The technical aspects of this program will be pursued with the following specific aims:

1) Optimize acquisition methods for 3D sonography using reconstruction and real-time techniques. 2) Develop techniques for registering anatomical images from 2D and 3D ultrasound with those obtained from prior ultrasound examinations and from magnetic resonance and computed tomographic imaging, considered “gold standards” for non-invasive anatomical imaging. 3) Develop tools for abstracting, in an automated fashion, anatomic changes from serial 3D and 2D ultrasound studies. 4) Develop algorithms for the optimal compression of 3D ultrasound images. 5) Assess the ability of novice examiners to obtain 3D sonographic datasets after minimal training.

These objectives will be pursued using data from a variety of in vitro, animal and clinical models. In particular, we will take advantage of a well-established collaboration with the National Institutes of Health, which permits highly sophisticated chronic animal models to be examined with a minimum of additional resources. Although the tools developed here should be applicable to any organ of the body, we will focus our efforts in two areas that present unique challenges and also are highly relevant to spaceflight. 1) The Kidneys: Bone demineralization during prolonged spaceflight may increase the risk for kidney stones. From a technical perspective, the kidney may be a particularly easy organ to register in three dimensions with prior examinations, being static and having well-defined anatomical landmarks. Using computed tomography as the reference standard, we will examine serially patients with renal and ureteral calculi, paying particular attention to the development and resolution of hydronephrosis and calyceal dilation. 2) The Heart: Manned space flight and prolonged exposure to microgravity may result in significant myocardial atrophy. Furthermore, cardiovascular emergencies such as myocardial infarction and cardiac trauma produce structural changes that are important in guiding therapy. From a technical perspective, the heart is a more difficult organ to register than the kidney, since it is dynamic. The presence of well-defined anatomical landmarks, however, assists in this process. To better understand the process of cardiac remodeling with unloading, we will examine two groups of patients in whom sudden changes in chronic loading conditions are known to produce profound ventricular remodeling: 1) patients with aortic stenosis undergoing aortic valve replacement; 2) patients with chronic severe aortic insufficiency undergoing aortic valve replacement or repair. These patient groups will provide insight into the determinates of ventricular remodeling, while forming an important testbed for the assessment of our registration and comparison algorithms for detecting changes in ventricular morphology over time.

In addition to the practical application to manned spaceflight and the ISS, these systems each represent significant societal problems for which the incidence of disease is common enough to study within the time period of this grant. Furthermore, ultrasound is the standard diagnostic tool used in the management of patients with each of these conditions. At the conclusion of this project, we anticipate delivering to the National Space Biomedical Research Institute and its Smart Medical Care System a set of algorithms and software for the nonrigid morphological registration and comparison of serial two and three dimensional ultrasound datasets and validated algorithms for optimal compression of 3D ultrasound data. In addition to these technical deliverables, our validation work on nephrolithiasis will provide important diagnostic clues for assessing this condition in manned spaceflight. Similarly, the work on cardiac mass regression following unloading will be invaluable to the NASA research and medical operations community in assessing the impact of long-term spaceflight on cardiac atrophy and utility of prophylactic countermeasures.

Compression, Visualization, Segmentation, And Registration Techniques For 3D Ultrasound: Despite significant advances in ground-based and satellite data communications, digital transfer of data remains a formidable challenge. Optimistic NASA estimates indicate that the overall Deep Space Network (DSN) bandwidth will not exceed 400 kbps until the year 2004 and transfer rates not exceeding 1 Mbps until the year 2010. Therefore, assuming 100% error-free, 100% efficiency/availability, and complete dedication of the entire DSN bandwidth, transmission of a single cardiac cycle 3D volumetric dataset with an estimated size of 60 MB would take over 15 minutes. Clearly, it is easy to appreciate the need for high degrees of volume compression if diagnostic and potentially therapeutic ultrasound imaging is to be applied to future NASA deep space missions. Even the relatively low resolution Volumetrics device generates up to 60 MB of data per second and has proven cumbersome for visualization and quantification. To make this device applicable to space use, we validated the use of a 3D wavelet packet transform for 100:1 compression of 3D echo. We have shown the ability of maintaining a higher signal-to-noise ratio using a modified set partitioning in hierarchical trees algorithm (MSPIHT) over the conventional technique. Also, we have developed highly compact software for visualization, segmentation, and registration. These now can register pre- and post-exercise images and cross-modality data as described next.

Mutual Information-Based Registration of Ultrasound Volumes: We have shown that both rigid-body and affine misalignments can be recovered between volumetric ultrasound images using mutual information-based registration. In this approach, one of the two images (primary) is kept stationary and the other (secondary) is transformed iteratively until the optimal alignment, corresponding to the maximum of the mutual information (MI) function, is found. MI is an intensity-based measure, which conveys the amount of information that the primary contains about the secondary. MI is more effective than other intensity-based measures such as simple cross-correlation, since it is a statistical measure that works even if voxel intensities in the primary are nonlinearly related to those in the secondary. We search for the maximum of the MI by searching for a set of geometric transformation parameters, which produce a 4 x 4 transformation matrix T. Upon convergence, the MI is maximized between the primary (A) and the transformed secondary (TB). T refers to rigid-body transformation if it incorporates rotation and translation only. The transformation is affine (nonrigid) if it includes scaling and shearing, as well.

Registration of Resting Cardiac Ultrasound and SPECT: We have also performed MI-based registration of gated Sestamibi SPECT and RT3D ultrasound images in a pilot study involving nine subjects. Since ultrasound collects more frames per cardiac cycle, 8 original SPECT frames were interpolated with the aid of associated electrocardiograms to match the available ultrasound frames. The average rating for success made by 5 experts was 3.58, with a 1 meaning poor registration and 5 meaning near perfect spatiotemporal registration. Of note, we detected virtually all clinically relevant abnormalities, such as a perfusion defect in the right coronary artery territory that matched with a wall motion abnormality in the inferior wall by 3DUS. Only 7 of 45 evaluations were scored below 3; these were primarily due to poor endocardial definition in ultrasound images – attributed to the quality of the Volumetrics 3DUS device rather than the registration procedure itself. Recently, the image quality of 3D ultrasound has improved significantly. We will further explore cross registration capabilities in Specific Aim #2.

Teaching Novice Examiners to Obtain 2D and 3D Data With Minimal Training: A major issue with diagnostic imaging in space is the limited training that astronauts are likely to have, given the other demands on their schedule. In an important recent study, 5 novices with only 4 hours of training were able to perform technically adequate 2D echocardiograms with remote coaching from an experienced sonographer. Quantitative parameters (e.g., LV mass, ejection fraction, mitral inflow) agreed with (<±10%) and correlated well with (r>0.85) measurements by an expert. 3D studies of both the heart and kidney are ongoing, but anecdotal observations from this work suggest that 3D imaging is even easier to learn. Precise orientation of the probe is not necessary for 3D acquisitions; however, images obtained that must include the structure of interest and be free of near-field and other artifacts.

Facilitate ISS Ultrasound Use and Optimize Digital Downlink Technology: After 4 years of work with JSC engineers during our first grant (NCC9-60), a Philips HDI-5000 ultrasound system was launched to the International Space Station in March, 2001. Since then we have worked closely with Medical Operations physicians and engineers accomplishing several goals: optimizing the probe availability for this unit, testing digital file transfer from ISS to the ground, and developing live video feed capability for guiding the acquisition and real-time interpretation of data by a ground based expert. Digital echocardiography is now fully feasible on the ground with over 250 studies being stored daily at the Cleveland Clinic (~3 terabytes annually). The American Society of Echocardiography has made all-digital storage and analysis a major policy goal of the organization, and the PI now chairs that effort.

Development of wireless technology to interface with a handheld digital ultrasound system: We continue to work with Kevin Montgomery of the Stanford Biocomputation Center/NASA Ames Research Center. Their approach has been to develop small, wearable sensors that register heart rate, respirations, temperature, etc.; then, digitize these signals, and transmit them using wireless protocols. The goal is to capitalize on the wireless development effort and the infrastructure being developed at Ames and Stanford. This would allow full resolution (640x480x30 Hz) ultrasound to be made available in real-time on a web site with a latency of less than 2 seconds, short enough to allow realistic coaching and guidance of the acquisition by an off-site expert. To accomplish this, a wireless testbed within the Cleveland Clinic echocardiography laboratory is utilized to interface to the portable Sonoheart ultrasound device (Sonosite, Bothell, WA) via a PC with wireless capability using a video capture card. We have investigated echocardiographic exam review using Windows CE devices for their ability to view ultrasound data when connected using the 802.11 protocols.

2003; 41: 520A , Jan-2003

2003; 41: 430A , Jan-2003

2003;16:501. , Jan-2003

2003;16:501. , Jan-2003

2003;16:551, Jan-2003

2003;16:546. , Jan-2003

2003;16:551, Jan-2003

2003;16:552. , Jan-2003

2003; 16: 546 , Jan-2003

2003; 16: 517 , Jan-2003

2003; 108: IV-338. , Jan-2003

2003; 108:IV-554. , Jan-2003

2003; 108:IV-595, Jan-2003