McMaster University
OPTIMIZING MAGNETIC RESONANCE IMAGING SAMPLING PATTERNS FOR ACCELERATED MP2RAGE BRAIN T1 MAPPING
Abstract
dc:description.abstractThis work targets acceleration of MP2RAGE‐based brain T1 mapping by data‑driven k‑space sampling mask design. This is achieved through a new optimization algorithm that selects measurements to maximize joint reconstruction fidelity of the two MP2RAGE anatomical volumes under a fixed scan time (acceleration) budget. Using public MP2RAGE datasets, missing raw data (k‑space) was synthesized from images to emulate acquisition for training and evaluation. A fitness function was developed that scores candidate masks by the joint quality of reconstructed INV1 and INV2 images and derived T1 maps, and existing sampling‑pattern designs were improved upon by combining variable‑density Poisson‑disc (VDPD) sampling with an adapted bias‑accelerated subset‑selection procedure. The optimization method is adapted to MP2RAGE’s paired‑contrast physics and joint reconstruction, allowing the two masks to be optimized together while respecting a fixed total acceleration. This thesis presents (i) an optimization framework for generating paired sampling masks specialized for MP2RAGE and joint reconstruction; (ii) a data processing pipeline that can convert limited image data into more realistic k‑space for algorithm development; and (iii) the resulting data‑driven masks that consistently outperform VDPD alone, having the potential to yield shorter scans with minimal compromise in image quality. These advances contribute new knowledge on how anatomy‑aware, learned sampling can be tailored to multi‑contrast acquisitions and joint reconstructions rather than treated as single‑contrast problems. By improving the time-efficiency of MP2RAGE T1 mapping, this work supports quantitative neuroscience – including studies of brain aging, learning, and disease via biomarkers such as myelin content. The approach provides a practical route to faster T1 mapping and establishes a general recipe for extending data‑driven sampling optimization to other quantitative MRI protocols that benefit from joint, multi‑contrast reconstruction.
Degree
thesis:*- Department dc:contributor.department
- Physics and Astronomy
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Karapetov, Artur
- Advisor dc:contributor.advisor
-
- Rowley, Christopher
Subjects
dc:subject × 18Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Identifier URI
- https://doi.org/10.71548/25186