Session V46: Physics in Medicine: Modeling, Imaging, and Treatment

Advances in Modeling, Imaging, and Treatment of Cancer

Medical physics is intimately connected with medicine, and is progressing along a similar path. General trend of medicine, particularly oncology, towards personalized treatment gave rise to precision medicine, which addresses the highly complex nature of disease. In the past, little could be done to tackle this complexity, but the emergence of targeted therapies is bringing personalized therapies within reach. However, there are severe obstacles to overcome. For example, cancers evolve in time to become harder targets to treat. Understanding treatment resistance, and its development, often connected with the highly heterogeneous nature of the disease, is the key obstacle. Use of multi-modality imaging techniques such as molecular imaging is one of the solutions that medical physics can offer. Radiomics, where large amounts of useful data are extracted from a single medical scan, is turning into a promising emerging field. However, much more data are available from genomic and other tests, which need to be integrated into the picture as well. To analyze such large amounts of data, we must learn from "big-science" physics approaches. For example, when analyzing data to detect the Higgs boson, CERN physicists relied on a deep understanding of the underlying fundamental physics. Likewise, when we move to medicine, we have to better understand the underlying biological principles that drive observations. However, medical physicists cannot do this alone. To achieve this effectively, it's essential that physicists much more effectively partner with biologists and other scientists beyond current typical collaborative frameworks. In summary, in order for medical physics to remain at the forefront of scientific research, it will have to move beyond the current boundaries towards better addressing advanced disease, better integration of “big-science” physics approaches, better understanding of the biological basis of the disease, and better collaborations beyond current collaborative frameworks.   

Cancer Dormancy and Criticality from a Game Theory Perspective

Background: The physics of cancer dormancy, the time between initial cancer treatment and re-emergence after a protracted period, is a puzzle. Cancer cells interact with host cells via complex, non-linear population dynamics, which can lead to very non-intuitive but perhaps deterministic and understandable progression dynamics of cancer and dormancy. Results: We explore here the dynamics of host-cancer cell populations in the presence of (1) payoffs gradients and (2) perturbations due to cell migration. {\bf Conclusions:} We determine to what extent the time-dependence of the populations can be quantitively understood in spite of the underlying complexity of the individual agents and model the phenomena of dormancy.   

The Emergence of Polyploid Giant Cancer Cells as the Reservoir of Theraputic Resistance

We demonstrate with a microfluidic stress landscape that the generation of polyploid giant cancer cells (PGCCs) is an emergent response by PC3 cancer cells under high stress chemotherapies. We tracked the emergence of polyploid cancer cells using PC3 human prostate cancer cells in a docetaxel gradient and investigated the population dynamics, morphological variations and cell motilities as a function of stress, time and space over the stress landscape. Beyond a sharp transition in cell mortality across the stress landscape, in the highest chemotherapy concentration regions, the PGCCs are the primary survivors. This implies that the PC3 cells acquired survival advantage after polyploidy emergence. We further show that the PGCCs possess stem-like properties, including the expression of cancer stem cell markers and the ability to undergo neosis during the off-drug period in the intermittent chemotherapy cycles, which reveals PGCCs' tumorigenic potential as well as their ability to drive resistance to chemotherapy. We argue that the generation of PGCCs are the adaptive response to chemotherapy on both individual and collective levels, and may be a hallmark of elevated cancer evolution dynamics and therefore could be potentially taken as the target of new treatment strategies.   

T-cell Receptor Diversity During Acute Thymic Atrophy and Resumption

The human body hosts an immense pool of T-cells, and each is capable of responding to a specific antigen. The thymus provides a constant supply of new cells to the peripheral blood. While most circulating T-cells are unique in their antigen specificity, some divide to create clones of identical cells. The thymus experiences acute atrophy during physiological duress, for example during infection, starvation, or psychological distress. Atrophy is accompanied by a dramatic decrease in T-cell export, from which the thymus typically recovers after removal of the stressor. We present an ODE model quantifying the effect of this atrophy and subsequent recovery on the size of the peripheral T-cell pool, which is able to distinguish between clones of different sizes. We find that the time scales of T-cell eradication during thymic collapse and regeneration after recovery are proportional to the disparity in homeostatic cellular proliferation and death rates.   

Quantification of the Rupture Potential of Intracranial Saccular Aneurysms under Contact Constraints

The rupture predictability of intracranial aneurysms is an important medical challenge. While most intracranial aneurysms are asymptomatic, the rupture potential of both symptomatic and asymptomatic lesions is relatively unknown. Moreover, an intracranial aneurysm constrained by a nerve tissue might be a common scenario for a physician to deal with during the treatment process. In this work, we develop a computational model of an intracranial saccular aneurysm constrained by nerve tissues to investigate the protective role of constrained tissue on the aneurysm. A comparative parametric study for constraints of varying length, for aneurysms of varying neck size and aneurysm wall with varying geometric and material models are considered. Our computational results will demonstrate the influence of contact constraints on the level of stress near the fundus and provide an insight on when these constraints are protective and when they are destructive.   

A Momentum-Based Acceleration of the Diffeomorphic Demons Algorithm for Registration of MRI and CT Images of the Brain

Accurate deformable registration of multiple 3D imaging modalities is vital to many areas of diagnostic and interventional radiology and surgery. Diffeomorphic Demons algorithm have been previously reported for mono-modality registration. We extend such methodology to multi-modal (MRI-CT) images of the brain using point-wise mutual information (pMI) with momentum-based acceleration of the optimization. Preprocessing via automatic histogram stretch improved robustness and accuracy of registration in studies involving CT and T1-weighted MRI of a head phantom and clinical studies of five neurosurgery patients. Performance was compared to B-spline Free-Form Deformation (FFD) and Symmetric Normalization (SyN). pMI-Demons achieved target registration error of 0.21±0.07 mm (median±iqr) in phantom and 1.57±0.52 mm in clinical studies, providing alignment comparable to the voxel size without statistically significant difference from FFD and SyN. The pMI-Demons and SyN methods yielded diffeomorphic transformations, whereas FFD yielded unrealistic deformations. pMI-Demons provided a 66% runtime reduction (10 min vs. 30 min for SyN) that facilitates application in rapid image-guided neurosurgery workflows.   

Medical Imaging Physics: Quantitative imaging and analysis of the pediatric spinal cord to detect pathologies.

Diffusion Tensor Imaging (DTI) is a magnetic resonance imaging technique, based on water diffusion in tissue. Using Einstein’s equation of diffusion and the Bloch equation, we can obtain quantitative diffusion indices, such as fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD). We analyzed DTI data from pediatric patients at Vanderbilt Children’s Hospital to quantitatively evaluate the health of the spinal cord. We obtained DTI from 17 pediatric patients, aged 1 to 16 (mean: 7 years, SD: 4 years). Thirteen patients were identified as normal, and four patients were diagnosed with neurological disorders, including Chiari I, tethered cord, and an intra-spinal tumor. We hypothesize that quantitative diffusion indices could assist existing clinical MRI in the diagnosis and prognosis of some pathologies of the cord. We believe that our new data analysis can produce indices that may be able to differentiate between healthy spinal cord and disease. In the presentation, we will review the physics and mathematical concepts utilized in DTI. Then, we will explain how we process the images to obtain DTI metrics. Finally, we will discuss our results, and how they may be potential clinical biomarkers.   

Voxel Based Morphometry in Myotonic Dystrophy Type I

Myotonic Dystrophy Type 1 (DM1) is a degenerative and hereditary disorder; its more typical symptoms are muscle weakness and hypotonia, which may lead to several complications like respiratory failure and cardiac arrest. The aim of this project is to find biomarkers that help us to characterize the evolution of this disorder. In the present work, 27 DM1 patients and 27 healthy controls volunteer participants were matched by age, sex and level of education, then underwent an MRI session in a 3T Philips Ingenia scanner with a 32-channel head sense coil. T1 weighted high resolution images were acquired (FOV=240x240, 180 sagittal slices, spatial resolution 1mm³). For the analysis of data a t-test was applied for unpaired samples, and the FSL 5.0.8 software was used. For the post processing T1 images were reoriented to the standard space, then artifact corrected and the voxel based morphometry script was applied.
Clusters with statistical significance (P<0.05) were found in the regions of angular rotation, temporal lobe, frontal cortex and insula.
Atrophy in gray matter is significant in frontal brain areas, which may imply that different cognitive control processes could be altered.
  

Monte Carlo Bloch Simulation of T1, T2 uncertainties in NMR and MRI pulse sequences

Bloch simulators have been used extensively in NMR and MRI to develop and understand new pulse sequences, investigate artifacts, and more recently for advanced imaging techniques that depend on dictionary lookup tables. Here, we present a new class of Bloch solvers designed to investigate uncertainties in quantitative MRI techniques. Our system couples a Bloch simulator to a Monte Carlo based sampling of a large number of uncertainties and non-idealities that are inherent in NMR and MRI. These include B0 and B1 inhomogeneities, time base jitter and imperfections, transmit and receive phase jitter, temperature variations, coil and electronic noise, variations in initial conditions, nonlinear gradients, jitter in gradient and RF waveforms, errors in fitting due to noise and imperfect data. The Monte Carlo Bloch solver was used to establish uncertainties in primary calibrations of SI-traceable phantom solutions where absolute accuracies are on the order of 1%. After validation of the solver using precision NMR techniques, the system is then applied to MRI systems in which the pulse sequences are less ideal due to the need to minimize imaging time and RF dose. Output simulations are then compared with international MRI T1, T2 round robin studies using the NIST MRI system phantom.   

Fluorescence Lifetime Imaging for Characterization of Cancer Biomarkers: Application to HER2 Positive Tumors

Advances in tumor biology created a foundation for targeted therapy aimed at inactivation of specific molecular mechanisms responsible for cell malignancy. In this paper, we used in vivo fluorescence lifetime imaging with HER2-targeted fluorescent probes as an alternative imaging method to investigate the efficacy of targeted therapy
We have designed an imaging system is a time-resolved fluorescence technique for calculating the lifetime of the contrast agent. HER2-specific Affibody (His6-ZHER2:GS-Cys) (Affibody, Stockholm, Sweden) conjugated with the Dylight750 fluorescent probe (Thermo-Fisher-Scientific, Waltham, Massachusetts) was used as contrast agent and six human cancer cell lines, known to express different levels of HER2/neu are used in athymic mice xenografts.
We have shown that HER2-specific Affibody conjugated with the Dylight750 fluorescent probe as contrast agent, we demonstrated in live animals that fluorescence lifetime of the bounded contrast agent can be used to assess the high to mid expression of HER2 expressing tumors in-vivo. This method can replace invasive , and in-vitro biopsies.   

Improving Ultrasound Transducer Control and Tumor Targeting for 3-D Acoustic Radiation Force Impulse Imaging Guided Prostate Biopsy

Identification and grading of prostate cancer (PCa) is important for determining a proper treatment plan. One current PCa diagnostic procedure involves transrectal B-mode ultrasound for systematic biopsy of the prostate to determine whether cancer is present. The multi-focal nature of prostate cancer, and the difficulty visualizing lesions with ultrasound B-mode, often necessitates 10 or more biopsy samples to determine if a patient has PCa. Even after biopsy the extent and severity of the disease may not be fully understood due to the random sampling afforded by systematic sampling. 3-D acoustic radiation force impulse (ARFI) imaging has demonstrated sensitivity to prostate cancer lesions and their location within the imaged volume. For this study, a graphical user interface was developed to facilitate 3-D ARFI volume acquisitions in the prostate. Additionally, a 3-D Slicer targeting module was created to relate positions in the scanned volume to biopsy sampling locations. Integrating these software tools with ultrasound imaging enabled biopsy targeting with target registration errors of less than 2mm in a prostate mimicking phantom.   

Novel Operating methods of operation for the Rotating Gamma System for safer and more effective stereotactic Radiosurgery near critical organs at risk.

Stereotactic radiosurgery (SRS) with the Rotating Gamma System (RGS) has been used to effectively treat functional disorders of the brain, such as trigeminal neuralgia, arteriovenous malformations, and benign and malignant tumors within the brain with minimal dose spillage to the surrounding healthy tissue. Although under normal operation the RGS produces a sharp penumbra, treatment of lesions near critical organs at risk (OAR) is risky or impossible. Here we present two new operation modes of the RGS, the Intensity Modulated Radiosurgery (IMRS) and Speed Modulated Radiosurgery (SMRS) modes, which could further sharpen the penumbra of the RGS to target lesions near critical OARs. Geant4 based Monte Carlo simulations of the RGS dosimetry were performed for the normal, IMRS and SMRS operation modes of the RGS. Results from the normal and IMRS models were validated with comparisons to experimental data collected with EBT3 films. For the IMRS and SMRS modes the penumbra significantly sharpened along the semi-major axis of the dose profiles; however, the IMRS mode requires a longer dose delivery time compared to the SMRS mode. Nonetheless, both operation modes could be used to target lesions near critical OARs that are not currently possible with the RGS or other SRS devices.   

Implementation of a Parallel Simulating Annealing Algorithm for Intensity Modulated Radiation Therapy Optimization

The purpose of this study is to elucidate the performance improvement of the simulating annealing algorithm (SAA) by parallelizing it on graphics processing unit (GPU) in highly dimensional optimization tasks, such as the Intensity Modulated Radiation Therapy (IMRT) in prostate and lung cancer cases. A MATLAB based implementation of treatment planning for radiation therapy was accomplished by using the computational environment for radiotherapy research (CERR). The planning target volume (PTV) was defined as a quadratic error function, while dose-volume constraints (DVCs) were applied for the dose that the Organs at risk (OARs) would receive separately. The SAA was implemented to determine the optimal intensities that deliver the prescribed dose in the PTV, while satisfying the dose-volume constraints for the OARs. For the parallelization of SAA on the GPU, the Parallel Computing Toolbox in MATLAB version 2016a was employed and the code was launched on four different GPUs. The performance comparison between the different GPUs was established on the speedup factors between the serial and parallelized SAA for different beamlets sizes. In prostate and lung cancer cases, a maximum speedup factor of ~33 for 0.2x0.2 cm² beamlet size was achieved when the K40m card was utilized.