.3586956

This article presents TPS-PSO, a hybrid deformable image registration framework integrating deep learning, non-linear transformation modeling, and global optimization for accurate inter-subject, intra-modality 3D brain MRI alignment. The method combines a 3D ResNet encoder to extract volumetric features, a Thin Plate Spline (TPS) model to capture smooth anatomical deformations, and Particle Swarm Optimization (PSO) to estimate transformation parameters efficiently without relying on gradients. Evaluated on the BraTS 2022 dataset, TPS-PSO achieved state-of-the-art performance with a Dice Similarity Coefficient (DSC) of 85.7%, Mutual Information (MI) of 1.23, Target Registration Error (TRE) of 3.8 mm, HD95 of 6.7 mm, and SSIM of 0.92. Comparative experiments against five recent baselines confirmed consistent improvements. Ablation studies and convergence analysis further validated the contribution of each module and the optimization strategy. The proposed framework generates topologically plausible deformation fields and shows strong potential for clinical and research applications in neuroimaging.