Congratulations to Dr Wallis
Last Tuesday 1 June, David Wallis successfully defended his thesis at the Université Paris-Saclay. His examination board was chaired by Ronald Boellaard, and two other HYBRID PIs, Johan Nuyts and Julia Schnabel, were part of the jury as well.
David's thesis, entitled 'A study of machine learning and deep learning methods and their application to medical imaging' can be summarized as follows.
'We first used Convolutional Neural Networks (CNNs) to automate mediastinal lymph node detection using FDG-PET/CT scans. We built a fully automated model to go directly from whole-body FDG-PET/CT scans to node localisation. The results show a comparable performance to an experienced physician. In the second half of the thesis we experimentally tested the performance, interpretability, and stability of radiomic and CNN models on three datasets (2D brain MRI scans, 3D CT lung scans, 3D FDG-PET/CT mediastinal scans). We compared how the models improve as more data is available and examined whether there are patterns common to the different problems. We questioned whether current methods for model interpretation are satisfactory. We also investigated how precise segmentation affects the performance of the models.
Finally, we tested different methods for multi-centre learning. We compared methods using a mixed training set, transfer learning, and using ComBat feature harmonisation.'
The full thesis defence is available on Youtube.
We congratulate David on obtaining his PhD degree and wish him all the best for his future career! Of course, we also extend our congratulations to his supervisor, Irène Buvat.