The 3D medical image reconstruction (3D MIR) is important in the diagnosis and analysis of coronary artery disease (CAD), as it presents a detailed picture of the vascular structures. Nevertheless, it is difficult to reconstruct coronary arteries from 2D angiographic images in the presence of noise, low contrast, complex topology of the vessel and loss of spatial information. The current methods are inefficient in preserving the fine vessel structures and continuity, which results in less accuracy in the reconstruction process. In this paper, an advanced DL-based CARDIO-3DNet for 3D coronary artery reconstruction technique has been proposed by incorporating the advanced feature learning and image processing techniques. Preliminary filtering is done with the Optimally Oriented Flux (OOF) filter to obtain better vessel structures, and the coronary arteries are segmented with U-Net to obtain accurate extraction results. Then a 3D Convolutional Neural Network (3D CNN) is used to implement efficient 3D feature extraction and reconstruction. The proposed method is intended to enhance the continuity of the vessels, maintain the spatial distribution, and produce accurate 3D models of coronary arteries. The significance of this work lies in integrating vessel enhancement deep segmentation and hybrid 3D feature learning within a single framework. The proposed method provides improved reconstruction accuracy and can support early diagnosis and clinical decision-making in CAD analysis. The performance of the proposed CARDIO-3DNet is assessed using measures such as accuracy, precision, recall, specificity, and F1-score. Experimental results show that the model has an accuracy of 99.02% and an F1-score of 97.23%, showing that it is effective in 3D medical picture reconstruction of coronary arteries. Moreover, the proposed CARDIO-3DNet increases the AY by 0.79%, 0.52%, 0.39%, and 0.13% over AlexNet, ShuffleNet, SqueezeNet and DenseNet respectively.