An efficient 3D transformer network-based adaptive level set algorithm for joint registration and segmentation over multi-modal medical images

A joint adaptive method is proposed for the registration and segmentation of Computed Tomography (CT) and Magnetic Resonance Imaging (MRI). The main contribution of this work is the introduction of a model named 3D Trans-Registration with Segmentation Network (3D-TRSegNet) for registration and segmentation. Here, the segmentation is performed using the Adaptive Level Set Algorithm (ALSA) and the proposed model jointly performs registration and segmentation. Here, the model hyperparameters are optimized using the Integrated Pelican and Chimp Optimization Strategy (IPCOS) algorithm. Finally, the experimental results are evaluated to assess the effectiveness of the proposed method. Compared with the baseline approaches, the proposed method achieves improved performance in both registration and segmentation tasks. Therefore, the proposed method achieves a Structural Similarity Index Measure (SSIM) of 0.622, a Root Mean Square Error (RMSE) of 23.86, and an accuracy of 93.73%, thus demonstrating its effectiveness for medical image registration and segmentation.