Research

Cherenkov radiation is a phenomenon observed when charged particles travel through a medium at speeds exceeding the speed of light in that medium, resulting in a distinctive blue glow. This effect becomes relevant in the context of radiopharmaceuticals, which are compounds used in medical imaging and treatment that emit radiation. When these radiopharmaceuticals decay within a biological system, they can generate Cherenkov radiation, which can be captured and imaged using advanced optical imaging techniques. By developing sophisticated imaging methods, we aim to visualize this radiation, providing valuable insights into the distribution and intensity of the absorbed dose within experimental models, such as mice. This capability is especially crucial for assessing the dose delivered by daughter isotopes of alpha-emitters, which can redistribute during radiopharmaceutical therapy. Accurate imaging and dose prediction are essential for optimizing treatment protocols and minimizing side effects, thereby improving the efficacy and safety of targeted radiotherapy.

Dysphagia is a common clinical symptom swallowing dysfunction that causes difficulty or inability to move an alimentary bolus from the mouth to the esophagus. The prevalence of dysphagia is 30–50% in the elderly ( >65 years old), 40–80% in patients with stroke, 80% in patients with Alzheimer disease, 60% in patients with Parkinson disease, and 50% in patients with head and neck cancers. Currently, the most used procedure to diagnose dysphagia is the videofluoroscopic swallow (VFS) study. VFS studies use a barium contrast-enhanced radiographic procedure designed to define anatomy and physiology during a patient’s swallow in order to identify swallowing-related problems, such inefficient swallowing leading to residue in the mouth and pharynx and aspiration of material into the lungs. Given the complicated biomechanics related to swallowing, there is a compelling clinical need for automated image processing tools to provide objective, quantitative analysis of fluoroscopic studies in busy and resource limited clinics. To this end, we are developing and validating an artificial neural network-based software that will segment and track clinically important swallowing structures on a frame by frame basis within swallowing videos. Segmentation and tracking will automatically occur post acquisition with no needed input or video editing. Frame by frame auto-segmentation of regions of interest will allow for quantitative metrics to be determined algorithmically removing the time-consuming nature of these metrics.

Radiopharmaceutical therapy is emerging as an attractive treatment option for a broad spectrum of tumor types because it has the potential to simultaneously eradicate both the primary tumor site as well as the metastatic disease throughout the body. Patient-specific absorbed dose calculations for radiopharmaceutical therapies are important for reducing the risk of normal tissue complications and optimizing tumor response. Our lab has developed a Monte Carlo internal dosimetry platform called RAPID (Radiopharmaceutical Assessment Platform for Internal Dosimetry) in order to calculate patient-specific voxelized dose distributions in a clinically feasible time frame and to examine and quantify the dosimetric impact of various parameters and methodologies used in 3D internal dosimetry methods. This platform utilizes serial PET/CT or SPECT/CT images to calculate voxelized 3D internal dose distributions with the Monte Carlo code Geant4. Dosimetry can be computed for any diagnostic or therapeutic radiopharmaceutical and for both pre-clinical and clinical applications.

Microdosimetric and nanodosimetric calculations are sometimes needed to understand and interpret the mechanisms of radiobiological effects from RPT. This is especially true for Auger emitting and alpha emitting RPT agents. Additionally, deterministic biological effects including tumor response and normal tissue toxicity for RPT often do not correlate with mean absorbed dose metrics as they do for external beam radiotherapy. Our prior work on organ- and tumor-level dosimetry is being expanded to much smaller scales ranging from millimeter tissue samples to nanometer DNA strands.

Ultrasound has been used for decades as an important medical tool to observe the inside of the body in a cheaper and more efficient manner than other imaging modalities such as MR or CT. However, there are a variety of medical applications that could benefit from having the ability to acquire ultrasound images simultaneously with these other imaging modalities. In collaboration with GE Global Research Center we have developed the first e4D MR and CT compatible ultrasound probe. The probe is designed for hands free operation. Both the probe and transmit cable are RF shielded permitting use in an MR environment. The probe is being evaluated as a novel tool to help improve motion management for external beam radiation therapy. Other applications of the probe include image-guided surgery, biopsies, and drug delivery.

Features in the body such as tumors, organs and vessels move as a result respiratory and cardiac motion. Several medical applications require the robust tracking of these features in order to deliver a treatment more effectively. Computer algorithms can be written that import medical image data and track the movement of relevant anatomy during image-guided therapies such as external beam radiation therapy and interventional procedures. These algorithms require fast imaging processing so that changes in the therapy can be made in real-time to account for the influence of motion. We have developed a GPU-based feature tracking algorithm that is capable of real-time tracking of two and three-dimensional structures in ultrasound and MR images.

Clinical implementation of motion management systems require that they first be validated through measurement. Robust testing of these systems requires a phantom that is able to simulate translational motion and deformation, be compatible with a variety of imaging modalities, and be reusable for multiple experiments. Ideally, these phantoms would also provide three-dimensional (3D) dosimetry and steps towards the clinical implementation of deformable 3D dosimetry have been made. We have developed a deformable anthropomorphic phantom that incorporates deformable 3D gel dosimetry and is compatible with a variety of imaging modalities including both MRI and ultrasound. This phantom and dosimeter pairing has the ability to perform the static deformation measurements required for testing deformable dose accumulation algorithms and the dynamic deformation measurements required for testing motion management systems.

Clinical implementation of Monte Carlo (MC) dose calculation algorithms has been hindered primarily by enormous computational and time requirements.  In close collaboration with Virtual Phantoms Inc. and Rensselaer Polytechnic Institute, we are helping to develop and validate ARCHER, which is a Monte Carlo-based dose calculation engine capable of accurate, sub-minute dose calculations. ARCHER uses a symmetric execution model in order to efficiently make use of concurrent CPU and GPU computation. In addition, a benchmarking platform for ARCHER was created, using the well-benchmarked EGSnrc code package for a Varian TrueBeam accelerator.

Future radiation treatment technology will require new inverse planning solutions to account for an increasing number of physical and biological parameters used for advanced delivery modes. The difficult challenge of preserving dosimetric accuracy while finding an optimal treatment plan over a vast search space would greatly benefit from Monte Carlo based dose calculation methods. However, the dosimetric accuracy afforded by conventional Monte Carlo treatment planning methods come at the cost of being computationally time-consuming and memory-intensive to implement. This has largely relegated Monte Carlo treatment planning to research endeavors, where the computational time and memory considerations are less important than the accuracy afforded by Monte Carlo. While advancements in computer hardware and software have greatly improved the speed of Monte Carlo dose calculation methods, the computational demands of newer radiation therapy optimization problems still outpace the scalability of the current optimization paradigm. Our group has developed a novel approach to fluence optimization that directly incorporates optimization into the Monte Carlo transport process itself, in a manner that reduces both the computational time and memory penalties of performing transport and optimization separately.

Conventional x-ray irradiators used in radiobiology are designed to provide maximum versatility to radiobiology researchers. This added versatility often impedes the overall sensitivity and specificity of an experiment resulting in a trade-off between the number of absorbed doses (or dose rates) and biological endpoints that can be investigated in vivo or in vitro in a reasonable amount of time making modern irradiator designs incompatible with current research needs. Furthermore, important dosimetry and calibration characteristics (i.e., dose build-up region, beam attenuation, and beam scatter) of these irradiators are typically unknown to the end user, which can lead to significant deviation between delivered dose and intended dose to cells that adversely impacts experimental results. Our laboratory has developed new dosimetry formalisms for novel radiobiology irradiators. We have also developed the first fully automated and highly accurate high-throughput micro-irradiator to aid in radiation sensitivity investigations.