Integrated Neural-Hybrid System for Efficient Tumor Detection and Object Reconstruction

In computer vision and robotics, reconstructing multi-view 3D images is essential for accurate object representation from 2D data. In the first study, optimised weights through Adaptive School of Fish Optimisation are combined with 2D and 3D networks to introduce a Residual Network-50 model for deep learning-based 3D image reconstruction. On the ShapeNet dataset, this method demonstrates superior accuracy (0.993), F-score (0.734), and IoU (99.3%). Using Concurrent Excited DenseNet (CED)for feature extraction and Attention-Dense GRUs for prediction, the second study introduces the Concurrent Attentional Reconstruction Network(CARN) for reconstructing point clouds from single 2D images, achieving over 99% accuracy with low EMD and CD values. By combining convolutional layers, inception modules, and attention mechanisms with preprocessing steps like Ex_NLMF for noise reduction and Up_FKMA for accurate disease area identification, the third method,Twin Attention-aided Convolutional Inception Capsule Network (TA_CICNet), performs exceptionally well in medical image reconstruction and classification when it comes to diagnosing brain tumours.