Bilkent University
Department of Computer Engineering

---

M.S.THESIS PRESENTATION

 

PET-MRI Integration for Metabolic Activity Localization and Multimodal Image Fusion

 

Yasir Ali Khan
Master Student
(Supervisor:Assoc. Prof. Dr. A. Ercüment Çiçek )

Computer Engineering Department
Bilkent University

Abstract: Positron Emission Tomography (PET) and Magnetic Resonance Imaging (MRI) provide complementary functional and anatomical information for the assessment of neurological disorders. While FDG-PET reflects cerebral metabolic activity, its relatively low spatial resolution leads to partial volume effects, causing cortical and subcortical metabolic signals to appear spatially diffused and to extend into surrounding white matter. This limitation hampers accurate localization of metabolic abnormalities, which is critical in applications such as epilepsy evaluation. Although hybrid PET/MRI systems enable simultaneous acquisition, effective integration of these modalities remains challenging due to their inherent resolution mismatch. In this work, a two-stage PET-MRI integration framework is proposed that explicitly decouples metabolic activity localization from multimodal image fusion. In the first stage, PET activity is transported onto MRI-derived cortical regions to obtain anatomically consistent functional representations and mitigate signal spillover across tissue boundaries. In the second stage, the cortex-aligned PET representation is fused with MRI using an autoencoder-based architecture that learns joint structural-functional feature embeddings. This design enables effective exploitation of complementary information while preserving anatomical fidelity. Experimental evaluations on a paired PET-MRI dataset demonstrate that the proposed method consistently outperforms state-of-the-art fusion approaches in terms of quantitative metrics and qualitative assessment.

 

DATE: August 04, Tuesday @ 11:00

Place: EA 516