Gallbladder Shape Estimation Using Tree-Seed Optimization Tuned Radial Basis Function Network for Assessment of Acute Cholecystitis

In this paper, computerized scheme for automatic volume estimation of inflamed gallbladder in ultrasound images has been investigated. Diagnosis of acute cholecystitis at an early stage is an arduous task as the difference between normal shape and inflamed gallbladder shape cannot be visualized in ultrasound images. This paper comes out with an unsupervised machine learning algorithm—tree-seed optimization algorithm tuned radial basis function network for segmentation of gallbladder in ultrasound images. Tree-seed optimization algorithm, which optimizes function and parameters in real values, is a population-based stochastic search algorithm. Prior to the classification, speckle reduction and feature extraction process were successfully used. These features are then used in classification process to define the gallbladder and non-gallbladder regions. The proposed optimized classifier system is evaluated in real-time clinical datasets with cholecystitis and cholelithiasis. The inherent differentiation of the proposed intelligent classifier is analyzed using standard evaluation parameters. Comparison with expert decisions provides further evidence that the optimally tuned radial basis function network has important implications for diagnosis of acute cholecystitis.