Authors : Madhuri Avula,, L.V. Narasimha Prasad and Murali Prasad Raja
Abstract: Cancer, a condition involving unregulated cell growth is a dangerous disease. After many research studies, almost 100 different types of cancers that occur in the human body have been detected. Of these, one of the most common is bone cancer which leads to death. The detection of bone cancer is very arduous and its occurrence is unpredictable. Currently, most studies on cancer exploit data mining methods and the image processing techniques used for medical image analysis. The data and knowledge collected from large databases and related web sites have been used by many scientific researchers for formulating predictions. Association rule mining Supports Vector Machines (SVMs), fuzzy theory and probabilistic neural networks, learning vector quantization, k-means and C-means are the methods most frequently used for the detection, classification and segmentation of bone cancer. In this study, a region growing algorithm has been applied for bone image segmentation. The segmented image is further processed to provide bone cancer detection by evaluating the mean pixel intensity of identified area. Threshold values are proposed for the classification of medical images according to the presence or the absence of bone cancer. This method utilizes jpeg images but can also be applied to medical images in the original format of DICOM (Digital Imaging Communication of Medicine) after some modification. The method provides an accuracy rate of 100% and the required computation time is relatively short.
Madhuri Avula,, L.V. Narasimha Prasad and Murali Prasad Raja, 2018. Detection of Bone Cancer Using Region Growing Algorithm and Mean Pixel Intensity Thresholding. International Journal of Soft Computing, 13: 69-80.