Determination of kidney area independent unconstrained features for automated diagnosis and classification
K. Bommanna Raja, M. Madheswaran · 2007
An effort has been taken to establish a set of unconstraint features that are independent to kidney area variations for automated diagnosis and classification of kidney categories. The highly reported three kidney categories namely normal (NR), medical renal diseases (MRD) and cortical cysts (CC) are considered and images are acquired using ultrasound as imaging modality. A pre-processing procedure that includes image segmentation, rotation and unbounded pixels elimination has been employed to retain the pixels of kidney region. Using six feature extraction techniques, 36 features are extracted to study the texture patterns of kidney region. For making automated diagnosis and classification, the multilayer back propagation network and hybrid fuzzy-neural module are developed. The dependency of features on kidney area is studied by performing F-test, estimating Pearson product moment correlation coefficient and calculating R-squared value between two data sets with sixth order polynomial regression analysis. The result obtained shows that most of the features are independent to kidney area variations and can reliably be used for computer-aided diagnosis.