Motion-Compensated µMRI of Tethered Insect Locomotion

Magnetic resonance imaging (MRI) provides a powerful, non-invasive means to probe biological structure and function across scales, offering unparalleled versatility and making it a promising modality for studying small, freely behaving model organisms. However, living systems are inherently dynamic, and motion presents a persistent challenge for MRI,
mainly due to its relatively slow, sequential data acquisition. In clinical and preclinical imaging, well-established strategies exist to address the effects of physiological motion and small involuntary movements on image quality. These approaches, however, do not address large-scale behavioral motion. While such motion is rarely considered in clinical imaging, studies of model organisms often aim to observe biological processes in vivo, where natural
movement is an integral part of the phenomena under investigation.

Micro-magnetic resonance imaging (µMRI), the high-resolution adaptation of MRI for microscopic scales, offers a unique window into the internal morphology of small model organisms and represents an ideal modality for such investigations. Realizing this potential requires new imaging frameworks that treat motion as an intrinsic part of the MRI measurement, allowing natural behavior while modeling and correcting its effects during acquisition and reconstruction. ... mehrInsects serve as excellent models for developing and validating these approaches. Towards this goal, this thesis explores the feasibility of MRI in small moving organisms and examines the limitations of implementing such a framework within a µMRI system. Initial findings motivate the development of an experimental platform designed to study live, active insects under minimally constrained conditions using an in situ treadmill, combined with methods for motion-compensated imaging during tethered locomotion.

The MR-compatible treadmill system enables tethered insect locomotion within a 15.2T ultra-high-field preclinical scanner. This configuration constrains motion largely within the coil field of view (FOV) while allowing near-natural walking behavior. In addition, an integrated MR-compatible computer vision system was developed, employing an optical flow algorithm for real-time behavioral motion tracking (mean tracking error of 2.04 ± 1.13 pixels), which provided a robust means to address motion-related challenges during imaging. A vision-based prospective gating approach was first introduced to compensate for the semiperiodic abdominal pitching during locomotion. This enabled snapshot imaging of a 2D sagittal section of the insect at 117 µm isotropic resolution, as well as an axial section through the thorax at 39 µm resolution, clearly delineating internal morphology during behavior. While this enabled motion-compensated imaging, temporal resolution remained limited and performance depended on behavioral variability of the insect. Retrospective gating improved flexibility and enabled the possibility of dynamic imaging, but residual artifacts persisted due to the inherently non-rigid nature of insect body motion and measurement uncertainties.

To address the limitations of gating-based motion compensation, a deep learning motion correction step was applied after gating as part of a combined approach, leading to greater robustness against motion artifacts and improved image quality. Achieving true dynamic imaging, however, requires faster acquisition. To this end, acceleration was achieved via undersampled Cartesian acquisition, with a deep learning–based framework employed to mitigate the resulting reconstruction artifacts. In the absence of parallel imaging hardware, which limited spatial acceleration, integrating a view-sharing strategy with deep learning proved more effective, leveraging temporal redundancy to improve initialization and restore image quality in dynamic 2D imaging. While the resulting temporal resolution of 1.2 s/frame at an acceleration factor of 2.5X remain modest, the proposed approach represents a promising direction for future non-invasive dynamic biomechanical studies of tethered insects.

These contributions provide a systematic framework for motion-inclusive µMRI, bridging the gap between high resolution structural imaging and the study of small organisms in active, near-naturalistic conditions.