Simultaneous morphological and quantitative imaging of cartilage using phase-cycled bSSFP
File(s)
Author(s)
Balaji, Krithika
Type
Thesis
Abstract
Knee osteoarthritis (KOA) is a prevalent disease diagnosed after significant cartilage degeneration has occurred. Early, non-invasive detection methods are needed to monitor progression and support the development of Disease Modifying Osteoarthritis Drugs.
Cartilage T2 is sensitive to changes in collagen hydration and organisation, making it a valuable early KOA biomarker. While conventional 2D imaging methods can estimate T2 using monoexponential fitting, they are too slow for 3D clinical imaging. As KOA affects the entire joint and cartilage is thin, rapid, high-resolution 3D techniques are needed that provide both T2 maps and morphological images for whole-joint analysis.
Quantitative Double Echo Steady State (qDESS) enables both morphological imaging and T2 mapping within approximately five minutes and is widely used in KOA imaging research. Another promising approach is phase-cycled balanced Steady State Free Precession (pc-bSSFP), which offers rapid 3D acquisition and high signal-to-noise ratio (SNR) efficiency. PLANET estimates T2 analytically using at least six complex pc-bSSFP acquisitions, which can be time-consuming to acquire.
In this thesis, I developed a Random Forests (RaFo) algorithm to estimate cartilage T2 from fewer pc-bSSFP acquisitions than PLANET. Models trained on simulated four and six phase-cycled acquisitions (RaFo-4 and RaFo-6) were evaluated on in vivo data. Both models preserved spatial T2 variations seen in reference and qDESS maps and demonstrated greater robustness at lower SNR levels.
A preliminary validation using test–retest analysis compared optimized, 5-minute qDESS and 4 pc-bSSFP acquisitions in healthy volunteers. RaFo-4 maintained spatial consistency in T2 estimation, showed improved robustness to outliers, demonstrated good repeatability, and produced comparable or better morphological image quality.
Additionally, coil combination techniques were evaluated for phase preservation, critical for accurate T2 estimation. ESPIRiT, Adaptive Reconstruction, and Intrinsic Multichannel Phase Alignment performed best.
Overall, RaFo-4 bSSFP is a promising five-minute alternative to qDESS, warranting further validation in healthy and clinical populations.
Cartilage T2 is sensitive to changes in collagen hydration and organisation, making it a valuable early KOA biomarker. While conventional 2D imaging methods can estimate T2 using monoexponential fitting, they are too slow for 3D clinical imaging. As KOA affects the entire joint and cartilage is thin, rapid, high-resolution 3D techniques are needed that provide both T2 maps and morphological images for whole-joint analysis.
Quantitative Double Echo Steady State (qDESS) enables both morphological imaging and T2 mapping within approximately five minutes and is widely used in KOA imaging research. Another promising approach is phase-cycled balanced Steady State Free Precession (pc-bSSFP), which offers rapid 3D acquisition and high signal-to-noise ratio (SNR) efficiency. PLANET estimates T2 analytically using at least six complex pc-bSSFP acquisitions, which can be time-consuming to acquire.
In this thesis, I developed a Random Forests (RaFo) algorithm to estimate cartilage T2 from fewer pc-bSSFP acquisitions than PLANET. Models trained on simulated four and six phase-cycled acquisitions (RaFo-4 and RaFo-6) were evaluated on in vivo data. Both models preserved spatial T2 variations seen in reference and qDESS maps and demonstrated greater robustness at lower SNR levels.
A preliminary validation using test–retest analysis compared optimized, 5-minute qDESS and 4 pc-bSSFP acquisitions in healthy volunteers. RaFo-4 maintained spatial consistency in T2 estimation, showed improved robustness to outliers, demonstrated good repeatability, and produced comparable or better morphological image quality.
Additionally, coil combination techniques were evaluated for phase preservation, critical for accurate T2 estimation. ESPIRiT, Adaptive Reconstruction, and Intrinsic Multichannel Phase Alignment performed best.
Overall, RaFo-4 bSSFP is a promising five-minute alternative to qDESS, warranting further validation in healthy and clinical populations.
Version
Open Access
Date Issued
2025-10-03
Date Awarded
2026-05-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Bangerter, Neal
Lally, Peter
Bharath, Anil
Sponsor
National Institutes of Health (NIH)
National Institute for Health Research (Great Britain)
Wellcome Trust (London, England)
Edmond J. Safra Philanthropic Foundation
Grant Number
NIH: R01EB002524
Wellcome Trust: 220473/Z/20/Z
Publisher Department
Department of Bioengineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)