Image Registration Guided Brain Tumour Segmentation and Advancements

In the field of medical imaging, detecting brain tumors through image registration poses a formidable challenge. Automated tools play a vital role in non- invasively identifying and documenting tumors. Significant advancements in medical image processing, facilitated by advanced image registration techniques, have greatly contributed to improved brain tumor detection. Based on such techniques this study discusses image registration-based brain tumor segmentation and their advancements. This experimental validation is performed with MRI images of 100 patients from BraTS2020 (Brain Tumor Segmentation) dataset, it involves training models without registered images (35 epochs) and with registered images (20,35 and 40 epochs). This research presents a unique image registration approach that sets it apart from existing methods by employing the Gradient Descent Linear Search algorithm for optimization and using Mutual Information as the Similarity Metric, our method demonstrates superior performance compared to conventional techniques. The integration of Linear Interpolation, 3D Euler Transformation, and Advanced Resampling distinguishes our work, further providing a more comprehensive and accurate solution for multi- dimensional image alignment.