Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

Bioinformatics, Computational and Systems Biology

Bioinformatics, Computational and Systems Biology - Poster Session D

Poster A9 - Predicting the response of undifferentiated pleomorphic sarcoma to neoadjuvant therapy with image-based modeling

Friday, October 25, 2024
3:30 PM - 4:30 PM EST
Location: Exhibit Hall E, F & G

Introduction:  Undifferentiated pleomorphic sarcoma (UPS) is an aggressive subtype of a heterogeneous group of soft tissue tumors that arise from mesenchymal tissues throughout the body. UPS is the most common soft tissue sarcoma in adults and exhibits aggressive biological properties. Localized disease is typically treated with neoadjuvant therapy (NAT; chemotherapy and/or radiotherapy) followed by surgical resection. Determination of therapeutic efficacy currently requires histological analysis of the surgically resected mass—after all NAT has been delivered—thus, there is a need to predict efficacy for individual patients prior to the conclusion of NAT. To this end, we propose combining biology-based mathematical models with advanced medical imaging data, especially with multi-parametric magnetic resonance imaging (mpMRI), which characterizes UPS tissue properties before, during, and after NAT. In this abstract, we assess the predictive power of image-based mathematical models to predict the tumor viability at the end of NAT.

Materials and

Methods:  mpMRI data consisting of diffusion-weighted, susceptibility-weighted, and perfusion-weighted MRI was collected in twenty patients diagnosed with UPS. Each patient was imaged at least twice before, during, or after NAT. The patients were treated at the MD Anderson Cancer Center with either radiotherapy, chemotherapy, or sequential chemotherapy and radiotherapy. Total tumor cell count was estimated at each imaging visit via diffusion-weighted MRI. We then constructed an ordinary differential equation model describing tumor proliferation and death due to chemotherapy (ifosfamide and doxorubicin) and fractionated radiotherapy. For each chemotherapy agent, literature values of the drug half-life were used to model clearance of the drug in tissue, while the linear quadratic model was used to model the efficacy of radiotherapy. We then used a 5-fold cross validation process where patients were randomly assigned into training (80%) and testing (20%) sets. Using the training data, we calibrated model parameters for proliferation, and death rate due to ifosamide/doxorubicin therapy, all of which were employed to predict total tumor cellularity (TTC) at the end of chemotherapy and at the end of radiotherapy for patients in the testing set. The correlation between the predicted and measured TTC was assessed using the Pearson correlation coefficient (PCC) and the concordance correlation coefficient (CCC).

Results, Conclusions, and Discussions: Figure 1 shows the results of the predicted outcomes due to chemotherapy (left panel) and radiotherapy (right panel). We observed a strong correlation (PCC = 0.88) and agreement (CCC =0.71) between predicted and measured TTC at the conclusion of chemotherapy. Likewise, we observed a strong correlation (PCC = 0.92) and agreement (CCC =0.91) between predicted and measured TTC at the conclusion of radiotherapy. The average and standard deviations for the model parameters were: tumor proliferation rate 0.046+/- 0.030 day-1, radiosensitivity parameter 0.030 +/- 0.020 Gy-1, and combined chemotherapy death rate 0.091 +/- 0.010 day-1.

Discussion: A 5-fold cross validation approach was used to determine cohort averaged model parameters which were then used to forecast response. Our image-based modeling approach which leverages diffusion-weighted MRI collected before and after NAT was able to achieve accurate predictions of patient’s TTC. Future efforts will investigate the incorporation of more mpMRI data to provide a more complete picture of the intra-tumor biology.

Conclusion: Integration of mpMRI with biology-based models can enable predictions of response to NAT for patients with UPS.

Acknowledgements (Optional):  This work is supported as a pilot project sponsored by The Oden Institute for Computational Engineering and Sciences, The University of Texas MD Anderson Cancer Center, and Texas Advanced Computing Center.