Optimizing Extracted Deep Features for Classification of Colon Cancer Histopathology Images

Doğan R. S., Güzel Ö. F., Altuntop M. S., Taşdemir K.

5th International Conference on Data Science and Applications (ICONDATA'22), Muğla, Turkey, 7 - 11 September 2022, pp.59

  • Publication Type: Conference Paper / Summary Text
  • City: Muğla
  • Country: Turkey
  • Page Numbers: pp.59
  • Abdullah Gül University Affiliated: Yes


Colorectal cancer is about ten percent of all cancer cases recorded worldwide and is the third most common type of cancer. Classification of medical images of colon cancer with high accuracy is one of the critical factors that determine the course of the disease. Colonoscopy is a method used for colon cancer diagnosis, and the optical and pathological images obtained from this provide important information in determining the location and type of tissue. Accurate classification of images sent to the pathology unit, increasing processing speed and reducing human error is very important. There are many studies on classification, and models with high accuracy have been developed in these studies using machine learning and deep learning methods. In real-time applications, the size and complexity of the model affects the speed of output from the model. In this study, it is aimed to train fewer layers faster and more efficiently, to get closer to realtime implementation and to reach accuracy rates similar to pretrained models. In addition, the effect of color normalization applied to histopathology images on the accuracy rate was also investigated.