Creative Commons License

Cengiz E., Yaylak F., Gülbandılar E.

Eskişehir Osmangazi Üniversitesi mühendislik ve mimarlık fakültesi dergisi (online), vol.30, no.3, pp.441-453, 2022 (Peer-Reviewed Journal) identifier


Recent advances in machine learning, particularly with regard to deep learning, help to recognize and classify objects in medical images. In this study, endoscopy images were examined and deep learning method was used to classify healthy and polyp cells. For the proposed  system,  a  database  was  created  from  the  archives  of  General  Surgery Department Endoscopy Unit in Kutahya Evliya Celebi Training and Research Hospital. The database contains 93 polyps and 216 normal images from 54 archive records. For image  multiplexing, a total  of 1236 images were obtained  by rotating each  image 90 degrees around its axis. K-fold Cross Validation method was used to reduce the variability of performance results. In this study, 48 different models were created by using differe nt activation  and  optimization  functions  to  find  the  best  classification  model  in  deep learning.  According  to  the  experimental  results,  it  was  observed  that  accuracy  of  the models depends on the selected parameters; the  best model  with the accuracy rate of 91% was obtained with 64 neurons in the hidden layer, ReLU activation function and RmsProp optimization method whereas the worst model with the accuracy rate of 76% was  obtained  with  32  neurons  in  the  hidden  layer,  Tanh  activation  and  RMSprop optimization  functions. Accordingly, classification performance of polyp images can   be optimized by utilizing different activation and optimization methods during the design of deep learning models.