Naslov
Bladder Cancer Recognition Using SIFT and SURF Features

Autori
Jelena Musulin, Klara Smolić, Zlatan Car

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
My First Conference book of abstracts / - Rijeka, 2020, 24-25

Skup
4th edition of annual conference for doctoral students of engineering and technology „MY FIRST CONFERENCE“

Mjesto i datum
Rijeka, Hrvatska, 24.09.2020

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Recenziran

Ključne riječi
Bladder cancer classification, Multilayer Perceptron, Scale-Invariant Feature Transform, Speeded-Up Robust Features

Sažetak
Bladder cancer is a heterogeneous disease, which starts when cells that make up the urinary bladder start to grow out of control [1]. The standard procedure for the diagnosis of bladder cancer is called cystoscopy, and it’s combined with standard biopsy and histopathological examination of the biopsied tissue. Such a procedure can be invasive, therefore a procedure that involves an optical examination of urinary bladder mucosa is imposed, since it is less invasive approach [2]. Early detection of bladder cancer can increase the chance of survival among people, but new technologies could be expensive and time-consuming. For this reason, image processing techniques with the aid of Artificial Intelligence (AI) tools are widely used in different medical fields for detection of cancer in early stages [3]. The application of AI algorithms in medicine has been receiving much attention lately because of the possibility of automated medical diagnosis with high accuracy. In this research, integration of Multilayer Perceptron (MLP) along with Scale- Invariant Feature Transform (SIFT) and Speeded- Up Robust Features (SURF) algorithm is proposed for bladder cancer classification. The dataset was obtained from the Clinical Hospital Center in Rijeka and consists of 1316 image dataset, where the training set consists of 1200 images while the testing set consists of 116 images. SIFT and SURF were first applied to extract features which were later used as the input vector for MLP. The maximum recognition accuracy of 99.92% has been achieved with a combination of MLP and SIFT algorithm (96x128) while using MLP and SURF algorithm (64x64) maximum recognition accuracy of 96.33% has been achieved

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo, Temeljne medicinske znanosti

POVEZANOST RADA