William Paul Segars ' Research

Medical Imaging Simulation Techniques and Computer PhantomsMedical Imaging Simulation Techniques and Computer Phantoms

My current research involves the use of computer-generated phantoms and simulation techniques to investigate and optimize medical imaging systems and methods. Simulation methods are finding an increasingly important role in medical imaging research. Simulation is a powerful tool for characterizing, evaluating, and optimizing medical imaging systems. They have become an important and indispensable complement to theoretical derivations, experimental methods, and clinical studies in medical imaging research and development.

Fig. 1. Computer-based medical imaging simulation.

Simulation involves computer generated phantoms, models of the imaging process, and fast computational methods, Fig. 1. Computer phantoms provide a model of the subject’s anatomy and physiology. Given a model of the physics of the imaging process, acquired data of a computer phantom can be generated using the computational methods. A major advantage to using computer-generated phantoms in simulation studies is that the exact anatomy and physiological functions of the phantom are known, thus providing a gold standard from which to evaluate and improve medical imaging devices and image processing and reconstruction techniques. Other advantages are that computer phantoms are always willing participants and can be altered easily to model different anatomies and medical situations providing a large population of subjects from which to perform research. It is frequently difficult both ethically and practically to test every combination of parameters on patients under clinical conditions.

A vital aspect of simulation is to have a realistic phantom or model of the subject's anatomy. Without this, the results of the simulation may not be indicative of what would occur in actual patients or animal subjects and would, therefore, have limited practical value. I have been leading the development of two realistic digital phantoms for use in medical imaging research, the 4D NCAT and 4D MOBY phantoms. The phantoms are distributed free-of-charge to academic institutions. Companies are also welcome to use them, but we do charge a small licensing fee. Email me (wsegars@jhmi.edu) for more information on how to obtain the phantoms.

The 4D NURBS-based Cardiac-Torso (NCAT) phantom [1-3] (Fig. 2) was originally developed to provide a realistic and flexible model of the human anatomy and physiology for use in nuclear medicine research, specifically single-photon emission computed tomography (SPECT) and positron emission tomography (PET). Non-uniform rational b-splines, or NURBS surfaces were used to construct the organ shapes in the NCAT phantom using the three-dimensional Visible Human CT dataset as their basis. NURBS surfaces can be altered easily to model anatomical variations and patient motion. The NCAT phantom was extended to four dimensions to model common patient motions such as the cardiac and respiratory motions using 4D tagged magnetic resonance imaging (MRI) data and 4D high-resolution respiratory-gated CT data respectively. Both datasets were acquired from normal patient volunteers. With its basis upon human data and the inherent flexibility of the NURBS primitives, the result is a computer-generated phantom that closely resembles the anatomical structures and cardiac and respiratory motions of a normal human subject. Combined with accurate models of the imaging process, the 4D NCAT is capable of simulating imaging data close to that of actual patients. The 4D NCAT phantom has provided an excellent tool with which to study the effects of anatomy and patient motions on SPECT and PET images [2-11]. It is widely used in nuclear medicine imaging research.

Fig. 2. (Left) Anterior view of the original 4D NCAT phantom. (Middle) Cardiac and respiratory motion models of the NCAT phantom. (Right) Emission and transmission simulations performed using the phantom.

Although capable of being far more realistic, the 4D NCAT phantom was originally designed for low-resolution nuclear medicine imaging research, and lacks the anatomical detail required for use in higher-resolution imaging modalities such as x-ray CT. At the same time, there is a lack of realistic and flexible computer-based phantoms for use in this area. We plan to fill that void by building upon the existing 4D NCAT phantom and other simulation tools developed in our laboratory. Unlike current computer phantoms used in x-ray CT, the NCAT has the advantage, due to its design, that its organ shapes can be changed to realistically model different anatomical variations and patient motion. The anatomy and physiology of the NCAT phantom is currently being updated to include the level of detail needed for use in high-resolution x-ray CT imaging research, Fig. 3. In addition, sophisticated models of the x-ray CT imaging process are being developed to generate CT images from the phantom that accurately mimic that obtained from actual patients.

As x-ray CT evolves towards 4D dynamic functional imaging, the simulation tools developed in this work will have applications in a broad range of imaging research in developing image acquisition strategies, image processing and reconstruction methods, and image visualization and interpretation techniques. Also, the tools provide the necessary foundation to achieve our longer range goal to optimize clinical CT applications so as to obtain the highest possible image quality with the minimum possible radiation dose to the patient. Such a task can only be practically and efficiently performed using accurate and realistic computer simulation methods which we are developing.

Fig. 3. (Left) Initial extension of the 4D NCAT anatomy. (Right) Simulated chest x-ray CT images from the extended 4D NCAT. Coronal (top row) and transaxial (bottom 2 rows) reconstructed slices are shown. Images are more realistic than those shown in Fig. 2.

The rapid growth in genetics and molecular biology combined with the development of techniques for genetically engineering small animals has led to increased interest in in vivo small animal imaging. With the rise of small animal imaging, new instrumentation, data acquisition strategies, and image processing and reconstruction techniques are being developed and researched. A major challenge is how to evaluate the results of these new developments. Simulation techniques can provide a vital tool to evaluate and improve molecular imaging devices and techniques. Currently, there is a lack of realistic computer-generated phantoms modeling the mouse anatomy and physiological functions for use in molecular imaging research.

The same methods and techniques used to develop the 4D NCAT phantom were used in the creation of a new 4D mouse whole body (MOBY) phantom [12]. The organ shapes are modeled with NURBS surfaces. High-resolution 3D magnetic resonance microscopy (MRM) data obtained from the Duke Center for In Vivo Microscopy was used as the basis for the formation of the surfaces. Cardiac and respiratory motions were modeled using a gated black-blood magnetic resonance imaging (bb-MRI) dataset of a normal mouse as the basis for the cardiac model and respiratory-gated MRI and known respiratory mechanics as the basis for the respiratory model. The gated MR images were obtained from the University of Virginia. In each case, the time-changing 3D surfaces are defined by a set of time curves to create time continuous dynamic or 4D NURBS surface models. The MOBY phantom provides a unique and useful tool in molecular imaging research, especially in the development of new imaging instrumentation, image acquisition strategies, and image processing and reconstruction methods.

Fig. 4. (Left) Anterior view of the 4D MOBY phantom. (Middle) Cardiac and respiratory motions of the MOBY phantom. (Right) MicroCT and MicroSPECT images simulated using the phantom.

Fig. 6. Registration results obtained for one patient. CT (gray level) is shown fused with the PET transmission data (orange). The non-rigid method better compensates for the respiratory differences (diaphragm position, size of lungs).

My current research also involves the development of a fast, automatic, non-rigid method for PET-CT image registration [14,15]. With the advent of the combined PET-CT scanner, registration of 3D CT and PET data to allow simultaneous viewing of structural and functional information is becoming an important area for research. The difference in the speed of image acquisition between CT and PET does, however, introduce a new problem, the effect of respiratory motion. Image acquisition for 3D CT data typically occurs during a single breath hold while 3D PET data is obtained over a longer acquisition time and includes many respiratory cycles. Therefore, the effect of respiratory motion must be an important consideration when attempting to register CT with PET data. Many non-rigid transformation techniques have been researched to compensate for the differences in PET and CT due to respiration. These techniques, however, do not take into account the underlying physiology of respiration in formulating the transform.

I seek to improve PET-CT image registration by taking into account the underlying physiology of respiration. Using the realistic respiratory model of the 4D NCAT phantom, I have been working on the development of an algorithm to non-rigidly transform 3D CT images obtained during a single breath hold to better match that of the 3D PET data of the same patient obtained during many respiratory cycles, Fig. 6.

Through my work in developing the phantoms and the non-rigid registration technique, I have gained a great deal of experience in using and developing different image segmentation methods. I have developed automatic and semi-automatic routines to help segment the organs from the imaging data used as the basis for each phantom. The segmented organs were then fit with 3D NURBS surfaces. For the registration program, automatic segmentation routines were developed to segment the body, lungs, liver, heart, and diaphragm when possible from CT and PET images. A comparison of the segmented organs from each data set helped to determine the non-rigid transformation due to the respiratory motion.

I recently became involved in a bioinformatics project to help develop a web-based interface to display phenotypic and genotypic prostate cancer data obtained from many patients within an anatomical framework. Organ models similar to those of the 4D NCAT are being developed based on imaging data to represent normal and diseased (large metastases and micro metastases) states. The genetic and phenotypic information of a particular patient will be displayed over the anatomical framework provided by the organ models using color-coding and other such methods.

An area of research in which I would like to get involved is computer-aided diagnosis (CAD). Much research is being performed to investigate methods to efficiently analyze and handle the vast amount of data that can now be collected for a given patient. These large volumes of data can be overwhelming and impractical to review in standard radiological practice. Various CAD methods are being developed to aid in the interpretation of the increasing amounts of medical image data. Computers have shown the potential to provide an excellent second opinion to the physician or radiologist making it an invaluable tool for use in patient diagnosis and treatment. Much of the research in CAD, however, has been limited to x-ray CT. In the future, I would like to investigate extending these methods to nuclear medicine.