Application of contour detectors in algorithms for diagnosis of bladder cancer

Rad se bavi temom primjene neuronske mreˇze za klasifikaciju slika mokra´cnog mjehura na dvije klase - karcinom i zdravo tkivo, sve u svrhu poboljˇsanja dijagnosticiranja karcinoma. Vrˇse se razna testiranja na skupovima podataka kako bi se pronaˇslo optimalno rjeˇsenje, tj. ostvarila ˇsto toˇcnija klasifikacija. Testiranje se vrˇsi na augmentiranim i neaugmentiranim skupovima podataka primjenjuju´ci razliˇcite vrste interpolacijskih metoda za skaliranje slika koji se filtriraju s filtrima za detekciju rubova te prilaˇzu kao ulazni podatak u neuronsku mreˇzu. Za programiranje se koristi Python 3 programski jezik s integriranim razvojnim okruˇzenjem PyCharm.This paper deals with the subject of usig an artificial intelligence to classify pictures of a bladder in two classes - cancer and healty tissue, all in purpose to improve diagnostics of cancer. Various tests are carried out on the datasets in order to find optimal solution, ie to achieve the most accurate classification. Testing is performed on augmented and non-augmented datasets using different types of image interpolation scalling methods that are filtered with edge detection filters and sent as an input to the neural network. Python programming language with integrated PyCharm development environment is used for programming