Quanzheng Li, Ph.D. - US grants

DESCRIPTION (provided by applicant): Positron emission tomography (PET) with FDG has become a widely accepted and used clinical molecular imaging tool for disease diagnosis, staging, treatment planning, management and evaluation. Although conventional static PET imaging provides high sensitivity in tumor detection, further improvement is important since even a small percentage of false negatives can have a major impact on treatment, cost and outcome. Visual inspection of static images is potentially inaccurate for small tumors due to limited spatial resolution and low lesion-to-background contrast. Computer aided detection (CAD) combined with use of dynamic PET data could assist in improving sensitivity and specificity for these small lesions. The goal of this exploratory Bioengineering Research Grant proposal is to investigate such a CAD method for dynamic FDG PET that integrates image reconstruction, lesion detection and thresholding in a statistical framework. The method will be optimized based on the properties of the dynamic PET data and the imaging system, and is designed to use standard dynamic data sets without the need for a measured blood input function. The CAD system will automatically provide a voxel-wise statistical map indicating probable lesion locations. By using a statistical detection algorithm that combines spatial and temporal information, we expect to be able to improve detection of small lesions that are not clearly visible in standard static scans and thereby provide improved diagnostic information to the radiologist. We will apply our maximum a posteriori (MAP) approach to PET image reconstruction to data from the new generation of clinical scanners, and optimize performance in terms of modeling and calibration procedures based on the characteristics of the scanner. The resulting images of estimated dynamic tracer uptake, as well as their approximate covariance, computed based on a theoretical analysis of the reconstruction algorithm, will be used as input to a matched subspace detector. This detector characterizes typical tumor and normal tissue dynamics using linear subspaces in combination with a generalized likelihood ratio test, to generate a voxel-wise statistical map indicating the likelihood of tumor presence or absence. Typical tumor and normal tissue subspaces will be obtained using a training dataset from multiple subjects with tumor and normal tissue regions of interest (ROIs) identified by a radiologist. The statistical detection map will then be thresholded to obtain a voxel-wise indication of likely tumor locations, while controlling for the effects of multiple comparisons. We will implement, optimize and perform preliminary evaluation of this CAD approach for dynamic data collected at USC using the Siemens Biograph TruePoint scanner. Evaluation will use Monte Carlo simulation and retrospective human studies. Human studies will focus on patients with liver metastases from colorectal cancer who are enrolled in an ongoing clinical trial. Serial imaging studies, with subsequent surgical resection and independent verification through pathology and intraoperative ultrasound, will provide a basis to evaluate the performance of our CAD detection approach. PUBLIC HEALTH RELEVANCE: Positron Emission Tomography (PET) has been widely used in cancer diagnosis, staging, treatment planning, management and evaluation. One of the main functions of PET is to detect tumors and metastatic lesions, which is conventionally done by visual inspection of a static volumetric image by a radiologist. This project is focused on using multiple images of the patient collected in a single session, in combination with a novel computer aided detection (CAD) method, to assist radiologists in detecting small tumors that may not be clearly visible using standard imaging protocols. Success of this project may lead to improved detection, staging and monitoring of metastatic disease.

DESCRIPTION (provided by applicant): Positron Emission tomography (PET) is a major molecular imaging tool in oncology, with applications ranging from diagnosis and staging to patient management. Despite the broad use of PET in the clinical environment, there is no quantitative PET imaging method available for routine clinical practice. The currently used static scan can provide a semi-quantitative measurement, standardized uptake value (SUV), for a whole body scan. However, it completely ignores the dynamic nature of radiopharmaceutical kinetics. The popular semi-quantitative dual time point method can approximate the kinetic differences at two time points by comparing activities but usually requires an extended waiting time for the second scan. The multiple time point method can calculate the net influx rate but still requires long scan duration and makes a whole body scan infeasible. The challenge of a quantitative whole body dynamic PET scan lies in how to estimate the quantitative functional values, such as net flux rate, using data from a short acquisition period, and how to accelerate the computation to make it practical in a clinical setting. We address this challenge by developing and optimizing a novel data analysis method and implementing it using a high performance computing tool. We take advantage of the linearity of Patlak graphic analysis to model the tracer activity in each voxel as a linear combination of the blood input function and its integral, weighted by the Patlak parameters including net influx rate. In addition, we derive a simplified model of the blood input function, based on the same assumptions used to derive Patlak parameters from the kinetic compartment model. We then estimate the Patlak parameters and the parameters in the blood input function in a penalized maximum likelihood estimation framework using the list mode data and its associated inhomogeneous Poisson statistical model. We also theoretically analyze the performance of our Patlak estimator in terms of noise, resolution and signal-to-noise ratio (SNR), and use the results to guide us in optimizing the scan duration and any movement of imaging bed to achieve the best SNR. The advanced estimation algorithm, along with an accurate imaging system model, can robustly compute the net influx rate using the list mode data in a short acquisition without a measured blood input function, and make whole body dynamic scans practical. Our algorithms will be implemented on an Nvidia Tesla GPU (graphics processing unit) based workstation, a new computing tool that provides computational power previously available only on a mini super computer. We will further accelerate our algorithms using a combination of efficient representation of the list mode data and the system matrix. We will evaluate the performance of the proposed method and compare it with SUV, the dual time point method, and the traditional Patlak method, using simulated and clinical data. We will use a range of performance metrics, including region of interest (ROI) bias, ROI variance, lesion detectability, and computer and human observers. This project will eventually provide a quantitative dynamic whole body PET imaging protocol that can potentially improve the sensitivity and specificity of PET imaging in oncology.

DESCRIPTION (provided by applicant): The long-term goal of this project is to develop a novel SPECT system design that tackles one of the most limiting aspects of SPECT instrumentations by offering a greatly improved sensitivity without sacrificing imaging resolution. This proposed approach is based on the use of a novel detection system called dense- camera-array (DCA). As we have demonstrated with a Monte Carlo study described in Sec. C.2, a small animal SPECT system based on the DCA detectors could a photon detection efficiency of >1% (as compared to the typical levels of 0.1%-0.01% found in modern pre-clinical SPECT instrumentations), while maintaining an excellent spatial resolution. This dramatic increase in sensitivity could potentially provide a radical change in how we might employ SPECT imagining in both pre-clinical and (potentially) clinical practice, by offering a dramatically lowered detecton limit and allowing for new imaging procedures that would be difficult to implement with the current generation of SPECT instrumentations. The design of dense camera array (DCA) is inspired by the compound eyes often found on small invertebrates, such as flies and moths. A DCA camera consists of a large number of independent micro-pinhole-gamma- camera-elements closely packed in a dense array (e.g. 10-20 independent camera-elements per cm2). Each of the micro-camera-elements covers a narrow view angular through the object. When constructing a SPECT system with multiple DCAs, there will be a very large number (up to several thousand) of micro-camera- elements in the system pointing towards the object and collecting gamma rays simultaneously. This is the key for attaining a super-high detection efficiency, while maintaining an excellent imaging resolution. One of the key challenges for constructing the DCA camera is the need for a state-of-art detector technology that offers an ultrahigh 3-D spatial resolution (e.g. 100mm), an excellent energy resolution, an adequate count-rate capability and a very high stopping power for energetic gamma rays. This would allow us to pack 10-20 independent micro-camera-elements into 1 cm2 area, and to ensure each micro-camera-element having a sufficient resolving power. For this project, we will utilize a recently developed small-pixel CdTe/CZT detector equipped with a hybrid pixel-waveform readout system to construct the prototype DCA cameras.

Neurodegenerative dementias are devastating neurological illnesses. Estimations point to more than 24 million people worldwide affected by dementia. Dementia ranks as one of the top causes of disability in elderly people, and the estimated annual cost of caring for dementia patients is around $175 billion in the United States. During the last two decades, research has particularly focused in describing the early stages of the disease with the hope of implementing preventative treatments with disease modifying agents. However, the recent expansion of our understanding of aging and preclinical Alzheimer?s disease (AD) through the development of molecular imaging biomarkers of amyloid-beta (A?), and, more recently, Tau using Positron Emission Tomography (PET) has revolutionized the field and shifted interest to earlier pathophysiologic events. AD neurodegeneration follows specific neuronal circuits of the brain. However, it is still unknown how Tau and A? accumulation progresses along those large-scale brain systems, and how this may produce the structural/functional breakdown associated with symptomatic AD phases. The network nature of AD-related pathology has been recently proven by our group. Our findings have supported that A? accumulates following specific routes of the brain connectome. Moreover, our preliminary results have shown that both Tau and A? networks interact in temporal and cingulate areas at the early stages of AD pathology. In this proposal, we therefore hypothesize that AD propagates along connectivity networks between the medial temporal and cortical areas, leading to clinical conversion to cognitive impairment or AD. In this R01 effort, we will develop, optimize and validate imaging and network analysis and apply them to the existing data in the Harvard Aging Brain Study (HABS) to test this hypothesis. We will first develop methods to efficiently and accurately estimate the covariance of distribution volume ratio (DVR) image and then de-couple the covariance of DVR estimation from the physiological correlation. We will then use graph regression model algorithms to build the high-resolution brain pathological networks, and apply Cortical Hubs, Interconnectors and Stepwise Connectivity analyses to study the properties of these networks. We will also analyze the longitudinal data in HABS using Dynamic Bayesian Network and study the origin and pathways of the propagation of AD pathology, particularly focused on the relationship between the medial temporal lobe and the cingulate cortex. Finally, we will develop a graph theory tool to evaluate an individual subject?s risk to AD based on a matched subspace approach.

Abstract Simultaneous PET/MR can be considered as an integrated imaging modality only if the information of both modalities is integrated together. In current routine PET/MR applications, the PET and MR scans are performed separately, and the images are reconstructed separately as well. The information is integrated only at the application level. Here we propose unified methodologies of joint PET/MR image reconstruction, a paradigm shifting new way to integrate information of PET and MR to significantly maximize the outcome of PET/MR. The PET and MR scanners indeed measure different physical or physiological signals, but there are still redundant information (e.g. tumor boundary and mutual information) between the images obtained with the two modalities that can be utilized to build connection between PET and MR images in a potential joint reconstruction. In addition, if the compartmental model is taken into account, the physiological parameters estimated from PET and MR can have overlaps, and therefore the parametric image (voxel-wise kinetic parameters) estimated from one modality could be directly used to help the estimation of the parametric image of the other modality. Therefore, there are inter- connections between these two modalities that we can use to develop elegant methods of joint reconstruction. We will first take advantage of the simultaneous acquisition of PET/MR to develop a static image reconstruction with anatomic prior derived from MR images, and to develop methods to jointly reconstruct gated PET images using a motion field computed from MR images. We believe in both cases, the quality of PET images will be significantly improved compared to traditional approaches. For PET/MR, there are many novel ways to jointly model the dynamic PET and MR images. We will thus develop an alternating direction method of multipliers (ADMM) to directly estimate the voxel-wise kinetic parameters of dynamic PET and dynamic MR together from raw data. This will achieve the maximum signal noise ratio of parametric images for both dynamic PET and MR. We will also investigate novel approaches to parametric imaging of non-stationary kinetic modeling in which not only the images are estimated but also the uncertainty on those estimates of the parametric images. The knowledge of uncertainty is important when making decisions about progression/regression of the disease, signal detection, etc. We will use a method developed in our laboratory in which the noise in raw PET data will be transferred to parameter images using origin ensemble algorithm.