Main Objective
Classification of prostate cancer tumors in multiparametric MRI using Deep Learning techniques
Detecting benign (ClinSig False) and malignant (ClinSig True) prostate tumors using convolutional neural networks
White: the pixel belongs to the class ; Black: does not belong to the class ; Database: PROSTATEx
Tasks Performed
Under the supervision of Eric Moulton from Guerbet and Nicolas Brunel from ENSIIE school
- Literature review, familiarisation with prostate imaging and pathologies
- Medical image preprocessing: normalisation, data augmentation, segmentation
- Development of convolutional network models in Python, with the TensorFlow 2.0 library
- Transformation of radiology and biopsy reports into structured data
- Use of Git and Cloud Computing services (Microsoft Azure)
- Regular presentation of the progress during the team meeting
- Internship report writing
Technologies Used
- Python 3.7, Tensorflow 2.0 and scikit-learn libraries
- Microsoft Azure
- Github