Abstract:
Elderly individuals experience lower back and leg pain, often due to sciatica, due to nerve root compression, main cause of which is lumbar spondylolisthesis, a vertebral...View moreMetadata
Abstract:
Elderly individuals experience lower back and leg pain, often due to sciatica, due to nerve root compression, main cause of which is lumbar spondylolisthesis, a vertebral slippage. Accurate diagnosis is crucial for treatment planning. Deep learning algorithms are commonly used in medical image classification for musculoskeletal problem analysis. This study's aims to estimate the severity of spondylolisthesis using Magnetic Resonance Imaging modality (MRI) images and deep learning networks. The proposed computer-assisted diagnosis system provides accurate grading and minimize interobserver differences. The deep learning algorithms that have been used in this study are AlexNet, VGG16, VGG19 and ResNet50. The classifier was built to classify four classes (normal, grade 1, grade2 and retrolisthesis). The models were trained using a dataset constructed from authentic 245 MRI images of male and female subjects aged 25–80 years old (154 normal, 53 gradel, 4 grade2, and 34 retrolisthesis). Data augmentation was used to increase the number of images in the training dataset and to generate a diverse and unbiased dataset, resulting in a total of 514 images (308 normal, 106 grade1,32 grade2, and 68 retrolisthesis). The classification results show that VGG 16 achieved the highest accuracy of 97.43 % compared to other algorithms.
Published in: 2024 IEEE 8th International Conference on Signal and Image Processing Applications (ICSIPA)
Date of Conference: 03-05 September 2024
Date Added to IEEE Xplore: 26 September 2024
ISBN Information: