BRAIN MRI IMAGES ADAPTIVE DENOISING BY APPLYING NEURAL NETWORKS AND RANGE-BASED PARTITIONING

Abstract

Introduction. The denoising autoencoder model for enhancing the quality of T1-weighted brain magnetic resonance imaging (MRI) images are considered.

Methods. This model utilized deep learning methods and incorporated additional components to improve performance and maintain structural integrity. The key elements included convolutional layers for feature extraction, batch normalization for stability, dropout for regularization, and a unique gradient constraint layer to preserve gradient smoothness.

Results.  A specialized activation function classified pixel values based on predefined ranges, ensuring the model's adaptation to varying intensity levels in MRI data. The architecture consisted of an encoder-decoder structure with multiple steps of downsampling and concatenation, facilitating the reconstruction of high-quality images. The combination of these elements resulted in a robust model capable of reducing noise levels while preserving important anatomical details.

Conclusions. The implementation of this autoencoder model enhanced the clarity and diagnostic value of MRI scans, providing radiologists with improved tools for accurate assessment. The study was conducted using a comprehensive dataset of T1-weighted brain MRI images.

References

 Aja-Fernández, S., & Vegas Sánchez-Ferrero, G. (2016). Statistical Analysis of Noise in MRI. Springer. https://doi.org/10.1007/978-3-319-39934-8

Cleary, J.O.S.H., & Guimarães, A.R. (2014). Magnetic Resonance Imaging. In L.M. McManus & R.N. Mitchell (Eds.), Pathobiology of Human Disease (pp. 3987–4004). Academic Press. https://doi.org/10.1016/B978-0-12-386456-7.07609-7

Ding, K., Ma, K., Wang, S., & Simoncelli, E.P. (2022). Image Quality Assessment: Unifying Structure and Texture Similarity. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(5), 2567–2581. https://doi.org/10.1109/TPAMI.2020.3045810

Ghalambaz, M., Sheremet, M.A., Khan, M.A., Raizah, Z., & Shafi, J. (2024). Physics-informed neural networks (P INNs): Application categories, trends and impact. International Journal of Numerical Methods for Heat & Fluid Flow, 34(8), 3131–3165. https://doi.org/10.1108/HFF-09-2023-0568

Mason, A., Rioux, J., Clarke, S.E., Costa, A., Schmidt, M., Keough, V., Huynh, T., & Beyea, S. (2020). Comparison of Objective Image Quality Metrics to Expert Radiologists' Scoring of Diagnostic Quality of MR Images. IEEE Transactions on Medical Imaging, 39(4), 1064–1072. https://doi.org/10.1109/TMI.2019.2930338

Mudeng, V., Kim, M., & Choe, S. (2022). Prospects of Structural Similarity Index for Medical Image Analysis. Applied Sciences, 12(8), Article 8. https://doi.org/10.3390/app12083754

Sliusarenko, D., Netreba, A., & Radchenko, S. (2023). MRI Denoising Neural Network Architecture Convolution. 2023 IEEE 12th International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS) (1, pp. 968–971). IEEE Explore. https://doi.org/10.1109/IDAACS58523.2023.10348629

Thakur, R.S., Chatterjee, S., Yadav, R.N., & Gupta, L. (2023). Chapter 5–Medical image denoising using convolutional neural networks. In S.S. Rajput, N.U. Khan, A.K. Singh, & K.V. Arya (Eds.), Digital Image Enhancement and Reconstruction (pp. 115–138). Academic Press. https://doi.org/10.1016/B978-0-32-398370-9.00012-3