Classification Algorithms for the Detection of the Primary Tumor Based on Microscopic Images of Bone Metastases


Sladan Kantar, Aleksandar Pluskoski, Igor Ciganovic and Jelena Vasiljevic, University of Belgrade, Serbia


This paper presents the analysis of techniques for microscopic images in order to find a primary tumor based on the of bone metastases. Was done algorithmic classification into three groups, kidney, lung and breast. In order to speed up the treatment of the patient and easier for doctors and therefore reduce room for human error. Digital microscope images of bone metastases were analyzed, for which it is known that the primary tumor is in one of the three human organs: kidney, lung or breast. We tested several solutions for classification, were tested two methods of image analysis. Multifractal analysis and convolutional neural network. Both methods were tested with and without preprocessing image. Results of multifractal analysis were then classified using different algorithms. Images were processed using CLAHE and kmeans algorithm. At the end, the results obtained using a variety of techniques are presented.


Cancer classification, Microscopic images, Image preprocessing, Multifractal analysis, Classification algorithms

Full Text  Volume 7, Number 8