Related Experiment Video
Updated: Jun 12, 2026
07:00
Diffusion Tensor Magnetic Resonance Imaging in Chronic Spinal Cord Compression
Published on: May 7, 2019
9.0K
Nerve Root Compression Analysis to Find Lumbar Spine Stenosis on MRI Using CNN
Turrnum Shahzadi1, Muhammad Usman Ali2, Fiaz Majeed1
1Department of Information Technology, University of Gujrat, Gujrat 50700, Pakistan.
Diagnostics (Basel, Switzerland)
|September 28, 2023
Summary
This study introduces a convolutional neural network (CNN) for automated lumbar spine stenosis (LSS) detection using MRI images. The AI model achieved high accuracy, aiding clinical diagnosis.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- Spinal Diagnostics
Background:
- Lumbar spine stenosis (LSS) is a condition causing nerve compression and low back pain, with current detection methods having limitations in accuracy and versatility.
- Accurate and efficient detection of LSS is crucial for timely diagnosis and patient management.
- Existing segmentation algorithms for LSS lack the desired accuracy and adaptability for widespread clinical use.
Purpose of the Study:
- To develop and evaluate an automated method for categorizing lumbar spine stenosis (LSS) using magnetic resonance imaging (MRI).
- To leverage convolutional neural networks (CNNs) for precise LSS detection and grading from MRI scans.
- To enhance diagnostic accuracy and support clinical decision-making in LSS diagnosis.
Main Methods:
- A convolutional neural network (CNN) model was developed for the automated detection and grading of LSS from axial-view MRI images.
- Radiological grading was performed on a public dataset, defining four regions of interest (ROIs) for normal, mild, moderate, and severe LSS.
- Experiments utilized 1545 MRI images, with datasets split into multi-ROI and single-ROI categories, employing an 80:20 training/testing ratio and fivefold cross-validation.
Main Results:
- The proposed CNN-based method demonstrated high diagnostic accuracy, achieving 97.01% for the multi-ROI dataset and 97.71% for the single-ROI dataset.
- The computer-aided diagnosis approach significantly improved accuracy in simulated clinical workflows.
- The CNN model's efficacy was validated across different datasets, outperforming existing state-of-the-art methods.
Conclusions:
- The developed CNN-based approach for MRI image segmentation effectively detects and grades lumbar spine stenosis (LSS).
- This automated system shows significant potential to enhance diagnostic accuracy and assist medical experts in clinical decision-making for LSS.
- The proposed method offers a superior and robust solution compared to current state-of-the-art techniques for LSS diagnosis.
