Defence of doctoral thesis in the field of neuroscience and biomedical engineering, Msc (Tech) Jan Schreier
Title of the doctoral thesis is "Clinical Applicability of Deep Learning for Organ-At-Risk Segmentation in Radiotherapy Planning"
With the number of cancer cases rising globally, there is an increasing need for automation in the treatment planning process. This dissertation evaluates the clinical applicability of automatic organ delineation for radiotherapy planning and gives evidence that deep learning can be used to a) reduce the time spend for organ delineation, b) performs on par with clinical practice for the male pelvis on cone-beam CT and c) generalises to other institutions in most cases.
For this, we developed a state-of-the-art deep neural network that can segment organs-at-risk in the female breast, female pelvis and male pelvis region. We then measured the time needed to correct the resulting structures and compared these to the delineation from scratch. Further, doctors were scoring structures produced by our deep neural network and those originating from clinical practice in the male pelvis. Here, we could see that the scores for the neural network are equal or better. Finally, we compared how the neural network performs on patients from clinics that were not present during the training of the neural network. This provides guidance to which extend neural networks developed by a third party can be readily applied in clinical practice. Here, we see that, for the inner organs tested, the network performs well as long as the contouring practice is coherent
Opponent: Professor John A. Lee, Université catholique de Louvain, Belgium
Custos: Professor Lauri Parkkonen, Aalto University School of Science, Department of Neuroscience and Biomedical Engineering
Contact information of the doctoral candidate: Jan Schreier, [email protected], 041 7420632
The defence will be organised via remote technology (Zoom). Link to the defence
Electronic doctoral thesis (aaltodoc.aalto.fi)
The doctoral thesis will be publicly displayed 10 days before the defence in the electronic publication archive of Aalto University (aaltodoc.aalto.fi).