Event Time: Tuesday, April 29, 2025 | 10:00 am - 12:00 pm
Event Location:
(In-person): 15th Floor Meeting Room, BC Cancer Research Institute, 675 W 10th Ave, Vancouver, BC V5Z 0B4
Add to Calendar 2025-04-29T10:00:00 2025-04-29T12:00:00 Deep-Learning-Guided Image Generation, Enhancement and Analyses: With Applications to Nuclear Medicine Imaging Event Information: Abstract:Modern nuclear medicine imaging pipeline involves image generation, enhancement, and analysis, each facing challenges in reconstruction fidelity, quantitative reliability, and automated interpretation. This thesis presents deep learning approaches to overcome these limitations throughout the nuclear medicine pipeline. In Chapter 2, we propose **DIP-SPECTNet**, an unsupervised approach leveraging the inductive bias of convolutional networks to denoise SPECT projections without paired training data. Exploiting the deep image prior technique, our method separated noiseless photopeak projections from Poisson noise while preserving anatomical features in low-count regimes. In Chapter 3, we present **DAWN-FM**, a novel framework for solving ill-posed inverse problems through data and noise-aware flow matching. Incorporating embeddings for measured data and noise characteristics into the training process, we solved deblurring and tomography inverse problems in the presence of noise. Our method’s ability to sample from the learned posterior enables the exploration of the solution space and facilitates uncertainty quantification. In Chapter 4, we develop a **comprehensive framework for evaluating deep-learning methods for lymphoma quantitation**. Rigorous comparison with expert annotations showed that networks match physician performance for large lesions while revealing shared limitations in detecting small, low-contrast abnormalities. In Chapter 5, we introduce **IgCONDA-PET** to overcome annotation scarcity for training networks for anomaly detection in PET. Our weakly-supervised approach combined attention-mechanisms with counterfactual diffusion modeling to localize lesions without pixel-level supervision, outperforming other competing methods across diverse cancer phenotypes. In Chapter 6, we propose **Thyroidiomics**, a machine-learning framework for thyroid disease classification using scintigraphy imaging. Integrating deep segmentation with radiomics analysis achieved physician-level accuracy while reducing additional test requirements and assessment time. Finally, in Chapter 7, we present **Multiscale Stochastic Gradient Descent**, addressing computational challenges in high-resolution network training. The computation of the gradient of loss using a coarse-to-fine strategy with novel mesh-free convolutions enabled efficient convergence while maintaining resolution consistency, which is crucial for training deep learning models where training compute is often the bottleneck, especially in the case of high-resolution imaging. Together, these AI solutions have the potential to enhance nuclear medicine from acquisition to diagnosis by addressing core challenges in data quality, annotation needs, and computational efficiency, bridging innovation with clinical implementation.  Event Location: (In-person): 15th Floor Meeting Room, BC Cancer Research Institute, 675 W 10th Ave, Vancouver, BC V5Z 0B4
Event Time: Tuesday, April 29, 2025 | 1:00 pm - 2:00 pm
Event Location:
HENN 318
Add to Calendar 2025-04-29T13:00:00 2025-04-29T14:00:00 High-parallel field spectrometer extends capability of TRIUMF beta-detected nuclear magnetic resonance Event Information: Abstract: This thesis reports the design and implementation of a new high-parallel field spectrometer, which extends the capability of TRIUMF beta-detected nuclear magnetic resonance (beta-NMR) facility with fields up to 200 mT parallel to the sample surface.  The magnetic field range and spectrometer configuration are designed to allow nm-scale depth-resolved studies of superconducting RF (SRF) materials up to the superconducting critical field of Nb, the main material for SRF cavities. SRF cavities are the main technology behind high-energy and high-power linear accelerators (linacs) worldwide, allowing charged particle acceleration using radiofrequency (RF) accelerating gradient up to several tens of MV/m. The accelerating electric fields along the cavity axis are accompanied by the RF magnetic fields parallel to the cavity wall, which induce dissipation due to the penetrating magnetic fields contained within 100 nm layer of the cavity surface in the flux-free superconducting Meissner state. The ability of the SRF materials to screen and contain magnetic fields within the penetration depth, as well as the maximum field limit before strongly dissipative magnetic fluxes enter the bulk of the material (and induce RF quenches of the SRF cavity), have been found to be very sensitive to different types of surface treatments. The magnetic field-dependent surface dissipation affects the operational cost of SRF cavities, and the maximum magnetic field that can be sustained in the Meissner state ultimately limits the maximum accelerating gradient of SRF cavities. Various surface treatment recipes using heat treatment and/or impurity diffusion have been developed which demonstrate enhanced performance of SRF cavities. Complete understanding of the underlying mechanism of this enhancement, however, requires a more controlled microscopic study of the near surface layer. Depth-resolved measurements of the magnetic field screening below the surface of SRF materials are made possible with this new spectrometer, which combines local magnetic field measurements via spin-polarized radioactive ion beam (RIB) produced at TRIUMF ISAC facility (commonly uses Li-8 positive ions), and the suitable spectrometer which allows high-parallel field combined with implantation depth-control of the probing ions via deceleration of the their momentum. The new spectrometer requires modifications of the existing beta-detected nuclear quadrupole resonance (beta-NQR) beamline, and an additional ~1 m beamline extension. The magnetic field configuration parallel to the sample surface (and initial spin polarization of the probe) but transverse to the beam momentum deflects the beam vertically and requires compensation via electrostatic steering of the RIB to deliver beam to the target sample. The details of design, assembly, various stages of beamline installation, and operations of the various elements both along the beamline and the new spectrometer are all presented in this thesis.Also provided are the test results of the new/modified components, and the commissioning results proving the functionality of the new spectrometer using RIB. Depth-resolved measurements on two Nb samples with different surface treatments typically applied to SRF cavities have been performed on the new spectrometer up to the maximum available fields (of 200 mT). The results demonstrate the sensitivity of the beta-NMR technique in characterizing the magnetic field screening, and provide a working method for future SRF study. These results also provide comparison of the different screening responses of various Nb samples to the applied magnetic fields due to the modified surface layers. The change in the magnetic penetration depth with increased fields are then compared to various theoretical predictions on the role of the modified surface. Outlook on future experiments on different SRF materials (such as layered superconductors), as well as potential applications of the new spectrometer for other materials are proposed.  Event Location: HENN 318