In vivo biomechanics

In collaboration with neighboring labs we perform studies with human participants to better understand soft-tissue and full body biomechanics with the goal of improving clinical outcomes. 

Measuring in vivo strain is challenging. Previously, researchers have measured strain in vivo by combining medical imaging with classical image texture correlation algorithms, e.g., direct deformation estimation (DDE) and digital image correlation (DIC). However, these techniques are often limited by low signal-to-noise ratios, poor image texture, and other image artifacts. Thus, our lab has developed a deep convolutional neural network, StrainNet, to automatically measure patient-specific in vivo strains from high-frequency ultrasound images (Fig. A). Applying StrainNet to ultrasound images of human flexor tendons (Fig. B), we have been able automatically to segment the tendon as well as accurately capture strain distributions (Fig. C) by passing image sequences through a deep convolutional neural network (Fig. D). Currently, StrainNet has demonstrated a greatly improved predictive power with a mean strain error 90% lower than DDE and DIC (Fig. E). We believe StrainNet will generalize to other biological domains beyond soft tissues as we aim to uncover in vivo tissue behavior under loading, in order to better understand tissue injury and rehabilitation mechanisms.

Patients with chronic orthopedic conditions such as lower back pain (LBP) often experience complex and diverse movement and force patterns that are difficult to quantify. In collaboration with the UCSF Digital Orthopaedic Biomechanics Lab our lab has developed an algorithm that computes a kinematic profile (K-profile) and a kinematic score (K-score) that provides an understanding of full-body motion quality through calculating a patient’s deviation from an average control populations’ full body movement trajectory. These metrics offer a clinical solution to understanding the motion quality of patient populations which can be useful in understanding post-treatment (surgical or rehabilitation) biomechanical improvements. Our lab is now working to further understand the specific biomechanical strategies adopted by orthopedic patients such as LBP patients through time-series analysis of movement and force profiles.The ultimate goal is to develop algorithms to better understand, diagnose and predict orthopedic pathologies using time-series biomechanics data.

If you are interested in applying machine learning to medical images, please feel free to contact Blythe Dumerer (blythedumerer at berkeley dot edu).