Classification of Lumbar Disc Disorder from MRI and CT images using Iterative Differential Approach

Abstract

The prime objective of this study is to recognize and segment Lower Lumber Spine from the collected sample and then perform classification to separate affected and non-affected regions by lower lumbar spine disease. The proposed model first of all performs identification and separation of regions from the sample. This was performed by converting RGB cell image into gray colour scale. Background subtraction algorithm was applied to extract only cell structures from the image by eliminating the background completely and region-props. In second phase, features from the segmented regions were extracted. These features include homogeneity, contrast, energy, correlation and some hybrid features. In the third phase, digital differential analyzer optimization(DDAO) algorithm was applied to select the significant features. In the final phase, different classifiers were used to validate the performance of proposed optimization approach. The proposedmodel was applied on well-known benchmarked dataset. The obtained results corresponding to identification and separation were 92, 88 and 80% of segmentation accuracy, sensitivity and specificity, respectively. This result was best among other published papers worked on same dataset. Classification accuracy was notably higher as compared to other models not following DDA optimization algorithm. Validation of results was further extended through feature reduction ratio and still remarkable results in terms classification accuracy of 90% was achieved.