The PhD defence and trial lecture will be streamed. The chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience physically present in the auditorium to ask ex auditorio questions.
→ Live streaming of trial lecture and disputation
→ Request for thesis copy (available until the disputation starts)
Trial lecture
Time and place: October 8, 2024; 10:15 AM, - Lille Fysiske auditorium (V232) - Fysikkbygningen
Title: "The clinical work-up in neuro-oncology – the role of medical imaging"
Main research findings
Time management is crucial in the ever-changing clinical workflow. Artificial intelligence has shown great promise in automating time-consuming clinical tasks, improving MR image quality, and reducing imaging times, all to improve brain tumor assessment. In this thesis, artificial intelligence was used to automate detection and annotation of brain metastases and to reconstruct MRI from images with highly reduced imaging times.
Consistent with previous works, the methods developed in this thesis have been shown to accurately segment and detect brain metastases from different clinical protocols and number of MR sequences across two medical institutions. This was further extended with semi-supervised segmentation, where it was shown that additional unlabeled data further improved the segmentation and detection performance. In addition, we developed an artificial intelligence-based MRI reconstruction model that was evaluated on MRIs imaged during tumor resection surgery. The expert readers were overall positive, but additional work is needed to improve the resulting image quality due to the challenging MRI setting during tumor resection surgery.
The results from this thesis suggest that artificial intelligence can be successfully used to segment and detect brain metastases, and to reconstruct MRI in a highly challenging unorthodox setup.
Hovedfunn (Norwegian)
I den stadig endrende kliniske hverdagen er effektivt tidsbruk kritisk. Kunstig intelligens har vist lovende resultater innen automatisering av tidskrevende kliniske oppgaver, forbedre bildekvaliteten i MR og å redusere avbildningstiden under MR, alt dette for å forbedre den kliniske evalueringen av hjernekreft. I denne oppgaven blir kunstig intelligens baserte metoder utviklet og evaluert for segmentering og deteksjon av hjernemetastaser og rekonstruksjon av høyt akselererte MRI-opptak.
Likt som tidligere arbeid innenfor fagfeltet har metodene utviklet i denne avhandlingen vist evnen til å kunne nøyaktig segmentere og detektere hjernemetastaser fra MR bilder tatt med varierende antall sekvenser og forskjellige kliniske protokoller fra to forskjellige sykehus. Segmenteringen av hjernemetastaser ble videre bygget på med semi-supervised segmentering hvor det ble vist at ekstra ikke-annotert data kunne forbedre den automatiske segmenteringen og deteksjonen av hjernemetaster. Utover segmentering har vi utviklet en kunstig intelligens basert modell for rekonstruksjon av MRI tatt under reseksjonskirurgi av hjernetumor. Ekspertene som evaluerte bildene var positive, men metoden krever ytterligere forbedringer grunnet de vanskelige forholdene under en såkalt intraoperativ MRI.
Resultatene i denne oppgaven tilsier at kunstig intelligens kan bli brukt for å segmentere og detektere hjernemetastaser samt rekonstruere MRI under høyst uortodokse og vanskelige forhold.
Adjudication Committee
- Associate Professor Mattijs Elschot, MR cancer group, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology, Norway
- Dr. Kicky van Leeuwen, Diagnostiv Image Analysis Group, Radboud University Medical Center, Netherlands
- Associate professor Nina Frederike Jeppesen Edin, Department of Physics, The Faculty of Mathematics and Natural Sciences, Norway
Supervisors
-
Dr. Endre Grøvik, Department of Radiology, Helse Møre of Romsdal, Norway
-
Prof. Atle Bjørnerud, Computational Radiology & Artificial Intelligence, Oslo University Hospital, Norway
-
Dr. Kyrre Eeg Emblem, Department of Physics and Computational Radiology, Norway
-
Professor Anne Helene Schistad Solberg, Department of Informatics, University of Oslo, Norway
Chair of defence
Head of Department Susanne Friederike Viefers, Department of Physics, University of Oslo, Norway
Candidate contact information
LinkedIn: https://www.linkedin.com/in/jon-andre-ottesen/
Contact information to Department: Line Trosterud Resvold