Adding value to clinical data by linkage to a public death registry.

We describe the methodology and impact of merging detailed statewide mortality data into the master patient index tables of the clinical data repository (CDR) of the University of Virginia Health System (UVAHS). We employ three broadly inclusive linkage passes (designed to result in large numbers of false positives) to match the patients in the CDR to those in the statewide files using the following criteria: a) Social Security Number; b) Patient Last Name and Birth Date; c) Patient Last Name and Patient First Name. The results from these initial matches are refined by calculation and assignment of a total score comprised of partial scores depending on the quality of matching between the various identifiers. In order to validate our scoring algorithm, we used those patients known to have died at UVAHS over the eight year period as an internal control. We conclude that we are able to update our CDR with 97% of the deaths from the state source using this scheme. We illustrate the potential of the resulting system to assist caregivers in identification of at-risk patient groups by description of those patients in the CDR who were found to have committed suicide. We suggest that our approach represents an efficient and inexpensive way to enrich hospital data with important outcomes information.

Tissue composition affects measures of postabsorptive human skeletal muscle metabolism: comparison across genders.

Despite clear anthropomorphic differences, gender differences in human skeletal muscle protein and carbohydrate metabolism have not been carefully examined. We compared postabsorptive forearm glucose, oxygen, and lactate balances and forearm protein kinetics between 40 male and 36 female subjects. Forearm composition was measured in a subset of 17 subjects (8 males and 9 females) using multislice magnetic resonance imaging. Oxygen uptake, net phenylalanine release, and estimated rates of forearm protein synthesis and degradation were greater in male than in female subjects when expressed as the rate per 100 mL forearm volume (P < 0.05). In males, however, muscle accounted for 58% of forearm volume, compared with 46% in females (P < 0.001). When phenylalanine balance, protein degradation and synthesis, and glucose and oxygen uptake were expressed per 100 mL forearm muscle, there were no significant differences across gender. Likewise, the extraction fractions for oxygen, glucose, phenylalanine, and labeled phenylalanine were comparable in males and females. We conclude that cross-gender comparisons of metabolic variables must accommodate differences in tissue composition. These data indicate that in the postabsorptive state, skeletal muscle metabolism of glucose, protein, and oxygen do not differ by gender in healthy young humans.

Proton magnetic resonance spectroscopy in patients with glial tumors: a multicenter study.

The authors represent a cooperative group of 15 institutions that examined the feasibility of using metabolic features observed in vivo with 1H-magnetic resonance (MR) spectroscopy to characterize brain tumors of the glial type. The institutions provided blinded, centralized MR spectroscopy data processing long with independent central review of MR spectroscopy voxel placement, composition and contamination by brain, histopathological typing using current World Health Organization criteria, and clinical data. Proton 1H-MR spectroscopy was performed using a spin-echo technique to obtain spectra from 8-cc voxels in the tumor and when feasible in the contralateral brain. Eighty-six cases were assessable, 41 of which had contralateral brain spectra. Glial tumors had significantly elevated intensities of choline signals, decreased intensities of creatine signals, and decreased intensities of N-acetylaspartate compared to brain. Choline signal intensities were highest in astrocytomas and anaplastic astrocytomas, and creatine signal intensities were lowest in glioblastomas. However, whether expressed relative to brain or as intratumoral ratios, these metabolic characteristics exhibited large variations within each subtype of glial tumor. The resulting overlaps precluded diagnostic accuracy in the distinction of low-and high-grade tumors. Although the extent of contamination of the 1H-MR spectroscopy voxel by brain had a marked effect on metabolite concentrations and ratios, selection of cases with minimal contamination did not reduce these overlaps. Thus, each type and grade of tumor is a metabolically hetero-geneous group. Lactate occurred infrequently and in all grades. Mobile lipids, on the other hand, occurred in 41% of high-grade tumors with higher mean amounts found in glioblastomas. This result, coupled with the recent demonstration that intratumoral mobile lipids correlate with microscopic tumor cell necrosis, leads to the hypothesis that mobile lipids observed in vivo in 1H-MR spectroscopy may correlate independently with prognosis of individual patients.

Functionality of gray-scale display workstation hardware and software in clinical radiology.

This article examines the functional factors crucial for the successful conversion from film-based radiography to radiologic gray-scale display systems, including hardware architecture and software requirements, radiologic workstation operations, and a multilayered intelligent user interface. Radiologic workstation operations are logically decomposed into case preparation, case selection, case presentation, case interpretation, and documentation and presentation of the diagnosis. A multilayered software architecture for an adaptive, intelligent user interface is proposed: a hardware interface layer, an object-oriented layer, and a knowledge-based layer. The knowledge-based layer is composed of three elements: image presentation based on context-dependent models of diagnostic requirements, knowledge-based expert systems for assistance in diagnostic decision making, and computer-assisted diagnosis to alert the radiologist to potential lesions or abnormalities.

Simulation of spatial and contrast distortions in keyhole imaging.

Keyhole imaging is a scheme introduced to improve temporal resolution in dynamic contrast-enhanced MRI by a factor of four or more. A "full" acquisition before contrast administration is followed by truncated acquisitions sensitive primarily to changes in image contrast. Simulations of the point-spread functions that obtain, and their effect on contrast and spatial resolution, reveal significant degradation only for the smallest objects. Our simulations also address the feasibility of three-dimensional keyhold imaging, and demonstrate a potential 16-fold increase in temporal resolution. This suggests roles for keyhole imaging in conventional (nondynamic) precontrast and postcontrast studies and other applications.

Modeling of analog film-file radiographic retrievals. A Markov chain.

RATIONALE AND OBJECTIVES: Existing retrieval models for radiology film libraries have not incorporated the influence of previous retrievals. A Markov chain model for the retrieval rates from an analog film library is proposed as a means of considering this effect. METHODS: A Markov chain model was developed for the retrieval rates of an analog film library. The Markov chain model required identification of the states of the Markov chain, the required one-step transition probabilities between states, and the initial state probabilities. RESULTS AND CONCLUSIONS: The results from the Markov chain model compared favorably with the 30-day measurements (25,775 retrievals), but a large enough sample to determine a statistical confidence level was not considered.

Comparison between retrospective gating and ECG triggering in magnetic resonance velocity mapping.

ECG-triggered cinematographic studies of the cardiovascular system are hampered by several technical restrictions such as the inability to image end-diastole, ghosting, varying signal intensity, and phase contributions from eddy currents. Retrospective gating may solve these problems, but involves signal manipulation such as interpolating raw data from a time window. In this study, the performance of the two gating strategies was compared in quantitative MR velocity mapping on the abdominal aorta in eight healthy volunteers and on a pulsatile flow phantom. The results were compared to a one-dimensional velocity mapping technique and Doppler ultrasound. Finally, the consequence of decreasing the time window in the raw data interpolation used for retrospective gating was also examined. With retrospective gating, a low-pass filtering was seen, causing significantly prolonged duration and decreased amplitude of flow pulses. However, by reducing the time window retrospectively gated flow measurements were in good agreement with those that are ECG triggered. When fulfilling the demand of a narrow time window for interpolation, retrospective gating offers several advantages in MR velocity mapping.

T2-weighted three-dimensional MP-RAGE MR imaging.

The application of three-dimensional (3D) magnetization-prepared rapid-gradient-echo (MP-RAGE) imaging to the acquisition of T2-weighted 3D data sets has been investigated, with a 90 degrees x-180 degrees y-90 degrees-x pulse set (driven equilibrium) for the T2 contrast preparation. A theoretical model was used to study the contrast behavior of brain tissue. The effects of radio-frequency and static-field inhomogeneities and eddy currents on the T2 contrast preparation and the effects of eddy currents on the gradient-echo acquisition resulted in blurring and intensity banding artifacts. With a multistep gradient preparation, these artifacts could be suppressed. With further development, this technique may yield a clinically practical method for obtaining T2-weighted 3D data sets of relatively large volumes (eg, the whole head) suitable for multiplanar reformatting.

Electrocardiograph-independent, "wireless" cardiovascular cine MR imaging.

A new electrocardiograph (ECG)-independent, "wireless" gating technique for cine magnetic resonance (MR) imaging was evaluated in 23 cases of cardiovascular disease; in each case, standard ECG-dependent image loops were available for comparison. The ECG-independent strategy references cine MR imaging data retrospectively to inherent periodic changes in MR signal related to the cardiac cycle. With a "double-section" method, both timing data reflecting such changes and imaging data can be acquired simultaneously. "Artificial R waves" are extracted from the timing data acquired with a projection approach. The ECG-independent image loops were diagnostic in 91% of cases. Their overall image quality was at least equal to that of available ECG-dependent versions in only 39% of cases, but this proportion increased to 53% if cases with suboptimal imaging orientations for monitoring of the motion-dependent signal changes were excluded. Orientation appeared to be the primary technical limitation associated with this ECG-independent technique; however, poor ventricular function also significantly impaired performance. Improvement in the performance of the ECG-independent strategy is anticipated with technical advances.

Wireless retrospective gating: application to cine cardiac imaging.

A new "wireless" method of cardiac imaging is introduced, which, unlike ECG triggering, allows imaging the heart at end-diastole, and greatly reduces smearing artifacts in the phase-encoding direction. It is an improvement over ECG-driven retrospective gating, in that patients with poor ECGs can be imaged. This method extends the applicability of cardiac imaging, and since it requires no physiological monitoring hardware, can be implemented easily on any MR imager. The images produced by this method are superior to those from ECG triggering, especially when viewed in a "cine" loop. The technique described herein is, furthermore, extendable to any area where periodic or quasi-periodic motion is a problem.

Magnetic resonance imaging-based brain morphometry: development and application to normal subjects.

A semiautomated computerized method of in vivo morphometric analysis that is based upon high-resolution three-dimensional magnetic resonance imaging has been developed. This morphometric method is efficient and is of greater analytical precision than any other morphometric method currently applied to living human tissue. Including error inherent in image data acquisition, the aggregate error of the methodology, as estimated by the phantom studies, ranges from 4.5 to 9.6%, with incremental error above 4.5% a function of magnetic resonance slice thickness. This method was applied to magnetic resonance scans of 7 normal volunteers. The derived volumes of whole brain and of individual substructures were closely concordant with previously published volumes of normal fresh (unfixed) brains obtained post mortem. This morphometric methodology is potentially applicable to any structure or lesion that can be visualized by magnetic resonance imaging.