Quant Imaging Med Surg. 2026 Sep 1;16(9):727. doi: 10.21037/qims-2026-0675. Epub 2026 Aug 3.
ABSTRACT
BACKGROUND: Peripheral arterial disease (PAD) is mainly defined by macrovascular obstruction, while microvascular dysfunction remains insufficiently characterized by conventional diagnostics. Arterial spin labeling (ASL) magnetic resonance imaging (MRI) enables non-invasive, contrast-free quantification of skeletal muscle blood flow (SMBF) in the calf, with potential applications in PAD diagnosis, risk stratification, and perioperative assessment of microcirculation. Deep learning enabled reconstruction of dynamic ASL data offers potential to generate temporally and spatially resolved high-quality perfusion maps. The purpose of this study was to determine whether this dynamic ASL could distinguish PAD from healthy controls (HCs), both on spatial and temporal domains.
METHODS: Dynamic SMBF maps of the calf were acquired with 15-sec temporal resolution before, during, and after a 3-minute isometric plantar flexion exercise (hyperemia) using a MR-compatible ergometer in 10 patients with mild to severe PAD and 10 age- and sex-matched healthy volunteers. PAD was confirmed by significantly lower ankle-brachial index values compared with HCs (0.60±0.21 vs. 1.25±0.1, P<0.0001). A convolutional neural network was trained on synthetic ASL data and applied to the dynamic series to reconstruct spatiotemporal SMBF maps with full-sampling and under-sampling schemes. Regions of interest were defined in five calf muscles (medial gastrocnemius, soleus, lateral gastrocnemius, tibialis anterior, and lateral compartment) and perfusion was quantified on the SMBF maps. Linear mixed-effects modeling evaluated SMBF across time, muscle, and cohort.
RESULTS: Peak exercise perfusion did not differ significantly between two cohorts across muscle groups (all P>0.05), except in the medial gastrocnemius when normalized for baseline perfusion [PAD: 54.0%±41.4 % vs. HC: 83.4% (66.2%, 128.4%), P=0.05]. In contrast, the hyperemic recovery time constant (τ), a time-dependent kinetic parameter describing post-exercise perfusion return towards baseline, was significantly prolonged in patients with PAD across all muscle groups (P<0.05), except the tibial compartment (P=0.094). Linear mixed-effects analysis demonstrated higher resting SMBF in patients with PAD compared with controls (difference +8.3 mL/min/100 g, P=0.011), but revealed a significantly attenuated exercise-induced perfusion response, reflected by a robust time × group × muscle interaction (P<0.0001). Deep learning reconstruction of under-sampled ASL data produced SMBF maps with relatively low bias compared to fully sampled references (Bland-Altman: -1.2 to -2.0 mL/min/100 g across conditions), potentially improving temporal resolution by 100%.
CONCLUSIONS: Deep learning-enhanced, temporally resolved ASL MRI identifies attenuated exercise-induced perfusion and delayed recovery after exercise in patients with PAD that are not evident using static peak perfusion metrics alone. Active hyperemia recovery kinetics provide a sensitive marker of microvascular dysfunction, while deep learning reconstruction enables high temporal resolution imaging within clinically practical scan times.
PMID:42701578 | PMC:PMC13545620 | DOI:10.21037/qims-2026-0675