<b>Lumbar MRI Dataset for Disc Herniation Classification</b>

<p dir="ltr"><b>Overview</b><br>This dataset comprises lumbar spine MRI images used for binary classification of disc conditions into Normal and Herniated categories, supporting research in medical image analysis.</p><p><br></p><p dir="ltr"><b>Data Collection</b><br>All data were anonymized and used for research purposes only. The dataset consists of a to tal of 1,426 MRI samples collected from multiple sources. A subset of 800 samples was obtained from publicly available medical imaging reposito ries, including platforms such as Kaggle, the Open Access Series of Imaging Studies (OASIS), and other open-access medical imaging sources.

In ad dition, 258 samples were collected from a private hospital, while the remaining 368 samples were obtained from a governmental hospital. All data were aggregated to ensure diversity and variability in lumbar spine MRI cases, covering both normal and herniated conditions.</p><p dir="ltr"><br></p><p dir="ltr"><b>Dataset Composition</b><br>• Total: 1426<br>• Normal: 671<br>• Herniated: 755</p><p dir="ltr"><br></p><p dir="ltr"><b>Preprocessing</b><br>• Grayscale conversion to standardize image representation<br>• Resizing to a fixed resolution (200×200) for consistency<br>• Bilateral filtering to reduce noise while preserving structural details<br>• Pixel intensity normalization to the range [0, 1] for stable model training</p><p dir="ltr"><br></p><p dir="ltr"><b>Usage</b><br>This dataset is used to evaluate the performance of classical machine learning models, including LightGBM, Random Forest, and SVM, with PCA for dimensionality reduction.</p><p dir="ltr"><br></p><p dir="ltr"><b>Purpose</b><br>The purpose of this dataset is to support research in medical image classification by enabling the development and evaluation of efficient classical machine learning approaches.</p><p dir="ltr"><br></p><p dir="ltr"><b>Key Notes</b><br>• Lightweight and computationally efficient pipeline<br>• PCA-based dimensionality reduction for handling high-dimensional data<br>• Focus on classical machine learning methods instead of deep learning<br>• Balanced trade-off between performance and computational cost<br>• Suitable for research, benchmarking, and educational purposes</p><p dir="ltr"><br></p><p dir="ltr"><b>License</b><br>This dataset is made available solely for research and educational purposes and should not be used for commercial applications.</p>