HIC-net: A deep convolutional neural network model for classification of histopathological breast images
Computers and Electrical Engineering, cilt.76, ss.299-310, 2019 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 76
- Basım Tarihi: 2019
- Doi Numarası: 10.1016/j.compeleceng.2019.04.012
- Dergi Adı: Computers and Electrical Engineering
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
- Sayfa Sayıları: ss.299-310
- Anahtar Kelimeler: Cancer classification, CNN, Convolutional neural networks, Histopathological image, Whole-slide
- Ankara Hacı Bayram Veli Üniversitesi Adresli: Evet
Özet
In this study, a convolutional neural network (CNN) model is presented to automatically identify cancerous areas on whole-slide histopathological images (WSI). The proposed WSI classification network (HIC-net) architecture performs window-based classification by dividing the WSI into a certain plane. In our method, an effective pre-processing step has been added for WSI for better predictability of image parts and faster training. A large dataset containing 30,656 images is used for the evaluation of the HIC-net algorithm. Of these images, 23,040 are used for training, 2560 are used for validation and 5056 are used for testing. HIC-net has more successful results than other state-of-art CNN algorithms with AUC score of 97.7%. If we evaluate the classification results of HIC-net using softmax function, HIC-net success rates have 96.71% sensitivity, 95.7% specificity, 96.21% accuracy, and are more successful than other state-of-the-art techniques which are used in cancer research.