Exploring Deep Relationship Learning for Regression: A Case Study on Brain Age Estimation

Abstract

To understand how the brain develops and matures, an accurate age estimate utilizing MRI pictures is essential. There has been encouraging progress in this area using deep learning-based methods, especially CNNs, which can extract important features from MRI scans. By combining the VGG16 and EDCNN networks, this study presents a new method for determining brain age. For feature extraction, we use VGG16, and for brain age prediction, we use EDCNN. The proposed method was evaluated using the Kaggle repository, a freely available benchmark dataset that includes magnetic resonance imaging (MRI) images of 562 individuals ranging in age from 20 to 86. Following the preprocessing of the dataset, the images were adjusted to 224 × 224 x 3 dimensions and normalized. Using a broad range of evaluation criteria, the proposed methodology's performance was contrasted with that of several state-of-the-art techniques. Additionally, ablation experiments were conducted as part of this study to determine the significance of various components of the suggested technique. The results showed that the method's performance was significantly improved when VGG16 was used as a feature extractor. In addition, this study found that the proposed method's performance was significantly improved after VGG16 network tuning. The research demonstrates that technologies based on deep learning might help us comprehend how the brain develops and ages. The approach uses EDCNN to forecast brain ageing and VGG16 to extract features. In order to decipher the structural subtleties seen in brain MRI images, the acclaimed feature extraction tool VGG16 is used as a base.