Highly-efficient technique for automatic segmentation of X-ray bone images based on fuzzy logic and an edge detection technique

Abstract

Development of medical image segmentation techniques has become one of the most important challenges in many applications that employ computers-based medical image analysis techniques. However, most of current-existing medical image segmentation techniques still have poor efficiency and complexity in calculations. A new technique for X-ray bone image segmentation has been presented in this paper. The proposed technique is designed to generate a highly-efficient quality of bone extraction from X-ray bone image at low calculation burdens. The proposed approach begins with obtaining the inverse of the original X-ray bone image then applying bounded-product operator and fuzzy controller to control the contrast of the inverse image. At the same time, an adaptive thresholding of the gradient magnitude of the original X-ray bone image is achieved to detect the edges of the bone regions. Accordingly, the bone edges are integrated with the contrasted image to give the bone segmented image. To ensure the efficient quality of the proposed algorithm, more than sixty-one of X-ray bone images were tested using much vision and statistical investigations. Then, evaluations employed several measures such as Dice similarity coefficient index, Confusion Matrix, Accuracy, Precision, Sensitivity, Specificity, and processing speed. Furthermore, results obtained using the proposed were compared to those of conventional image segmentation techniques (such as Watershed segmentation, Otsu-thresholding, K-means, and fuzzy C-means). Comparison results demonstrated the superiority of the proposed technique over other conventional techniques in both of quality and processing speed. All obtained results were obtained using MATLAB R20014a over Windows XP with processing speed 2 GHz. The high efficiency and processing speed of the proposed technique makes it such a promising solution to be implemented in many real applications.