Explainable deep learning and biomechanical modeling for TMJ disorder morphological risk factors
Shuchun Sun, Pei Xu, Nathan Buchweitz, Cherice N. Hill, Farhad Ahmadi, Marshall B. Wilson, Angela Mei, Xin She, Benedikt Sagl, Elizabeth H. Slate, Janice S. Lee, Yongren Wu, Hai Yao
Shuchun Sun, Pei Xu, Nathan Buchweitz, Cherice N. Hill, Farhad Ahmadi, Marshall B. Wilson, Angela Mei, Xin She, Benedikt Sagl, Elizabeth H. Slate, Janice S. Lee, Yongren Wu, Hai Yao
View: Text | PDF
Research Article Bone biology Metabolism

Explainable deep learning and biomechanical modeling for TMJ disorder morphological risk factors

Abstract

Clarifying multifactorial musculoskeletal disorder etiologies supports risk analysis, development of targeted prevention, and treatment modalities. Deep learning enables comprehensive risk factor identification through systematic analyses of disease data sets but does not provide sufficient context for mechanistic understanding, limiting clinical applicability for etiological investigations. Conversely, multiscale biomechanical modeling can evaluate mechanistic etiology within the relevant biomechanical and physiological context. We propose a hybrid approach combining 3D explainable deep learning and multiscale biomechanical modeling; we applied this approach to investigate temporomandibular joint (TMJ) disorder etiology by systematically identifying risk factors and elucidating mechanistic relationships between risk factors and TMJ biomechanics and mechanobiology. Our 3D convolutional neural network recognized TMJ disorder patients through participant-specific morphological features in condylar, ramus, and chin. Driven by deep learning model outputs, biomechanical modeling revealed that small mandibular size and flat condylar shape were associated with increased TMJ disorder risk through increased joint force, decreased tissue nutrient availability and cell ATP production, and increased TMJ disc strain energy density. Combining explainable deep learning and multiscale biomechanical modeling addresses the “mechanism unknown” limitation undermining translational confidence in clinical applications of deep learning and increases methodological accessibility for smaller clinical data sets by providing the crucial biomechanical context.

Authors

Shuchun Sun, Pei Xu, Nathan Buchweitz, Cherice N. Hill, Farhad Ahmadi, Marshall B. Wilson, Angela Mei, Xin She, Benedikt Sagl, Elizabeth H. Slate, Janice S. Lee, Yongren Wu, Hai Yao

×

Figure 6

Exploration of the TMJ disc’s biological responses to mechanical loading and its relationship with mandibular size with solute diffusion and energy metabolism model.

Options: View larger image (or click on image) Download as PowerPoint
Exploration of the TMJ disc’s biological responses to mechanical loading...
(A) Implementation of solute diffusion and energy metabolism models to assess the impact of mandibular size on TMJ disc oxygen and glucose availability, lactate accumulation, and ATP production. Simulation of TMJ disc compression during static biting reveals the nutrient environment’s response to localized solute exchange reduction in the contact area and mechanical strain–dependent solute diffusion in the disc’s loading volume. Solute movements are dictated by Fick’s second law, with oxygen, glucose consumption rates, and ATP production determined by the stoichiometry of intracellular energy metabolic reactions. (B) Analysis indicates that participants with smaller mandibles and larger joint forces experience compromised TMJ disc nutrient availability, increased lactate accumulation, and reduced ATP production (n = 16, 8 males and 8 females). 3D mandibular length exhibits the strongest correlation with these mechanobiological indicators. Oxygen versus mandibular length (11 N: P = 0.052, R² = 0.2441; 30 N: P = 0.002, R² = 0.4991; 60 N: P = 0.001, R² = 0.5906). Glucose versus mandibular length (11 N: P = 0.004, R² = 0.4507; 30 N: P = 0.002, R² = 0.5246; 60 N: P = 0.006, R² = 0.4347). Lactate versus mandibular length (11 N: P = 0.004, R² = 0.4628; 30 N: P = 0.001, R²=0.5753; 60 N: P = 0.004, R² = 0.4520). ATP versus mandibular length (11 N: P = 0.029, R² = 0.2960; 30 N: P = 0.002, R² = 0.5042; 60 N: P = 0.002, R² = 0.4951). Solid lines represent curve fittings where the differences are statistically significant (P < 0.05), and dashed lines represent curve fittings where the differences are not statistically significant (P ≥ 0.05). Elevated joint forces can lead to an expanded TMJ disc–condyle contact area and heightened mechanical strain, impeding nutrient transport and metabolic waste removal within the TMJ disc (Supplemental Figure 7). This model offers a mechanistic explanation for the increased susceptibility to degenerative changes in individuals with smaller mandibles.