Cerebral small vessel disease includes a wide spectrum of cerebrovascular diseases that involve endothelial dysfunction of capillaries, small arteries and small veins in the brain, leading to blood-brain barrier (BBB) dysfunction, impaired vasodilation, vessel stiffening, dysfunctional blood flow and interstitial fluid drainage, WM rarefaction, ischemia, inflammation, myelin damage, and secondary neurodegeneration [17]. The underlying neurobiological mechanisms are, to date, not clear, and SVD is generally assessed by looking at macroscopic or mesoscopic manifestations, such as WM hyperintensities, lacunas, microbleeds, perivascular spaces, or small subcortical infarcts. In terms of hemodynamic measures, a reduction in CVR and CBF has been observed [18]. Given the microvascular nature of the disease, mapping microvascular reactivity and the possible related metabolic impairment can help clarify the underlying mechanisms and pathological effects of the observed alterations [19].

AD is characterized by amyloid-β accumulation and impaired CVR [20], but the relationship between these alterations is not clear. One hypothesis is that amyloid-β accumulation may directly induce vascular dysfunction by impairing vasorelaxation, but therapies aimed at amyloid-β plaque reduction are not effective in reversing cognitive decline [21].

Another hypothesis identifies the microvasculature and related inflammatory processes as fundamental in the pathogenesis of AD [22]. In particular, a CBF reduction is currently the earliest known change associated with the disease, and the mechanism behind it might be a pericyte-driven constriction of capillaries [23]. It is hypothesized that this constriction and the associated hypoperfusion are the precursors of amyloid-β accumulation via upregulation of the BACE1 enzyme. Capillary dysfunction and microvascular CVR reduction were additionally observed to be associated with symptom severity [24, 25], and resting CMRO₂ was found to be reduced in a calibrated BOLD study in AD patients [26].

Brain tumors are accompanied by strong alterations in the vasculature (e.g., angiogenesis), which often lead to a local disruption in the neurovascular coupling, also termed neurovascular uncoupling (NVU). Even low grade brain tumors show some degree of NVU [27]. In practice, this may cause a reduced or absent BOLD response to functional tasks in the region affected, even if neuronal activation is present [28]. Distinguishing regions of NVU from regions effectively lacking neuronal response is very important in presurgical planning to avoid excess surgical resection [27, 28]. In place of task-based BOLD fMRI, CVR mapping is generally preferred as it relies on a purely vascular stimulus. CVR mapping was found to be superior (i.e., able to identify NVU regions also in low and intermediate brain gliomas) to perfusion mapping with T2* dynamic susceptibility contrast [27]. Increasing spatial resolution in CVR maps and adding a vascular contrast (such as CBV-weighted) is important for precise localization and characterization of the tissue affected.

Mapping of CBV with MRI without the use of contrast agents can be achieved with vascular-space occupancy [3]. VASO is an inversion recovery technique exploiting the different T₁ of blood and tissue (T_1,tissue < T_1,blood) for creating a contrast sensitive to CBV changes. In the original VASO formulation the images were acquired at the time of blood nulling [3], while more recent implementations showed that different inversion times can be used, as long as a difference in the T₁ weighting of the two compartments is achieved [29, 30]. When the vasculature dilates, either following a metabolic demand of the tissue or a vasodilatory stimulus, the volume of blood in a responding voxel increases. VASO measures quantitative changes in blood volume, which can be expressed as a percentage change or in volume units, but VASO does not quantify baseline CBV.

The slice-saturation slab-inversion VASO (SS-SI-VASO) [31], developed at 7 T, permits the acquisition of gradient-recalled echo (GRE)-BOLD and VASO contrasts interleaved by adding a second excitation pulse at each repetition time (TR). VASO demonstrated a high specificity for the capillaries, arterioles and intracortical arteries [32, 33], making it a valuable tool for high-resolution fMRI [34]. Recently, the VASO implementation acquiring BOLD volumes concomitantly has been adapted for high-resolution applications at 3 T, where it has been shown capable of distinguishing layer-dependent activation [13], thus reflecting a microvascular sensitivity.

Using the VASO sequence requires some effort in setting up a working protocol, dealing with the lower SNR and longer TRs compared to BOLD, as well as a less standardized data processing. Examples of protocols for submillimetric acquisitions can be found at https://github.com/layerfMRI/Sequence_Github/tree/master/3T.

Cerebrovascular reactivity (CVR) reflects the capacity of brain vessels to dilate or constrict in response to a vasoactive stimulus. MRI-based CVR is generally expressed as percent BOLD or ASL signal change per mmHg change in end-tidal carbon dioxide (EtCO₂) [2]. Despite the relative simplicity of the measure, it is one of the most reliable MRI-derived predictors of cerebrovascular impairment and has been applied in the study of several diseases, including SVD and AD [2, 35, 36].

Since GRE-BOLD signal changes are heavily weighted towards the venous macrovasculature, both due to an inherent weighting towards larger vessels and to dHb-containing compartments [37], CVR maps based on such contrast show similar features. In order to target microvascular CVR, different sequences should be employed, such as ASL, spin-echo (SE)-BOLD, or VASO. In particular, VASO-based CVR could be interesting due to the high resolutions achievable and the observed stability to hypercapnic stimuli [14]. In one study, VASO reactivity was found to be more negative (i.e., larger CBV increase) in patients with carotid artery disease compared to healthy controls [38].

BOLD signal changes in a voxel during a task of interest can be modeled according to Eq. 1 [4]: ∆ BOLD BOLD 0 = M 1 − CBV v CBV v , 0 CBF CBF 0 − β CMRO 2 CMRO 2,0 β (1) where CBV_v is the CBV contributing to BOLD signal changes (“venous” CBV), M is the calibration parameter which corresponds to the maximum BOLD signal change (BOLD = M for complete deoxy-hemoglobin washout from the vasculature), β is a magnetic-field dependent exponent, and the subscript 0 indicates quantities at baseline.

It is generally assumed that CBV and CBF are coupled via Eq. 2: CBV v CBV v , 0 = CBF CBF 0 α (2) where α is a coupling constant, often taken to be equal to 0.38 [43].

The goal of calibrated fMRI is isolating the CMRO₂ term from the Eq. 1: to this end, a pure CBF or CBV sensitive contrast needs to be acquired in addition to BOLD, and the calibration parameter M needs to be estimated. For extracting M a vasoactive, isometabolic (CMRO₂/CMRO_2,0 ∼ 1) stimulus is used, which introduces a BOLD dependency only on CBF or CBV (measurable) if Eq. 2 is assumed [4, 5].

Most calibrated fMRI studies use ASL sequences for mapping CBF and extract CBV via Eq. 2. Since the technique allows for voxelwise mapping, the resolution achievable by ASL at 3 T dictates the resolution of CMRO₂ maps. For submillimetric applications, CBV mapping can be used and CBF changes estimated by Eq. 2, while assuming a different coupling exponent for venous and total CBV changes [14, 15].

Other approaches for CMRO₂ mapping have been introduced using different gas mixtures or without exogenous gas manipulations. Existing acquisition techniques and modelling approaches have been recently reviewed [44, 45].

The proposed multiparametric acquisition can offer a multi-faceted window into microvascular dysfunction, by mapping vascular and metabolic features concomitantly at 3 T and at voxel volumes below 1 mm³.

The submillimetric VASO sequence has been introduced only recently at 3 T and has some limitations. First of all, the low SNR of 10–20 [13] limits its flexibility: tasks associated with small effect sizes are hard to detect, unless the acquisition time is largely increased. Moreover, the generation of the VASO contrast has some timing constraints that impact the TR duration: the typical TR is about 4–5 s (for the acquisition of a full BOLD-VASO pair) for slab protocols.

Additionally, the breathing challenge could create excessive discomfort in patient populations, both due to wearing a mask and being in a hypercapnic state. Therefore, it is important to familiarize each subject with the gas challenge prior to scanning. An earlier study showed an increase in participant dropouts for AD patients compared to healthy control for the breathing challenge [26].

While CVR mapping is reliable and has been applied in disease, the same does not hold for calibrated fMRI. Calibrated BOLD models rely on several assumptions, which, even at typical image resolutions, might not be a good approximation of the real underlying mechanisms. Such limitations could be exacerbated at higher resolutions where the spatial heterogeneity of the vasculature is resolved. Most notably, the flow-volume coupling expressed by Grubb’s relationship is unlikely to hold true with the same exponent for small voxel sizes and between healthy subjects and patients. Therefore, differences in CMRO₂ between a healthy group and a disease group could stem from a disrupted CBF-CBV coupling rather than a true metabolic difference. To avoid such bias, imaging CBF, CBV, and BOLD concomitantly [8] is preferable, but CBF mapping via arterial spin labeling at high resolution is challenging [49], therefore studying the flow-volume coupling relationship experimentally in humans is difficult.

If using VASO in the calibrated BOLD framework, the CBV at baseline needs to be assumed. A fixed fraction of 5.5% [50] might be acceptable considering that the microvascular blood density, which is the major factor responsible for VASO signal changes for short stimulations, is relatively homogeneous across the cortical depth [10, 51]. During hypercapnic stimulations, however, the fraction of VASO signal change coming from venous blood increases [52], challenging the assumption of a fixed baseline blood fraction.

Finally, the assumption of isometabolism during gas administration has been challenged in several studies. Although there is not a definitive answer, some evidence suggests a reduction, rather than an increase, in CMRO₂ of up to 13% during a 5%-CO₂ gas challenge [53]. However, this value depends on the concentration of arterial CO₂ and on the length of the respiratory challenge, with shorter epochs having less impact on CMRO₂ [2].