The combination of transfer learning and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) offers a new method to improve 3D brain reconstruction from MRI data in the fast-changing field of medical imaging. This study investigates the effectiveness of different pre- trained Convolutional Neural Network (CNN) models, including ResNet, InceptionV3, DenseNet, and the Vision Transformer (ViT), when used together with ANFIS. The project intends to enhance the input feature set for ANFIS by utilising the sophisticated feature extraction capabilities of CNNs, which have been trained on large picture datasets. This will result in improved analytical accuracy and efficiency. The approach entails conducting a comparative investigation of each CNN model's capacity to preprocess MRI data, with a specific emphasis on their flexibility and performance in extracting pertinent characteristics for distinguishing brain tissue. Afterwards, the modified characteristics are inputted into ANFIS for additional analysis, utilising its fuzzy logic to manage the complexities of medical imaging data. The main objective is to determine the transfer learning model that most effectively supports ANFIS for this job, with the aim of achieving a more accurate and dependable reconstruction of brain structures.