AdaBoost-based approach for detecting lithiasis and polyps in USG images of the gallbladder

Abstract

This article presents the application of the AdaBoost method to recognise gallbladder lesions such as lithiasis and polyps in USG images. The classifier handles rectangular input image areas of a specific length. If the diameter of areas segmented is much greater than the diameter expected on the input, wavelet approximation of input images is used. The classification results obtained by using the AdaBoost method are promising for lithiasis classification. In the best case, the algorithm achieved the accuracy of 91% for lithiasis and of 80% when classifyingpolyps, as well as the accuracy of 78.9% for polyps and lithiasis jointly. © 2011 Springer-Verlag.