Mehmet C. Onbaşlı - Tumor morphology recognition using transfer learning on pretrained models
|Study location||Turkey, Istanbul|
|Type||Summer Research Program - High School, full-time|
At least 2 reference(s) must be provided.
The student is going to use train Raman spectrum measurements on open source tumor morphology classifier networks. Common cancer morphologies can already be observed and diagnosed by pathologists via haematoxylin and eosin staining and microscope image inspection. In this project, the student is going to learn how to train a convolutional neural network and use transfer learning to enhance the classification accuracy of a morphology-based tumor classification model by training with the information on chemical bond profile (Raman spectra). The project includes coding in Python using Keras library in Tensorflow and the use of high-performance computational clusters on Koç University campus.