Detection of lymph node involvement has a significant negative impact on the treatment outcome for patients with head and neck squamous-cell carcinomas (SCC). Metastatic lymph nodes can be detected with high sensitivity by 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET). However, 18F-FDG uptake is not only increased in metastasis. For example, inflammation in irradiated regions have increased 18F-FDG uptake obscuring possible residual tumour. Additional imaging might improve the specificity of the 18F-FDG PET images.

Magnetic resonance imaging (MRI) is currently used in head and neck imaging to provide anatomical reference for the 18F-FDG PET. However, early detection of metastatic lymph nodes based on anatomical MRI is difficult. Physiological MRI uses sequences that allow for the measurement of physiological processes within a tissue. Changes of these physiological processes within lymph nodes might indicate a metastasis. Two types of physiological MRI were of interest for this study, diffusion weighted imaging-intravoxel incoherent motion (DWI-IVIM) and dynamic contrast enhanced (DCE)-MRI. Quantified data from these physiological MRI acquisition might provided added predictive value in head and neck PET/MR for the detection of metastatic SCCs lymph nodes.

For the prospective study, subjects were included between September and December 2022 scheduled for a standard diagnostic PET/MR exam of the head and neck region with (suspected) SCC. The DWI-IVIM and DCE-MRI sequences were additionally acquired during the standard diagnostic PET/MR. Lymph nodes identified by an experienced radiologist or nuclear medicine physician were included for processing and analysis. Metastatic lymph nodes were determined based on the pathology results. Each included lymph was described non-suspect, suspect or pathological based on the standard PET/MR images.

Masks of only the lymph node region of interests were used for processing of the DWI-IVIM and DCE-MRI images. From the DWI-IVIM model diffusion D and pseudodiffusion D* were calculated. Quantitative permeability parameters for DCE-MRI were modelled with the extended Tofts pharmacokinetic model. The outcome measures included permeability Ktrans, Kep and volume fraction Vp. Descriptive parameters were calculated for each lymph node from each quantitative parameter map.

In total 14 subjects were scanned with the additional physiological MRI sequences, of which ten were included with 41 analysed lymph nodes. Pathology was acquired for 13 lymph nodes, three malignant and ten non-malignant. Significant differences were found between the malignant and non-malignant lymph nodes for quantitative parameters of DWI-IVIM D, DWI-IVIM D*, 18F-FDG SUV, DCE-MRI Ktrans and DCE-MRI Kep. Furthermore, six of the ten non-malignant lymph nodes were suspected based on the standard PET/MR images.

The physiological MRI showed multiple quantitative parameters correlating to malignancy. Moreover, the relatively high number of suspected non-malignant lymph nodes depicts the potential specificity improvement physiological MRI can provide for head and neck PET/MR imaging. However, a larger database of analysed lymph nodes is required to determine more concrete correlations.

Literature on medical imaging segmentation claims that self-attention-based Transformer blocks perform better than convolution in UNet-based architectures. This recently touted success of Transformers warrants an investigation into which of its components contribute to its performance. Moreover, previous work has a limitation of analysis only at fixed data scales as well as unfair comparisons with others models where parameter counts are not equivalent. This work investigates the performance of the window-Based Transformer for prostate CT Organ-at-Risk (OAR) segmentation at different data scales in context of replacing its various components. To compare with previous literature, the first experiment replaces the window-based Transformer block with convolution. Results show that the convolution prevails as the data scale increases. In the second experiment, to reduce complexity, the self-attention mechanism is replaced with an equivalent albeit simpler spatial mixing operation i.e. max-pooling. We observe improved performance for max-pooling in smaller data scales, indicating that the window-based Transformer may not be the best choice in both small and larger data scales. Finally, since convolution has an inherent local inductive bias of positional information, we conduct a third experiment to imbibe such a property to the Transformer by exploring two kinds of positional encodings. The results show that there are insignificant improvements after adding positional encoding, indicating the Transformers deficiency in capturing positional information given our data scales. We hope that our approach can serve as a framework for others evaluating the utility of Transformers for their tasks.

Ablation is a medical procedure to treat liver cancer where a needle-like catheter has to be inserted into a tumor, which will then be heated or frozen to destroy the tumor tissue. To guide the catheter, Ultrasound(US) imaging is used which shows the catheter position in real time. However, some tumors are not visible on US images. To make these tumors visible, image fusion can be used between the inter-operative US image and a pre-operative contrast enhanced CT(CECT) scan, on which the tumors are visible. Several methods exist for tracking the motions of the US transducer relative to the CECT scan, but they all require a manual initialization or external tracking hardware to align the coordinate systems of both scans. In this thesis we present a technique for finding an initialization using only the image data. To achieve this, deep learning is used to segment liver vessels and the boundary of the liver in 3D US images. To find the rigid transformation parameters, the SaDE evolutionary algorithm was used to optimize the alignment between the blood vessels and the liver boundary between both scans.