Real-time motion-enabling positron emission tomography of the brain of upright ambulatory humans

The study addresses the need for motion-tolerant, upright neuroimaging that can capture activity in both cortical and deep brain structures during naturalistic behaviors. Existing modalities either require stillness and supine positioning (fMRI, standard PET/SPECT) or allow motion but lack deep brain sensitivity (EEG, NIRS/HD-DOT, OPM-MEG). Enabling upright, motion-compatible PET would allow quantitative assessment of metabolism and neurotransmission during real-world tasks. The authors developed a lightweight, head-worn PET prototype (AMPET) and aimed to test feasibility using an ecologically valid motor task. Three validation goals were specified: (1) demonstrate minimal motion-induced artifact during robust head/body motion (walking-in-place) versus standing-at-rest, (2) show task-related activation in a priori motor cortical regions (leg primary motor cortex and SMA) greater than non-motor regions, and (3) detect activity from deep brain structures (e.g., thalamus, basal nuclei) normally inaccessible to surface-limited modalities. The broader purpose is to establish feasibility for future ambulatory PET systems enabling studies in populations and tasks traditionally excluded due to motion constraints.

The article reviews limitations of current neuroimaging methods: fMRI and standard PET/SPECT require stillness and supine positioning and dedicated rooms, limiting ecologically valid tasks and excluding many patient groups (e.g., PD, dementia, ASD, epilepsy). Wearable modalities (EEG, NIRS, HD-DOT) are motion compatible but have shallow penetration and limited deep brain sensitivity (thalamus, basal nuclei, hippocampus). OPM-MEG allows modest movement in shielded rooms but cannot accommodate large movements like walking and has reduced deep sensitivity. PET offers unique quantitative measurement of metabolism and neurotransmitters; prior solutions include repeated H2[15O] injections or delayed-PET paradigms with FDG or receptor ligands during out-of-scanner tasks, but these have logistical/radiation burdens and lack same-session multi-task capability. Advances in SiPM detectors enable lighter head-dedicated PET systems, though motion robustness has been limited. Prior delayed-PET and fMRI studies of locomotion inform the a priori ROIs (leg M1, SMA) used here. The authors position AMPET as addressing the gap: upright, motion-tolerant PET with potential deep structure coverage for real-time or near-real-time behavioral imaging.

Ethics and participants: Approved by West Virginia University IRB; observational study using consenting adult patients scheduled for same-day clinical PET for non-small cell lung cancer. Inclusion: age >18, able to walk in place up to 10 min unaided. Exclusion: upper grip/spine/head/neck pain or mobility issues. Convenience sample intended n=10–12; demographics reported (mean age ~53 years; BMI mean 26.8; height mean 161.3 cm; weight mean 76.1 kg). One under-dosed participant excluded; for task analyses two excluded due to AMPET placement below leg motor ROI coverage. Final samples: n=10 for motion tolerance (Validation #1), n=8 for walking-in-place task activation (Validation #2), and n=4 (subset of n=10) for deep brain analysis (Validation #3). One participant had a right leg amputation with prosthesis and completed the task. AMPET hardware: Head-worn PET ring with 12 detector modules (each 48×48 mm active area) using Hamamatsu SiPM (MPPC) arrays coupled to 1.5×1.5×10 mm LYSO crystal arrays. Axial FOV 21 cm; spatial resolution at center: 2.0 mm tangential, 2.8 mm radial FWHM; weight 3 kg. Helmet supported from above via flexible cord/bungee to offload weight and permit natural head motion typical of walking. System upgrades included in-module amplifiers, temperature monitoring, and remote bias adjustment; module temperatures monitored and corrected for energy acceptance. Experimental protocol: AMPET fitted by adjustable fixation; positioned to include superior cortex (leg motor area). Immediately before tasks, ~1 mSv pre-dose of [18F]-FDG bolus administered (within total clinical dose ~10 mSv, ~370 MBq). Imaging started immediately; total session ~25 min. Initial 6 min post-injection: blocked design alternating 30 s standing-at-rest and 30 s walking-in-place. After initial period, minutes 11–16 involved either (a) seated-walking vs seated-rest with helmet positioned lower to include deep structures (subset), or (b) re-imaging leg-M1 coverage (subset) for cross-validation. Clinical scans and registration: Same-day clinical Siemens PET/CT brain scans used for registration and to assess resting activity patterns. AMPET and clinical PET images processed to DICOM; edge slices removed; 3×3 smoothing applied. Clinical PET aligned to CT via MIM BrainAlign; AMPET co-registered to PET/CT with automated then manual adjustments as needed (MIMneuro/MIMfusion). ROI selection: For motion artifact analysis, whole-brain outline traced on a middle axial slice and compared to entire FOV to quantify spillover. For task-related analyses, bilateral spherical ROIs (25 mm radius) centered on a priori coordinates for leg/foot primary motor cortex (M1) derived from prior walking PET/fMRI literature; SMA ROI and other cortical ROIs (precentral gyrus lateral M1, postcentral gyrus S1, precuneus, frontal lobe control) defined from MIM template and healthy control database, limited to AMPET coverage. Normalization: Relative uptake values computed as ratios to a lateral frontal cortex reference ROI present in all subjects and presumed uninvolved in the motor task, enabling comparison across conditions and with clinical resting scans. Motion tolerance analysis: In the initial 6 min window, twelve 30 s images (alternating rest/walk) generated. For each participant, ratio of whole-brain activity to entire FOV counts computed across intervals; variability compared between walking-in-place and standing-at-rest using a mixed-effects model accounting for within-subject correlation. Phantom-based calibration related ratio shifts to estimated misalignment distances, assuming changes due to motion rather than uptake dynamics or scatter. Statistics: Normality assessed with Shapiro–Wilk; independent two-tailed t-tests for ROI comparisons; significance threshold p<0.05. Mixed-effects model for motion artifact test. Activity quantified as mean intensity (sum voxel value/ROI volume). Supplementary tables contain exact p-values. Additional subsets: Deep brain subset (n=4 with adequate coverage after helmet repositioning) analyzed for caudate, putamen, thalamus versus cortical ROIs during minutes 11–16 seated walking/rest blocks. Leg-M1 re-imaged subset (n=5) assessed for stability of activation relative to frontal reference in the later time window.

- Motion tolerance (n=10): No visible motion artifacts (blurred edges/ghosting) detected. Mixed-effects analysis showed no significant difference in the whole-brain-to-FOV activity ratio between walking-in-place and standing-at-rest blocks (p=0.25). Estimated AMPET-to-brain misalignment during walking was ±1.3 mm if variance attributed to motion. - Task-related cortical activation (n=8): Bilateral leg-M1 ROI activity significantly elevated versus all other examined ROIs during the walking-in-place period (relative to frontal reference). Leg-M1 activity was ~1.75 times the mean activity of the brain within AMPET FOV (p=0.000006). Leg-M1 was significantly greater than SMA, premotor (precentral lateral), postcentral gyrus, precuneus, and frontal lobe ROIs (all p<0.001). SMA showed significantly greater activity than non-motor cortical ROIs (frontal, postcentral, precentral lateral) but less than leg-M1 (p=0.00019). Clinical PET resting scans showed no inherent leg-M1 elevation versus other ROIs, supporting task specificity of AMPET findings. - Amputee case (n=1): Participant with right leg amputation using a prosthesis exhibited greater right-hemisphere leg-M1 activation (representing intact left leg) compared with left, consistent with lateralized motor representation; in two-legged participants (n=7), left vs right leg-M1 activity did not differ (p=0.37). - Deep brain subset (n=4 with adequate coverage): Observed activity in basal nuclei and thalamus; trends for higher caudate activity versus lateral temporal cortex (p=0.051) and versus inferior frontal ROI (p=0.066). - Later window cross-validation (n=5): At 11–16 min post-injection, bilateral leg-M1 activation remained significantly greater than the frontal reference ROI (p=0.0008), indicating stable detection across time windows. Overall, the prototype tolerated robust upright motion while capturing expected locomotor-related cortical activation and preliminary deep structure signals.

Findings demonstrate feasibility of upright, motion-tolerant PET imaging using a lightweight head-mounted system during naturalistic locomotor behavior. The AMPET prototype maintained brain–imager alignment within a couple millimeters during walking-in-place, with no significant motion-induced artifact in quantitative measures. Functionally, AMPET detected robust activation in leg primary motor cortex and SMA consistent with prior locomotion literature, and showed preliminary evidence for deep structure engagement (basal nuclei, thalamus) in a subset. The amputee case highlighted expected unilateral dominance in the hemisphere controlling the intact leg, illustrating the potential to probe neuroplasticity and compensatory control in special populations. Compared to surface modalities (NIRS/HD-DOT, OPM-MEG), AMPET offers deeper sensitivity while allowing upright movement; unlike fMRI, it accommodates large movements such as walking and real-world tasks. These results support the potential of ambulatory PET for ecologically valid neuroscience and clinical research, including populations who cannot remain still (e.g., movement disorders, pediatrics) and tasks requiring upright posture and mobility. The study also informs priorities for system development: improved positioning (e.g., optical/IR tracking), increased axial/angular coverage to expand FOV, motion tracking/correction, and support mechanisms (e.g., gimbals/robotics) to accommodate multi-axis head motion. Methodological enhancements (e.g., infusion protocols, arterial/blood input functions) could enable higher temporal resolution functional PET and quantitative modeling during dynamic tasks.

The study provides proof-of-concept that a wearable, motion-tolerant PET helmet (AMPET) can perform upright brain imaging during ambulatory tasks, maintaining alignment and capturing expected task-related activation in motor cortices, with preliminary detection of deep brain activity. This extends PET’s applicability to ecologically valid paradigms and populations typically excluded due to motion constraints. Future work should focus on integrating precise optical positioning and motion tracking, expanding the detector geometry for whole-brain coverage, refining support systems to accommodate full head kinematics, and adopting quantitative protocols (e.g., infusion, arterial sampling or image-derived input functions) to enable multi-task, same-session functional PET with deep structure sensitivity. Larger, dedicated cohorts (including clinical populations and amputees), improved placement guidance, and longitudinal designs will help establish reliability, generalizability, and clinical utility.

- Small convenience sample with exclusions due to under-dosing and suboptimal helmet placement; deep brain analyses limited to n=4 with adequate coverage. - Single bolus FDG delivery without arterial blood sampling limited temporal resolution; alternating 30 s blocks could not be separated for differential analysis within the initial 6 min due to tracer sequestration dynamics. - Limited axial FOV and manual placement led to inconsistent coverage of target ROIs (leg-M1 and deep structures) across participants. - Motion analysis assumes ratio changes primarily reflect motion rather than uptake dynamics or scatter; potential confounds from changing physiology during walking. - No separate AMPET resting-state baseline; resting comparisons relied on clinical PET/CT performed later. - Support system with single-point suspension may limit tolerance to certain head motions (e.g., nodding/yaw), though adequate for walking-in-place. - Conducted at low dose within clinical constraints; not a fully optimized neuro-PET protocol; generalizability requires larger, dedicated studies.