Search

Radiology reports often contain findings about the condition of a patient which should be acted upon quickly. These actionable findings in a radiology report can be automatically detected to ensure that the referring physician is notified about such findings and to provide feedback to the radiologist that further action has been taken. In this paper we investigate a method for detecting actionable findings of appendicitis in radiology reports. The method identifies both individual assertions regarding the presence of appendicitis and other findings related to appendicitis using syntactic dependency patterns. All relevant individual statements from a report are collectively considered to determine whether the report is consistent with appendicitis. Evaluation on a corpus of 400 radiology reports annotated by two expert radiologists showed that our approach achieves a precision of 91%, a recall of 83%, and an F1-measure of 87%.

EMRs are essential in modern healthcare. At the same time there has been the rapid development of clinical research systems (CRS) which offer specific tools to facilitate research protocol development, execution and management. However, examples of integrating a commercial EMR and CRS are rare. We describe our experience in the analysis, resolution of issues, design, and implementation of EMR/CR integration in a large academic medical center with affiliated hospitals. Our initial results suggest that (1) neither the EMR nor CRS system alone will suffice to facilitate clinical research, (2) there are important benefits to integration of the EMR and CRS systems and (3) this integration can be leveraged across multiple institutions.

Purpose: To improve timely evaluation and management of newly identified, suspected, gastrointestinal (GI) malignancies discovered on radiologic imaging at a safety-net hospital through direct critical results messaging to surgical specialists., Materials and Methods: To address delays in evaluating patients for suspected GI malignancies identified on imaging, an enhanced workflow was created--electronically routed critical results messaging to the ordering provider was supplemented with parallel messaging to the surgical oncology clinic. Messaging data obtained for 10 months pre and post intervention were compared. Using chart reviews, time intervals were recorded to assess the impact on (1) being seen by a specialist, (2) completing a diagnostic workup and (3) initiating definitive management., Results: Significant improvements were achieved: (1) patients seen by a specialist increased from 45.9% to 98.0% (p<0.001), with median time decreasing from 35 to 7 days (p<0.001); (2) patients completing a diagnostic workup increased from 77.1% to 93.9% (p<0.05), with median time decreasing from 44 to 18 days (p<0.001); (3) patients with initiation of definitive management increased from 72.1% to 89.8% (p<0.05), with median time decreasing from 62 to 35 days (p<0.05). Further study is needed to assess impact on fragmentation of care and financial implications., Conclusions: Direct critical results messaging from the radiologist to the surgical oncologist at a safety-net hospital significantly improves the time to complete a diagnostic workup and initiate definitive management with significantly more patients being seen by a relevant specialist.

Aims: Left atrial (LA) structural and functional abnormalities may be subclinical phenotypes, which identify individuals at increased risk of adverse outcomes., Methods and Results: Maximum LA volume (LAmax) and LA emptying fraction (LAEF) were measured via cardiac magnetic resonance imaging in 1802 participants in the Dallas Heart Study. The associations of LAEF and LAmax indexed to body surface area (LAmax/BSA) with traditional risk factors, natriuretic peptide levels, and left ventricular (LV) structure [end-diastolic volume (EDV) and concentricity(0.67) (mass/EDV(0.67))] and function (ejection fraction) were assessed using linear regression analysis. The incremental prognostic value of LAmax/BSA and LAEF beyond traditional risk factors, LV ejection fraction, and LV mass was assessed using the Cox proportional-hazards model. Both increasing LAmax/BSA and decreasing LAEF were associated with hypertension and natriuretic peptide levels (P < 0.05 for all). In multivariable analysis, LAmax/BSA was most strongly associated with LV end-diastolic volume/BSA, while LAEF was strongly associated with LV ejection fraction and concentricity(0.67). During a median follow-up period of 8.1 years, there were 81 total deaths. Decreasing LAEF [hazard ratio (HR) per 1 standard deviation (SD) (8.0%): 1.56 (1.32-1.87)] but not increasing LAmax/BSA [HR per 1 SD (8.6 mL/m(2)): 1.14 (0.97-1.34)] was independently associated with mortality. Furthermore, the addition of LAEF to a model adjusting Framingham risk score, diabetes, race, LV mass, and ejection fraction improved the c-statistic (c-statistics: 0.78 vs. 0.77; P < 0.05, respectively), whereas the addition of LAmax/BSA did not (c-statistics: 0.76, P = 0.20)., Conclusion: In the general population, both LAmax/BSA and LAEF are important subclinical phenotypes but LAEF is superior and incremental to LAmax/BSA.

Purpose: To develop and validate an algorithm to automatically quantify white matter hyperintensity (WMH) volume., Materials and Methods: Images acquired as part of the Dallas Heart Study, a multiethnic, population-based study of cardiovascular health, were used to develop and validate the algorithm. 3D magnetization prepared rapid acquisition gradient echo (MP-RAGE) and 2D fluid-attenuated inversion recovery (FLAIR) images were acquired from 2082 participants. Images from 161 participants (7.7% of the cohort) were used to set an intensity threshold to maximize the agreement between the algorithm and a qualitative rating made by a radiologist. The resulting algorithm was run on the entire cohort and outlier analyses were used to refine the WMH volume measurement. The refined, automatic WMH burden estimate was then compared to manual quantitative measurements of WMH volume in 28 participants distributed across the range of volumes seen in the entire cohort., Results: The algorithm showed good agreement with the volumetric readings of a trained analyst: the Spearman's Rank Order Correlation coefficient was r = 0.87. Linear regression analysis showed a good correlation WMHml[automated] = 1.02 × WMHml[manual] - 0.48. Bland-Altman analysis showed a bias of 0.34 mL and a standard deviation of 2.8 mL over a range of 0.13 to 41 mL., Conclusion: We have developed an algorithm that automatically estimates the volume of WMH burden using an MP-RAGE and a FLAIR image. This provides a tool for evaluating the WMH burden of large populations to investigate the relationship between WMH burden and other health factors., (Copyright © 2012 Wiley Periodicals, Inc.)

Recognizing the anatomical location of actionable findings in radiology reports is an important part of the communication of critical test results between caregivers. One of the difficulties of identifying anatomical locations of actionable findings stems from the fact that anatomical locations are not always stated in a simple, easy to identify manner. Natural language processing techniques are capable of recognizing the relevant anatomical location by processing a diverse set of lexical and syntactic contexts that correspond to the various ways that radiologists represent spatial relations. We report a precision of 86.2%, recall of 85.9%, and F(1)-measure of 86.0 for extracting the anatomical site of an actionable finding. Additionally, we report a precision of 73.8%, recall of 69.8%, and F(1)-measure of 71.8 for extracting an additional anatomical site that grounds underspecified locations. This demonstrates promising results for identifying locations, while error analysis reveals challenges under certain contexts. Future work will focus on incorporating new forms of medical language processing to improve performance and transitioning our method to new types of clinical data.

Objectives: Electron beam computed tomography (EBCT) for coronary artery calcification can potentially evaluate liver fat, another marker of cardiovascular risk. We compared quantitative estimates of hepatic steatosis measured by EBCT with those obtained by a well-validated, accurate-measure, magnetic resonance spectroscopy ((1)H MRS)., Methods: EBCT and (1)H MRS were performed in 2159 subjects from the Dallas Heart Study. Forty subjects were randomly selected from each of 5 subgroups of liver fat percent by (1)H MRS (n = 200). EBCT average liver attenuation (HU) was determined in a 1- to 2-cm circular region of interest over the liver lobes. Pearson correlation coefficients were calculated. Using a previously defined (1)H MRS hepatic steatosis cut point (>5.5%), an optimized EBCT liver attenuation cut point was determined by receiver operating characteristic analysis., Results: (1)H MRS liver fat content and EBCT average right lobe liver attenuation were moderately negatively correlated (r = -0.64, P < 0.0001) in all subjects and in those with (1)H MRS hepatic steatosis (r = -0.71, P < 0.0001). This correlation did not improve with attenuation correction of the EBCT data using a standard calcium phantom or statistical transformation. Using an optimized receiver operating characteristic EBCT cut point (64.5 HU), sensitivity was 78% and specificity was 72% for detecting (1)H MRS hepatic steatosis, with a high false negative rate. Risk factors for hepatic steatosis (obesity, diabetes mellitus, insulin resistance, metabolic syndrome) were more strongly correlated with (1)H MRS than EBCT liver fat measures., Conclusions: Liver attenuation on EBCT acquired for coronary artery calcification screening correlates modestly with (1)H MRS measures of liver fat content, with a high false negative rate.

Background: Reference values and age-related changes of the wall thickness of the abdominal aorta have not been described in the general population. We characterized age-, race-, and gender-specific distributions, and yearly rates of change of mean aortic wall thickness (MAWT), and associations between MAWT and cardiovascular risk factors in a multi-ethnic population-based probability sample., Methods: Magnetic resonance imaging measurements of MAWT were performed on 2466 free-living white, black, and Hispanic adult subjects. MAWT race/ethnicity- and gender-specific percentile values across age were estimated using regression analyses., Results: MAWT was greater in men than in women and increased linearly with age in all the groups and across all the percentiles. Hispanic women had the thinnest and black men the thickest aortas. Black men had the highest and white women the lowest age-related MAWT increase. Age, gender, ethnicity, smoking status, systolic blood pressure, low-density lipoprotein-cholesterol levels, high-density lipoprotein-cholesterol levels, and fasting glucose levels were independent predictors of MAWT., Conclusions: Age, gender, and racial/ethnic differences in MAWT distributions exist in the general population. Such differences should be considered in future investigations assessing aortic atherosclerosis and the effects of anti-atherosclerotic therapies., (Published by Mosby, Inc.)

Background: The Health ABC Heart Failure score has recently been shown to predict 5-year risk of incident heart failure in the elderly. We tested whether this risk score is associated with subclinical phenotypes of heart failure in a younger population., Methods: We stratified participants in the Dallas Heart Study aged 30 to 65 years who had a cardiac magnetic resonance imaging and no self-reported history of heart failure or cardiomyopathy into 4 previously defined Health ABC Heart Failure risk groups: low (<5%), average (5%-10%), high (10%-20%), and very high (>20% risk for heart failure within 5 years). We compared left ventricular (LV) structural and functional parameters and levels of B-type natriuretic peptide (BNP) and N-terminal proBNP among the 4 groups., Results: In the study cohort (N = 2,540), the percentage of subjects in the low-, average-, high-, and very high risk groups was 78%, 15%, 6%, and 1%, respectively. Indexed LV mass (80 +/- 15 vs 90 +/- 20 vs 95 +/- 25 vs 116 +/- 41 g/m(2)), concentricity (1.6 +/- 0.3 vs 1.8 +/- 0.4 vs 2.0 +/- 0.5 vs 2.2 +/- 0.7 g/mL), median BNP (2.8 vs 3.7 vs 4.9 vs 7.5 pg/mL) and N-terminal proBNP (26 vs 30 vs 40 vs 58 pg/mL), and prevalent LV systolic dysfunction and LV hypertrophy progressively increased across risk groups (P < .001 for all) independent of gender or method of indexing LV mass., Conclusions: The Health ABC Heart Failure score was associated with subclinical cardiac structural changes in the general population 30 to 65 years of age, suggesting that it may be a valid tool for identification of young individuals at increased risk for heart failure., (2010 Mosby, Inc. All rights reserved.)

Purpose: To assess the agreement between three methods of calculation of mean aortic wall thickness (MAWT) using magnetic resonance imaging (MRI)., Materials and Methods: High-resolution MRI of the infrarenal abdominal aorta was performed on 70 subjects with a history of coronary artery disease who were part of a multi-ethnic population-based sample. MAWT was calculated as the mean distance between the adventitial and luminal aortic boundaries using three different methods: average distance at four standard positions (AWT-4P), average distance at 100 automated positions (AWT-100P), and using a mathematical computation derived from the total vessel and luminal areas (AWT-VA). Bland-Altman plots and Passing-Bablok regression analyses were used to assess agreement between methods., Results: Bland-Altman analyses demonstrated a positive bias of 3.02+/-7.31% between the AWT-VA and the AWT-4P methods, and of 1.76+/-6.82% between the AWT-100P and the AWT-4P methods. Passing-Bablok regression analyses demonstrated constant bias between the AWT-4P method and the other two methods. Proportional bias was, however, not evident among the three methods., Conclusion: MRI methods of measurement of MAWT using a limited number of positions of the aortic wall systematically underestimate the MAWT value compared with the method that calculates MAWT from the vessel areas., (Copyright (c) 2009 Wiley-Liss, Inc.)

Background: Caspase-3, an apoptosis protease, is expressed in atherosclerotic plaques. We examined the relationship between plasma caspase-3 levels, aortic compliance, and atherosclerosis., Methods: Caspase-3 was measured in 3,221 subjects from the Dallas Heart Study. Electron beam computed tomography measures of coronary calcium (CAC) (n = 2,404) and magnetic resonance imaging (MRI) measures of abdominal aortic wall thickness (AWT) (n = 2,208) and aortic compliance (AC) (n = 2,328) were obtained. Multivariate analyses were performed, adjusting for age, sex, ethnicity, body mass index (BMI), traditional cardiovascular risk factors, and cardiac medications., Results: In univariable analysis, caspase-3 associated with CAC (P < 0.0001), AWT (P = 0.002), and AC (P < 0.0001). After multivariable adjustment, 4th quartile caspase-3 (compared to 1st quartile) was significantly associated with CAC (P = 0.004), AWT (P = 0.02), and AC (P < 0.0001) with similar findings for caspase-3 as a continuous variable., Conclusions: Caspase-3 independently associates with CAC, AWT, and AC, suggesting a link between apoptosis and atherosclerosis.

Purpose: To study the effects of field strength and parallel imaging on image contrast and interstudy reproducibility of right and left ventricular (RV and LV) measurements using steady-state free precession (SSFP) cardiovascular magnetic resonance (CMR)., Materials and Methods: Thirty-two subjects (20 normal, 12 cardiac patients) underwent four SSFP cine short-axis imaging studies: two at 1.5T, one at 3T, and another at 3T with parallel imaging (SENSE). RV and LV contrast-to-noise ratios (CNRs) were compared between methods. Interstudy reproducibility of RV and LV measurements were assessed by Bland-Altman analyses., Results: 3T SENSE imaging reduced total imaging time from 8 +/- 2 minutes to 3 +/- 1 minute (P < 0.001). A significant gain in LV CNR was detected between 1.5T and 3T with SENSE (43.8 +/- 6.5 vs. 48.4 +/- 7.4, P = 0.01), but no significant gain was detected in RV CNR. The reproducibility of LV and RV measurements between two 1.5T studies was not significantly different from the reproducibility between a 1.5T study and a 3T study with SENSE., Conclusion: SENSE imaging at 3T is a reproducible technique for assessing RV and LV structure and function.

Background: The objective of the study was to evaluate racial differences in the prevalence of left ventricular (LV) dysfunction. Few data compare the relative frequency of reduced LV ejection fraction (EF) (LVEF) in blacks and whites. Because of the higher prevalence of risk factors for heart failure in blacks, including hypertension, obesity, and LV hypertrophy, we hypothesized that LV dysfunction would also be more common in this ethnic group., Methods: In the DHS, a probability-based sample of Dallas County, we performed cardiac magnetic resonance imaging on 1335 black and 858 white participants aged 30 to 67 years to measure LVEF and volumes. We compared the prevalence of reduced LV EF and distribution of ventricular volumes in the 2 ethnic groups., Results: The prevalence of a reduced LVEF, whether defined as < 50%, < 55%, or < 60%, did not appear to be different between black versus white women (P > or = .7 for each) or men (P > or = .4 for each). Similar findings were seen using a recently defined sex-specific threshold (men < 55% and women < 61%) for low EF (P = .1). Mean LV end-diastolic and end-systolic volumes indexed to body surface area were also comparable in black and white men (P > or = .3) and women (P > or = .1)., Conclusions: Despite having a higher prevalence of risk factors for heart failure, blacks as compared with whites did not have a higher prevalence of reduced LVEF in the general population.

In order for the conductance catheter system to accurately measure instantaneous cardiac blood volume, it is necessary to determine and remove the contribution from parallel myocardial tissue. In previous studies, the myocardium has been treated as either purely resistive or purely capacitive when developing methods to estimate the myocardial contribution. We propose that both the capacitive and the resistive properties of the myocardium are substantial, and neither should be ignored. Hence, the measured result should be labeled admittance rather than conductance. We have measured the admittance (magnitude and phase angle) of the left ventricle in the mouse, and have shown that it is measurable and increases with frequency. Further, this more accurate technique suggests that the myocardial contribution to measured admittance varies between end-systole and end-diastole, contrary to previous literature. We have tested these hypotheses both with numerical finite-element models for a mouse left ventricle constructed from magnetic resonance imaging images, and with in vivo admittance measurements in the murine left ventricle. Finally, we propose a new method to determine the instantaneous myocardial contribution to the measured left ventricular admittance that does not require saline injection or other intervention to calibrate.

The hypothesis that chronic kidney disease (CKD) is associated with increased coronary artery calcification (CAC) was tested using data from the Dallas Heart Study, a representative sample of Dallas County residents aged 30 to 65 yr. CKD was defined as presence of microalbuminuria and GFR > or =60 ml/min per 1.73 m(2) (stage 1 to 2), or GFR <60 ml/min per 1.73 m(2) (stage 3 to 5), excluding end-stage kidney disease. Logistic regression was used to examine the association between stages of CKD and CAC scores >10, >100, and >400 versus scores < or =10 compared with no CKD while adjusting for covariates. Analyses were repeated after stratifying by presence of diabetes. The mean age was 43.9 yr, and hypertension and diabetes were noted in 31.0 and 9.8%, respectively. No association was noted between stage 1 to 2 CKD and increased CAC scores. Compared with no CKD, stage 3 to 5 CKD was associated with CAC scores >100 (odds ratio, 2.85; 95% confidence interval, 0.92 to 8.80) and >400 (odds ratio, 8.35; 95% confidence interval, 1.94 to 35.95) in the total population after adjustment for covariates, but these associations were substantially reduced after exclusion of participants with diabetes. Participants with diabetes and stage 3 to 5 CKD had a ninefold increased odds of CAC scores >10 versus scores < or =10 compared with participants with diabetes and without CKD, whereas no association was noted between stage 3 to 5 CKD and CAC scores >10 in the nondiabetic population. In conclusion, stage 3 to 5 CKD is associated with increased CAC scores, but this association may be substantially stronger among adults with diabetes. These findings need to be confirmed in study populations that include adults >65 yr of age and a larger number of CKD cases.

Although cardiac myocyte apoptosis has been detected in explanted hearts from patients with end-stage dilated and ischemic cardiomyopathy, the relative contribution of apoptotic cell death to left ventricular (LV) remodeling and cardiac decompensation is not known. To determine whether progressive cardiac myocyte apoptosis contributes to the transition from a hypertrophic to a dilated cardiac phenotype that is observed in transgenic myosin heavy chain secreted TNF (MHCsTNF) mice with cardiac restricted overexpression of tumor necrosis factor (TNF), we assessed cardiac myocyte apoptosis (using a DNA ligase technique) in MHCsTNF mice and littermate control mice in relation to serial changes in LV structure, which was assessed using MRI. The prevalence of cardiac myocyte apoptosis increased progressively from 4 to 12 wk as the hearts of the MHCsTNF mice underwent the transition from a concentric hypertrophic to a dilated cardiac phenotype. Treatment of the MHCsTNF mice with the broad-based caspase inhibitor N-[(1,3-dimethylindole-2-carbonyl)-valinyl]-3-amino4-oxo-5-fluoropentanoic acid significantly decreased cardiac myocyte apoptosis and significantly attenuated LV wall thinning and adverse cardiac remodeling. Additional studies suggested that the TNF-induced decrease in Bcl-2 expression and activation of the intrinsic mitochondrial death pathway were responsible for the cardiac myocyte apoptosis observed in the MHCsTNF mice. These studies show that progressive cardiac myocyte apoptosis is sufficient to contribute to adverse cardiac remodeling in the adult mammalian heart through progressive LV wall thinning.

Objectives: We sought to compare the prevalence of coronary atherosclerosis in a cohort of middle-age African American (black) and non-Hispanic Caucasian (white) men and women from a population-based probability sample., Background: Blacks have a higher mortality from coronary heart disease (CHD) than whites, particularly among younger individuals, and yet several studies have reported that coronary atherosclerosis is less prevalent in blacks than in whites. Data from population-based samples comparing coronary atherosclerotic burden between blacks and whites are limited., Methods: The prevalence of coronary atherosclerosis in middle-aged blacks and whites was determined using coronary calcium measured by electron beam computed tomography in 1,289 men and women from a population-based probability sample from Dallas, Texas., Results: The population estimates of the frequency of a positive scan for coronary artery calcium were not statistically different between black and white men (37% vs. 41%, p = 0.36) or between black and white women (29% vs. 23%, p = 0.21). Although the prevalence of most of the coronary risk factors varied significantly between blacks and whites, mean Framingham coronary risk factor scores were identical in black and white men (10 +/- 4) but significantly higher in black women (13 +/- 4) than in white women (12 +/- 4)., Conclusions: Blacks in the general population have a prevalence of coronary atherosclerosis similar to whites. Factors other than coronary atherosclerotic burden, which are not reflected in the Framingham risk score, contribute significantly to the higher CHD mortality rate in blacks.

Low plasma levels of SHBG and free testosterone have been associated with increased insulin resistance and risk for type 2 diabetes in males. As truncal obesity, a condition accompanied by increased insulin resistance, is also associated with low SHBG and testosterone levels, the independent association of low free testosterone and SHBG with excessive insulin resistance remains to be determined. In this study we evaluated whether in normogonadic men, plasma levels of SHBG and free testosterone are primarily related to insulin resistance or to generalized and regional adiposity. Hyperinsulinemic-euglycemic clamps and iv glucose tolerance tests were performed in 24 healthy volunteer and 33 patients with mild type 2 diabetes. The 2 groups were chosen to have similar body mass index and were found to have similar body composition and fat distribution, assessed by underwater weighing, skinfold thickness, and magnetic resonance imaging of the abdomen. In the 2 groups combined, plasma levels of SHBG correlated inversely with fat accumulation in both sc and intraabdominal areas. Plasma levels of free testosterone correlated inversely with both truncal and peripheral skinfold thickness only in the nondiabetic men. No associations between plasma levels of sex steroid hormones and insulin resistance, hepatic glucose output, or insulin secretion were found to be independent of adiposity. Furthermore, although patients with diabetes were more insulin resistant than those without diabetes, the 2 groups had similar plasma concentrations of free testosterone (55 +/- 14 and 67 +/- 27 pmol/liter, respectively), SHBG (19 +/- 13 and 19 +/- 13 nmol/liter), estradiol (83 +/- 5 and 81 +/- 21 pmol/liter), and dehydroepiandrosterone sulfate (3.6 +/- 2.2 and 2.8 +/- 1.7 nmol/liter). We conclude that in normogonadal nondiabetic males, the variability in plasma bioavailable testosterone concentrations is predictive of the variability in fat deposition in the sc adipose tissue compartments of both truncal and peripheral areas. Low plasma levels of bioavailable testosterone do not independently predict excessive insulin resistance, beta-cell dysfunction, or hepatic glucose output in normogonadal men.