J20 - Non-Invasive Estimation Of Tumor Stiffness and ECM Fiber Architecture Using an Inverse Modeling Approach

Introduction: : Cancer remains a leading cause of morbidity and mortality worldwide, with diagnosis often, particularly in soft tissues, relying on ultrasound (US) imaging. Recent advances have revealed that stiffening of tumors may be an early indicator of malignancy. Tumor progression has also been associated with a significant architectural reorganization of the collagen fibers within the extracellular matrix (ECM), which is thought to promote tumor cell migration, invasion, and resistance to therapies. Consequently, characterizing both the mechanical stiffness and the fiber architecture of the tumor microenvironment (TME) has become necessary for understanding cancer progression and, hence, developing novel therapeutic strategies. While conventional methods such as second-harmonic generation (SHG) microscopy, atomic force microscopy (AFM), and ex vivo mechanical testing have been employed to study these phenomena, they are invasive, costly, and not feasible for in vivo applications. In this study, we present a computational framework that leverages realistic synthetic ultrasound images derived from in-silico finite element (FE) simulations of tumor inclusions under compression to estimate regional tumor stiffness and ECM fiber orientation maps.

Materials and

Methods: : We developed a two-dimensional FE model via plane stress formulation. The tumor was modeled as an irregular structure embedded in healthy tissue, and the tumor was prescribed a stiffness larger than that of the healthy tissue. The tumor stiffness was expressed as a multiplier of that of the surrounding tissue. Separate inverse modeling approaches were used to estimate (i) tumor tissue stiffness and (ii) the fiber architecture of the ECM. To mimic the hyperelastic behavior of the living tissue, the neo-Hookean model was used in the inverse model to predict the stiffness. However, in the case of ECM fiber orientation determination, we used an anisotropic hyperelastic material model for the tissue surrounding the tumor region. The fiber orientation angle was controlled by a Gaussian distribution defined by the mean fiber angle and the standard deviation. As the standard deviation increases, more fibers become disaligned from the mean fiber angle. The tumor-tissue assembly was fixed in the axial direction at the bottom faces, and axial compression loads were applied to the top faces. The displacement results were used to produce the synthetic Ultrasound Image samples, from which the ground truth displacement map was measured via speckle tracking. The proposed inverse model estimates the mechanical properties of the tumor and healthy tissue via a least square optimization comparing the speckle tracking (ST) displacement and FEM results. The first inverse model predicts the tissue stiffness of the tumor region, and the second model estimates the deviation in the fiber angle distribution.

Results, Conclusions, and Discussions:: The inverse model in this study predicted the optimized parameters with good accuracy. While estimating the tumor tissue stiffness, for different compression levels, the average error was 4.04%. The most accurate case had an error of 1.85%, and the highest error (6.87%) was found at the case for 6x tumor stiffness and 100 Pa compression load (Figure 1A). In Figure 2B, the coefficient of determination (R2 score) of our inverse model was calculated. With an average R2 of at least 0.98 on each aspect ratio, our inverse model recapitulates the mechanical properties of the tumor tissue on par with the ground truth values. We found that fiber orientation estimation was significantly accurate under most conditions. This model showed minimal error (average 4.85%) across varying fiber alignment and orientation angles (Figure 1B). Overall, our findings suggest that, in conjunction with US imaging, the inverse finite element method has the potential to measure the mechanical properties of tumor tissue non-invasively.

The mechanical properties of the tumor microenvironment are recognized as one of the key indicators of malignancy. Our findings highlight the potential of inverse finite element modeling as a tool for characterizing the tumor tissue biomechanics. This approach suggests a promising pathway toward non-invasive diagnosis and improved assessment of tumor progression.

Acknowledgements and/or References (Optional)::