A Deep Learning Framework for Denoising MRI Images using Autoencoders

Magnetic Resonance Imaging (MRI) has become an indispensable tool in the medical field for diagnosing and monitoring various diseases and conditions. However, the quality of MRI images can be degraded by noise, which can lead to inaccurate interpretations and diagnoses. In recent years, machine lear...

Full description

Saved in:
Bibliographic Details
Published in:International Conference on Bio-engineering for Smart Technologies (Online) pp. 1 - 4
Main Authors: Zein, Mohammad El, Laz, Wissam El, Laza, Malak, Wazzan, Taha, Kaakour, Ibrahim, Adla, Yasmine Abu, Baalbaki, Jad, Diab, Mohammad O., Sabbah, Maher, Zantout, Rached
Format: Conference Proceeding
Language:English
Published: IEEE 07.06.2023
Subjects:
autoencoder
Convolutional neural networks
Data augmentation
Deep learning
denoising
Magnetic resonance imaging
Medical diagnostic imaging
Monitoring
MRI
neural networks
noise
Noise reduction
ISSN:2831-4352
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Magnetic Resonance Imaging (MRI) has become an indispensable tool in the medical field for diagnosing and monitoring various diseases and conditions. However, the quality of MRI images can be degraded by noise, which can lead to inaccurate interpretations and diagnoses. In recent years, machine learning techniques have shown great potential in enhancing the accuracy of image denoising, especially in the medical domain. In this study, we propose a novel deep learning model based on autoencoders for denoising MRI images. We employed various preprocessing techniques such as data augmentation, resizing, normalization, and conventional denoising methods on our MRI images. Our model comprises a convolutional neural network autoencoder (CCNAE), which we fine-tuned by testing different parameters and layers to achieve optimal performance. Our results demonstrate a validation loss of approximately 0.0001, indicating a substantial improvement in denoising performance. Our work represents an important step towards developing an efficient and effective method for denoising MRI images without compromising critical data or time.
ISSN:2831-4352
DOI:10.1109/BioSMART58455.2023.10162068