Continuous oxygen delivery is essential for maintaining normal brain function and viability. Although the human brain accounts for approximately 2% of the total body weight, it consumes nearly 20% of the oxygen utilized by the entire body [1]. Notably, the brain has a very limited capacity to store oxygen; instead, its oxygen utilization depends primarily on real-time extraction from incoming arterial blood. As a critical physiological parameter reflecting cerebral energy metabolism, the Oxygen Extraction Fraction (OEF) has been proposed as a potential biomarker for a range of neurological and systemic disorders, including sickle cell disease (SCD) [2,3], Alzheimer's disease (AD) [4,5], multiple sclerosis (MS) [6], brain tumor [7,8] and other neurological disorders [[9], [10], [11]]. While significant progress has been made in OEF research targeting these aforementioned disorders, studies focusing on cerebral small vessel disease (SVD) remain relatively scarce. SVD is a heterogeneous group of cerebrovascular disorders characterized by pathological alterations in small arteries, capillaries, and venules within the brain. Its clinical manifestations include lacunar infarcts, white matter hyperintensities, and cerebral microbleeds， all of which contribute to cognitive impairment, gait disturbances, and an increased risk of stroke [12]. In patients with SVD, capillary dysfunction impairs cerebral autoregulatory capacity, leading to a subsequent reduction in cerebral blood flow (CBF) [13]. Several studies have indicated that to maintain a stable cerebral metabolic rate of oxygen (CMRO₂), the brain compensates by upregulating OEF [[14], [15], [16]]. This compensatory mechanism suggests that OEF may hold promise as a potential biomarker for distinguishing SVD patients from healthy controls (HC).

Current methodologies for quantifying OEF face inherent trade-offs between accuracy, invasiveness, and clinical feasibility. The gold-standard technique, oxygen-15 positron emission tomography (15O-PET), allows for the direct quantification of OEF [[17], [18], [19], [20]]. However, its reliance on ionizing radiation and invasive administration of radiotracers severely limits its clinical applicability, particularly for longitudinal monitoring and use in vulnerable populations [21]. These limitations have driven the development of non-invasive alternatives, primarily based on magnetic resonance imaging (MRI). Within brain capillaries, hemoglobin releases oxygen to surrounding tissues along the oxygen diffusion gradient, transitioning from the oxygenated state to the deoxygenated state. This transition not only facilitates oxygen delivery but also alters the magnetic properties of iron in hemoglobin: diamagnetic oxyhemoglobin loses oxygen and converts to paramagnetic deoxyhemoglobin. MRI leverages this magnetic susceptibility difference to quantify OEF, with the T₂-relaxation-under-spin-tagging (TRUST) technique being a well-established example [22]. TRUST avoids the ionizing radiation and invasiveness inherent to 15O-PET, making it safer for repeated measurements and longitudinal studies [19]. However, a critical limitation of TRUST is that its signal is predominantly derived from the superior sagittal sinus, a large superficial venous structure. For smaller venous vessels, the low signal-to-noise ratio (SNR) of TRUST leads to substantial estimation errors, rendering it inadequate for global OEF assessment.To extend oxygenation quantification to other specific vessels, Krishnamurthy et al. developed the T₂-relaxation-under-phase-contrast (trupc) technique [23]. This approach combines T₂ relaxation measurements with phase-contrast velocity encoding, enabling vessel-specific OEF assessment and offering a more flexible alternative to global sinus measurements. However, for applications requiring high-resolution mapping of the entire brain parenchyma.

In recent years, quantitative susceptibility mapping (QSM) has emerged as a promising MRI-based technique for OEF quantification [24]. Unlike conventional MRI sequences that indirectly infer tissue properties, QSM provides voxel-level quantitative measurements of tissue magnetic susceptibility—an attribute that is highly advantageous for global OEF evaluation [25]. Some studies have combined QSM with gas challenge protocols to measure OEF, which enables global assessment but requires complex control of gas inhalation and prolonged scanning sessions [[26], [27], [28]]. To address this limitation, Zhang et al. proposed a challenge-free approach: by assuming that CMRO₂ exhibits minimum local variance, they derived OEF maps from baseline QSM and CBF data acquired via arterial spin labeling (ASL) [29]. Despite its innovation, these methods rely on some assumptions—most notably, that CMRO₂ remains constant before and after gas manipulation and that a fixed linear relationship exists between cerebral blood volume CBV and CBF [30,31]. Furthermore, registering low-resolution EPI-based ASL images to high-resolution GRE-based QSM images introduces registration errors, which propagate to and distort final OEF calculations [26].The challenge-free QQ (QSM + qBOLD) method integrates phase and magnitude information to enable voxel-wise OEF mapping, avoiding the complexities of gas challenges and multi-modal registration [20,32]. However, this approach remains sensitive to noise and relies on idealized physiological assumptions that may not hold in diseased tissues.

Advancements in MRI hardware and reconstruction algorithms have improved the spatial resolution of QSM, enabling direct OEF measurement using large cerebral venous structures. These approaches require only a single static MRI acquisition, eliminates the need for gas challenge, and is unaffected by CBF fluctuations—making it more compatible with routine clinical MRI workflows [15,16,28,[33], [34], [35]]. However, a key challenge persists: while the magnetic susceptibility signal in QSM is dominated by the strong paramagnetism of deoxyhemoglobin, venous blood also contains a large amount of weakly diamagnetic oxyhemoglobin [36]. These two substances often coexist within the same voxel in brain tissue, leading to partial cancellation of their opposing susceptibility signals. Conventional QSM methods only provide voxel-averaged susceptibility values and cannot disentangle intravoxel susceptibility sources with opposite signs. In response, advanced susceptibility separation techniques have been developed to separate contributions from paramagnetic and diamagnetic components in QSM images [37].Notably, across the literature, the susceptibility values of oxyhemoglobin show considerable variability [36,38,39]. This variability, combined with the inherently small magnitude of the diamagnetic susceptibility component in venous blood, makes OEF estimation based on the diamagnetic component prone to large relative errors from residual separation noise and systematic bias due to uncertainty in the oxyhemoglobin susceptibility shift. In contrast, the paramagnetic susceptibility component, dominated by the stronger deoxyhemoglobin signal and supported by more established susceptibility constants, provides a more robust basis for OEF quantification, as it minimizes error amplification and better reflects deoxygenation-driven susceptibility changes. Therefore, the present study focuses on OEF quantification using the paramagnetic component derived from susceptibility separation via APART-QSM.

For this reason, the present study aims to quantify OEF using the paramagnetic component derived from susceptibility separation.

To evaluate the feasibility and potential of OEF estimation derived from the paramagnetic susceptibility component as a promising technique for quantitative brain oxygenation assessment, and to systematically compare its performance and numerical characteristics with those obtained via conventional QSM.

To compare OEF values between SVD patients and HC, and to validate the potential of OEF as a discriminative biomarker for SVD.