International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
Spinal Cord Gray Matter Segmentation Using Deep Dilated Convolution Swetha H U1, Paavana S S2, Rakshitha Y B3, Shwetha B4, Vandana B S5 1, Assistant Professor, Artificial Intelligence and Machine Learning, Bapuji Institute of Engineering and Technology,
Davangere, affiliated to VTU Belagavi, Karnataka, India.
2, Bachelor of Engineering, Artificial Intelligence and Machine Learning, Bapuji Institute of Engineering and
Technology, Davangere, Karnataka, India
3,Bachelor of Engineering, Artificial Intelligence and Machine Learning, Bapuji Institute of Engineering and
Technology, Davangere, Karnataka, India.
4, Bachelor of Engineering, Artificial Intelligence and Machine Learning, Bapuji Institute of Engineering and
Technology, Davangere, Karnataka, India
5, Bachelor of Engineering, Artificial Intelligence and Machine Learning, Bapuji Institute of Engineering and
Technology, Davangere, Karnataka, India ---------------------------------------------------------------------***--------------------------------------------------------------------
Abstract - Changes in gray matter (GM) tissue have been
time-consuming, labor-intensive, and error-prone operation that calls for specialized knowledge.
linked to a variety of neurological conditions and have recently been identified as a biomarker for impairment in amyotrophic lateral sclerosis. For contemporary research on the spinal cord, the capacity to automatically segment the GM is crucial. In this study, we develop a cutting-edge, straightforward, and fully automated human spinal cord gray matter segmentation technique using Deep Learning that can be applied to both in vivo and ex vivo MRI scans. When compared to conventional medical imaging architectures like U-Nets, we report state-of-the-art results in eight out of ten different evaluation metrics and significant network parameter reduction. We test our approach against six independently developed methods on a GM segmentation challenge.
The complex structure of the spinal cord is difficult for conventional techniques like threes holding and edge detection to handle, especially in high-dimensional pictures where noise, low contrast, and individual anatomical variances make it challenging to distinguish the gray matter boundaries. Consequently, there is an increasing need for automated, precise, and effective methods of spinal cord gray matter segmentation. Convolution neural networks (CNNs) are frequently utilized for a variety of medical imaging applications, such as organ, tissue, and lesion segmentation in brain and spinal cord scans.
1.1 Description
Key Words: Amyotrophic lateral sclerosis, spinal cord, grey matter segmentation, deep learning, MRI, automated segmentation, neuroimaging, and Bio markers.
The goal of this study is to employ a Deep Dilated Convolution Neural Network (DDCNN) architecture to automatically segregate spinal cord gray matter from MRI data. Due to their small receptive fields, traditional convolution models have trouble capturing fine-grained structural information in medical pictures. Dilated (Atreus) convolutions, which increase the receptive field without raising computing costs or sacrificing spatial resolution, are used by the suggested model to get around this. The network learns to differentiate gray matter from surrounding white matter and background tissues by processing axial MRI slices of the spinal cord. The model can accurately identify intricate anatomical features by capturing multi-scale contextual information thanks to dilated convolution layers. Using methods like data augmentation and post-processing to lower noise and improve borders, the system is trained on annotated medical datasets. The final model generates smooth, clinically effective gray-matter masks with good segmentation accuracy.
1. INTRODUCTION When it comes to sending neural messages from the brain to the rest of the body, the spinal cord is essential. It is in charge of processing sensory data, regulating reflexes, and coordinating motor activities. White matter and gray matter are the two primary components of the spinal cord. Gray matter comprises the nerve cell bodies that carry out processing tasks, whereas white matter is principally in charge of signal transmission. Understanding and diagnosing a variety of neurological problems, such as spinal cord injuries, neurodegenerative diseases (including multiple sclerosis and amyotrophic lateral sclerosis), and congenital spinal cord disorders, depends on gray matter segmentation. However, manually segmenting gray matter in spinal cord MRI images is a
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 1022