Physical intervention and data-driven biomechanics in breast cancer bone metastases
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Breast cancer patients are at a high risk of bone fracture due to aging, adverse effects such as hormone change and fatigue associated with cancer and cancer treatment, and, more importantly, cancer invasion (metastasis) into bone. The beneficial effects of physical activity and exercise on cancer patients include prevention of cancer occurrence and recurrence and improvement of locomotion and bone health, as demonstrated in considerable research and clinical trials. The aims of my thesis are (1) to understand the effects of different exercise modes on preserving the integrity of bone with metastatic breast cancer and (2) to provide new insight into the assessment of fracture risks based on images and machine learning. ☐ Two studies are included in this thesis. In the first study, three physical stimuli – namely moderate or intensive compressive tibial loading, and aerobic treadmill running – were applied to animals bearing bone metastases. A U-shaped load-magnitude-dependent development of bone lesion was found in this study. Moderate loading was found to suppress tumor induced osteolysis while intensive loading exacerbated the osteolytic lesions in bone. For the first time, this work demonstrated an inhibitory effect of five-week aerobic treadmill running on osteolytic bone metastases in immunocompetent mice bearing breast cancer in tibiae. Further cellular investigations revealed that mechanical stimuli altered the endpoint of bone metastases through, at least partially, suppression of the HIF signaling and osteocyte apoptosis. ☐ The second study of my thesis aims to develop a machine learning-based method for fracture assessment of metastatic bone disease utilizing the 260 longitudinal sequences of µCT scans acquired in the first study. A deep learning model was developed to forecast the bone structure from µCT images taken at previous time points. The stiffness of the predicted bone structure was estimated using Finite Element (FE) method. Although this work used only µCT images of murine bone metastases due to the lack of such human data, this is the first study to show that spatial-temporal neural network can be used to improve conventional and patient specific fracture assessment. ☐ Overall, my thesis has filled several gaps in bone metastasis research. The protective effects of aerobic treadmill running on bone metastases in skeleton-mature female mice was investigated for the first time. A large amount of in vivo µCT data was collected, enabling the development of a predictive model for fracture assessment based on deep learning and FE analysis. Even though all experiments were performed in mice, these preclinical findings suggest clinical potentials of physical interventions and data-driven fracture assessment in battling metastatic bone diseases.
Description
Keywords
Bone metastases, Breast cancer, Deep learning, Finite Element, Osteolytic lesions