Deep Convolutional Neural Network-Based MRI Image Compression: A Comparative Study of VGG19 and EfficientNetB7 Architectures

View at publisher Summarise with VIRAT
Authors: Rajalingam Arumuganainar, R. Anusuya, J. R. Arunkumar, V Vijaya Baskar
Cited by: 0
References: 14

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper examines the application of deep convolutional neural networks (CNNs) for MRI image compression and reconstruction, aiming to achieve high efficiency while preserving diagnostic quality. The graphic shows the baseline MRI scan next to the decoded results from two designs, VGG19 and DenseNet. This shows how well the networks can keep intricate structural details and tissue boundaries during the decompression process. We use the right performance metrics to measure the models' performance, such as processing time, peak signal-to-noise ratio (PSNR), compression ratio (CR), and root mean square error. The high compression ratio of$\mathbf{1 5 2. 4 2} \boldsymbol{(} \mathbf{R M S E} \boldsymbol{=} \mathbf{4. 8 4} \boldsymbol{)}$and PSNR of 34.43 dB showed that VGG19's reconstruction was of very high quality and had no distortion. EfficientNetB7 had a compression ratio of 152.12 and a slightly higher PSNR of 35.42 dB, although it took longer to encode (4.324 s) and decode (2.213 s) than VGG19. However, the image quality was about the same. Both models can compress and reconstruct MRI data, but the results show that VGG19 is preferable for real-time or largescale medical imaging applications because it strikes a good compromise between speed and picture quality. For highprecision diagnostic applications, EfficientNetB7's better noise suppression and better texture reconstruction more than make up for how hard it is to compute.

Fields of Study & Subjects

Advanced Data Compression Techniques Advanced Neural Network Applications Video Coding and Compression Technologies

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
882–887
Publication date
21 January 2026
Language
EN