Texture Feature Extraction of CTB Radiograph Image Using Derivative Gaussian filter With NN Classification to Diagnose Osteoporosis
Kavita. Avinash. Patil1, K. V. Mahendra Prashanth2 

1Kavita.Avinash.Patil, Department of Electronics & Communication. Visvesvaraya Technological University Karnataka, East Point College of Engineering & Technology, Bangalore, Karnataka State.
2Dr. K. V. Mahendra Prashanth , Professor & HOD Electronics & Communication Engineering, Visvesvaraya Technological University Karnataka, SJB Institute of Technology College, Bangalore, Karnataka.
Manuscript received on 02 June 2019 | Revised Manuscript received on 10 June 2019 | Manuscript published on 30 June 2019 | PP: 60-67 | Volume-8 Issue-8, June 2019 | Retrieval Number: G5866058719/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Texture feature based classification of bone radiograph images to diagnose osteoporosis is a challenge for researchers. The Calcaneus Trabecular Bone (CTB) micro architecture is one of the important factor to recognise osteoporosis because of which the entire structure of CTB is impaired. The method organized in this paper has two parts, to classify the normal (control cases) and abnormal (osteoporotic cases) CTB images. The first part of this method is to distinguish the CTB micro architecture predisposing using first and second order directional derivative of Gaussian filter with different standard deviation to obtain the extremum (maximum and minimum) responses. To differentiate the texture features of an image by transformation of extremum responses using linear and nonlinear operations on extremum responses. To reduce the entire dimension of the texture features, quantization and adjacent scale coding with weighted multipliers are used to reduce the intensity variations of features. The second part of this method uses the reduced histogram features as a training data set to classify the normal and abnormal CTB images using nearest neighbourhood (NN) classifier. The tested results gives effective classification accuracy of CTB images with less texture feature dimension. Since this method uses weighted multipliers and quantization to reduce the feature dimension of image texture. The selection of weighted multiplier plays an important role to improve the classification accuracy to diagnose osteoporosis for an input noisy and noiseless image. The overall system proposed in this paper results better diagnose accuracy than the existing system.
Keyword: Classification, CTB images, Gaussian derivativefilters, Osteoporosis ,Quantization, Texture features.
Scope of the Article: Classification