Abstract Body (Do not enter title and authors here): Introduction: Microscale mechanical dysfunctions of the heart underlie hypertrophic cardiomyopathies, yet modeling local tissue mechanics remains prohibitively challenging without extensive, high-resolution data. This difficulty stems from the heart’s inherent heterogeneity, anisotropy, and complex microstructural interactions. Atomic Force Microscopy (AFM) is the gold standard for measuring local mechanical properties, but it is limited by slow data acquisition and shallow imaging depth. To generate the large datasets required for accurate mechanical models, we propose correlating second-harmonic generation (SHG) images with mechanical measurements from AFM. This correlation will enable fast, non-destructive inference of mechanical properties from SHG images. In this study, we demonstrate that features in large-field SHG images can be correlated with local AFM measurements, providing a basis for mechanical inference from SHG images.
Methods: Left ventricular tissue from porcine hearts was fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned at 15. A sample was imaged using SHG ( by stitching together tiles to create a full-field image. Regions were manually segmented as cardiomyocytes, wavy collagen fibers, or aligned collagen fibers based on morphology. AFM force maps (, spacing) with a indentation velocity and a scan rate were acquired using a spherical silicon tip (, , ) after calibrating deflection sensitivity and determining the spring constant via thermal tuning. Young’s modulus was calculated at each point of the force map using a standard Hertzian contact model. A brightfield microscope coaligned with the AFM probe was used for coarse registration to the SHG image by matching sample geometry, and fine alignment was achieved by manually overlapping microstructural features. AFM measurements were then categorized by the region type identified in the SHG image.
Conclusion: We demonstrate a process for correlating AFM measurements with SHG images that enables constitutive modeling of the heart’s complex heterogeneous and anisotropic microstructures. Ongoing work includes using classical image processing and deep learning techniques to automate this process, implementing super-resolution techniques to improve spatial correlation, and using these results to derive physics-informed constitutive models of cardiovascular microstructural mechanics.