Research
Digital twins to patient-specifically optimize breast cancer response to chemotherapy
Digital twins employ mathematical and computational models to virtually represent a physical object (e.g., planes and human organs), predict the behavior of the object and enable decision-making to optimize the future behavior of the object. While digital twins have been widely used in engineering for decades, their applications to oncology are only just emerging. Due to advances in experimental techniques quantitatively characterizing cancer, as well as advances in the mathematical and computational sciences, the notion of building and applying digital twins to understand tumor dynamics and personalize the care of cancer patients has been increasingly appreciated.
In this project, we employed digital twins (i.e., mathematical models that provide virtual representation of individual patients and predict the changes at future time points) to systematically evaluate individual triple-negative breast cancer patients' responses to different neoadjuvant chemotherapy regimens, thereby patient-specifically optimizing treatments. The digital twins were established by integrating the longitudinal MRIs with the mechanism-based model. With parameters personalized, we simulated each individual patient's response to 128 clinically reasonable combinations of A/C/Taxol dosing schedules. The predicted response (pCR or non-pCR) from each alternative schedule was compared to the patient response from the actual treatment.
The results of our investigation indicate that without changing the total dose, shortening the duration of A/C/Taxol administration increased the treatment efficacy. The effectiveness of altering the schedules varied substantially in different patients. The ongoing efforts include:
- Predicting patient-specific response to various therapy types (beyond chemotherapy)
- Integrating multi-modality data (i.e., histopathological data, genetic sequencing from biopsy, patient demographic information, etc.) to improve accuracy in response prediction and treatment selection
- Accounting for toxicity induced by the treatments
Image-guided digital twins to predict breast cancer response to neoadjuvant therapy
Patients with locally advanced, triple-negative breast cancer (TNBC) typically receive neoadjuvant chemotherapy (NAT) to downstage the tumor and improve the outcome of subsequent breast conservation surgery. In this study, we integrated quantitative magnetic resonance imaging (MRI) data with biology-based mathematical modeling to address the currently unmet need for accurate predictions of TNBC response to NAT on an individual patient basis.
A biology-based mathematical model was established based on the reaction-diffusion equation to characterize the mobility of tumor cells, tumor proliferation and treatment-induced cell death. Pre- and mid-treatment images of the individual patient were used for model calibration on a patient-specific basis; thus, the methodology represents a significant step away from population-based predictions and towards individual-based predictions. The personalized model accurately predicted the spatiotemporal response of TNBC to NAT and achieved high accuracy and specificity for predicting the final pathological status for each individual patient. Ongoing efforts to extend and refine this project include:
- Integrating the mechanism-based model with deep learning approaches to enable pre-treatment prediction
- Uncertainty quantification for the model prediction
- Speeding up the computational simulation with reduced-order models
- Adapting the image-guided modeling to cancer in other sites
Automatic longitudinal mammography analysis for large breast cancer screening cohort
The task of image interpretation by radiologists involves two concomitant processes: the evaluation of morphologic features of a finding as well as the estimation of its temporal behavior (changes in size, number, shape, density, margin characteristics, etc.). The ability to detect temporal change in breast tissue and suspicious lesions on mammography has a potential to significantly improve the specificity of breast cancer detection and is therefore of great clinical value. However, the existing commercial and research computer-assisted detection (CAD) techniques (both traditional radiomic and deep learning (DL)-based) focus on interpretation of screening mammography at a single time point and cannot efficiently track temporal changes. This lowers the specificity of these tools and limits their clinical utility.
To address this unmet need, we aim to develop computational approaches for longitudinal mammography registration, interpretation and automatic image labeling using text-based radiology reports to enable efficient and comprehensive imaging data analysis. Modern clinical and translational research in the context of precision medicine requires analysis of large volumes of imaging data. Within this project, we will develop computational tools to maximize the utilization of existing clinical data for retrospective research purposes and decrease the manpower needed for image labeling. Furthermore, the image processing pipeline developed in this proposed study will be subsequently used for integrated imaging and blood biomarker analysis in the MERIT cohort (n > 7,300), to enable further breast cancer prevention and early detection research.
This project is currently ongoing on and is funded by the Joint Center for Computational Oncology in the 2023-2024 Pilot Project Program.
Patient-specific optimization of nanoparticle convection-enhanced delivery in rGBM
Glioblastoma multiforme (GBM) is the most common and deadliest of all primary brain cancers. One promising treatment strategy for patients with recurrent GBM is convection-enhanced delivery (CED) of Rhenium-186 (186Re)-nanoliposomes (RNL) to provide delivery of large, localized doses of radiation. The success of treatment by CED relies on proper catheter placement for therapy delivery to maximize tumor coverage and minimize the leakage to healthy tissue. In this project, we developed an image-guided physics-based model to optimize catheter placement for RNL delivery on a patient-specific basis.
The mathematical model consists of:
- The steady-state flow field generated via the catheter infusion and the Darcy flow through the 3D brain domain,
- The transport of RNL governed by an advection-diffusion equation, and
- The point-spread function to transform the RNL distribution into the SPECT signal.
Pre-delivery MRIs were used to assign patient-specific tissue geometries. Two scenarios were performed to personalize the model parameters: a) patients-specific calibration with longitudinal SPECT images monitoring RNL distributions, and b) population-based assignment with the leave-one-out cross-validation (LOOCV). Furthermore, in each patient, we used the image-guided model to simulate RNL distributions for all possible locations of catheter tip(s), so to minimize the ratio of the cumulative dose of RNL outside the tumor to that within the tumor. The results indicate that our image-guided model, with either patient-specific calibrated parameters or LOOCV assigned parameters, achieved high accuracy for predicting RNL distributions up to 24 hours after the RNL delivery. The placement of catheter(s) optimized via our modeling substantially reduced the off-target ratio of RNL delivery. These results proved the potential of our image-guided modeling to guide patient-specific optimization of catheter placement for convection-enhanced delivery of radio-labeled liposomes.
Quantitative MRI to characterize tumor microenvironment, vasculature, and blood supply
Dynamic contrast enhanced MRI (DCE-MRI) has been playing an essential role in the diagnosis, staging and prognosis of breast cancer, especially highlighted with its high sensitivity to detect suspicious lesions. However, one concern of DCE-MRI in the diagnosis of breast cancer is its moderate specificity (i.e., relatively high false-positive rate) in reported clinical studies. Thus, it is of great importance to develop new MRI acquisition and analysis methods that can improve the specificity for cancer diagnosis.
In this project, we used both high-spatial resolution and high-temporal resolution (ultra-fast) DCE-MRI to characterize tumor microenvironment features associated with vascular architecture, blood supply and interstitial transport, which are known to be hallmarks of cancer. We developed a novel image analysis framework to extract morphological and functional information of tumor‐associated vessels from the DCE-MRI. Moreover, we established an image-guided computational fluid dynamic model system to estimate blood and interstitial flow fields, as well as drug/nutrition supply dynamics associated with breast tumors. Both techniques provided new metrics, improving the accuracy and specificity of differentiating breast cancer from benign breast lesions. The results indicate that quantitative imaging characterization of morphological and functional features of breast vasculature and fluid transportation has the potential to improve breast cancer diagnosis.
in silico MRI validation framework
Quantitative evaluation of an image processing method to perform as designed is central to both its utility and its ability to guide the data acquisition process. Unfortunately, these tasks can be quite challenging due to the difficulty of experimentally obtaining the “ground truth” data to which the output of a given processing method must be compared. One way to address this issue is via “digital phantoms,” which are numerical models that provide known biophysical properties of a particular object of interest.
In this project, we established an in silico validation framework for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) acquisition and analysis methods that employs a novel dynamic digital phantom. The phantom, R2D2, provides a spatiotemporally-resolved representation of blood-interstitial flow and contrast agent delivery, where the former is solved by a 1D-3D coupled computational fluid dynamic system, and the latter described by an advection-diffusion equation. Furthermore, we establish a virtual simulator which takes as input the digital phantom, and produces realistic DCE-MRI data with controllable acquisition parameters.
We used this developed framework to assess the performance of the standard-of-care high-spatial resolution acquisition, as well as the ultra-fast acquisition. The results indicate that our in silico framework can generate virtual MR images that capture effects of acquisition parameters on the ability of generated images to capture morphological or pharmacokinetic features. This validation framework is not only useful for investigations of perfusion-based MRI techniques, but also for the systematic evaluation and optimization new MRI acquisition, reconstruction and image processing techniques.