An Adaptive Low-Rank Modeling-Based Active Learning Method for Medical Image Annotation.

Active learning is an effective solution to interactively select a limited number of informative examples and use them to train a learning algorithm that can achieve its optimal performance for specific tasks. It is suitable for medical image applications in which unlabeled data are abundant but manual annotation could be very time-consuming and expensive. However, designing an effective active learning strategy for informative example selection is a challenging task, due to the intrinsic presence of noise in medical images, the large number of images, and the variety of imaging modalities. In this study, a novel low-rank modeling-based multi-label active learning (LRMMAL) method is developed to address these challenges and select informative examples for training a classifier to achieve the optimal performance. The proposed method independently quantifies image noise and integrates it with other measures to guide a pool-based sampling process to determine the most informative examples for training a classifier. In addition, an automatic adaptive cross entropy-based parameter determination scheme is proposed for further optimizing the example sampling strategy. Experimental results on varied medical image datasets and comparisons with other state-of-the-art multi-label active learning methods illustrate the superior performance of the proposed method.

Physical activity predicts gray matter volume in late adulthood: the Cardiovascular Health Study.

OBJECTIVES: Physical activity (PA) has been hypothesized to spare gray matter volume in late adulthood, but longitudinal data testing an association has been lacking. Here we tested whether PA would be associated with greater gray matter volume after a 9-year follow-up, a threshold could be identified for the amount of walking necessary to spare gray matter volume, and greater gray matter volume associated with PA would be associated with a reduced risk for cognitive impairment 13 years after the PA evaluation. METHODS: In 299 adults (mean age 78 years) from the Cardiovascular Health Cognition Study, we examined the association between gray matter volume, PA, and cognitive impairment. Physical activity was quantified as the number of blocks walked over 1 week. High-resolution brain scans were acquired 9 years after the PA assessment on cognitively normal adults. White matter hyperintensities, ventricular grade, and other health variables at baseline were used as covariates. Clinical adjudication for cognitive impairment occurred 13 years after baseline. RESULTS: Walking amounts ranged from 0 to 300 blocks (mean 56.3; SD 69.7). Greater PA predicted greater volumes of frontal, occipital, entorhinal, and hippocampal regions 9 years later. Walking 72 blocks was necessary to detect increased gray matter volume but walking more than 72 blocks did not spare additional volume. Greater gray matter volume with PA reduced the risk for cognitive impairment 2-fold. CONCLUSION: Greater amounts of walking are associated with greater gray matter volume, which is in turn associated with a reduced risk of cognitive impairment.

Initial experience in using continuous arterial spin-labeled MR imaging for early detection of Alzheimer disease.

BACKGROUND AND PURPOSE: MR imaging of the brain has significant potential in the early detection of neurodegenerative disorders such as AD. The purpose of this work was to determine if perfusion MR imaging can be used to separate AD from normal cognition in individual subjects. We investigated the diagnostic utility of perfusion MR imaging for early detection of AD compared with structural imaging. MATERIALS AND METHODS: Data were analyzed from 32 participants in the institutional review board-approved CHS-CS: 19 cognitively healthy individuals and 13 with clinically adjudicated AD. All subjects underwent structural T1-weighted SGPR and CASL MR imaging. Four readers with varying experience separately rated each CASL and SPGR scan finding as normal or abnormal on the basis of standardized qualitative diagnostic criteria for observed perfusion abnormalities on CASL or volume loss on SPGR and rated the confidence in their evaluation. RESULTS: Inter-rater reliability was superior in CASL (kappa = 0.7 in experienced readers) compared with SPGR (kappa = 0.17). CASL MR imaging had the highest sensitivity (85%) and accuracy (70%). Frontal lobe CASL findings increased sensitivity to 88% and accuracy to 79%. Fifty-seven percent of false-positive readings with CASL were in controls with cognitive decline or instability within 5 years. Three of the 4 readers revealed a statistically significant relationship between confidence and correct classification when using CASL. CONCLUSIONS: Readers were able to separate individuals with mild AD from those with normal cognition with high sensitivity by using CASL but not volumetric MR imaging. This initial experience suggests that CASL MR imaging may be a useful technique for detecting AD.

Plasma amyloid levels and the risk of AD in normal subjects in the Cardiovascular Health Study.

OBJECTIVES: To examine the association between incident Alzheimer disease (AD), and plasma A beta 1-40 and A beta 1-42 levels in normal and mild cognitive impairment (MCI) subjects in a subgroup of participants of the Cardiovascular Health Study Cognition Study. METHODS: We determined the plasma A beta 1-40 and A beta 1-42 levels of 274 nondemented subjects (232 normals and 42 with MCI) in 1998-1999 and repeated the measurements in 2002-2003. The mean age of the subjects at baseline was 79.3 +/- 3.6 years. We examined the association between A beta levels and incident AD over the ensuing 4.5 years, controlling for age, cystatin C level (marker of glomerular function), apolipoprotein E-4 allele, Modified-Mini-Mental State Examination scores, and MRI-identified infarcts. RESULTS: In an unadjusted prospective model in normal subjects, both A beta 1-40 and A beta 1-42 levels in 1998-1999 were associated with incident AD (n = 55) in 2002-2003 (longitudinal analysis). In the fully adjusted multivariate model, neither A beta 1-42 nor A beta 1-40 nor their ratio was associated with incident AD. However, adjustment had a very small effect on point estimates for A beta 1-42, from an odds ratio (OR) of 1.61 (p = 0.007) in the unadjusted model to an OR of 1.46 (p = 0.08) in the fully adjusted model. In 2002-2003 (cross-sectional analysis), only the unadjusted models showed that both peptides were associated with AD. CONCLUSIONS: Plasma A beta levels are affected by age and by systemic and CNS vascular risk factors. After controlling for these conditions, A beta-40 and A beta 1-42 are weak predictors of conversion to Alzheimer disease (AD) in normal subjects and are only weakly associated with AD in cross-sectional analysis. Consequently, plasma levels of A beta do not seem to be useful biomarkers for AD.

Measuring flow reattachment lengths downstream of a stenosis using MRI.

Flow reattachment lengths (l(r)) are measured downstream of an abrupt axisymmetric 75% stenosis, located inside a cylindrical channel, for steady flow using ultra-fast magnetic resonance imaging (MRI). The MRI results are compared with those from other similar (non-MRI) studies. The MRI data confirm the existence of three flow reattachment regimes (laminar, fully turbulent, and transition) related to the flow Reynolds number (Re) measured inside the stenosis. Based on the MRI experiments, the laminar regime occurs at a stenotic Reynolds number below 250 with a slope (l(r)/Re) of 0.086. The fully developed turbulence occurs at a stenotic Reynolds number above 3600 with a minimum observed reattachment length of 5 step heights. The transition regime (occurring between the laminar and fully turbulent regimes) is characterized by a reattachment length plateau and then a drop with Re(-1.1). J. Magn. Reson. Imaging 2000;12:939-948.

Imaging obstructed ventilation with NMR using inert fluorinated gases.

We partially obstructed the left bronchi of rats and imaged an inert insoluble gas, SF(6), in the lungs with NMR using a technique that clearly differentiates obstructed and normal ventilation. When the inhaled fraction of O(2) is high, SF(6) concentrates dramatically in regions of the lung with low ventilation-to-perfusion ratios (VA/Q); therefore, these regions are brighter in an image than where VA/Q values are normal or high. A second image, made when the inhaled fraction of O(2) is low, serves as a reference because the SF(6) fraction is nearly uniform, regardless of VA/Q. The quotient of the first and second images displays the low-VA/Q regions and is corrected for other causes of brightness variation. The technique may provide sufficient quantification of VA/Q to be a useful research tool. The noise in the quotient image is described by the probability density function for the quotient of two normal random variables. When the signal-to-noise ratio of the denominator image is >10, the signal-to-noise ratio of the quotient image is similar to that of the parent images and decreases with pixel value.

Transforming NMR data despite missing points.

Some NMR experiments produce data with several of the initial points missing. The inverse discrete Fourier transform (IDFT) assumes these points are present so the data cannot be so transformed without artifact-ridden results. This problem is often particularly severe when projection imaging with free-induction decays (FIDs). This paper compares recent methods for obtaining a projection from incomplete data and elaborates on their strengths and limitations. One method is to write the transform that would take the desired projection to the truncated data set, and then solve the matrix equation by singular value decomposition. A second replaces the missing data with zeros, so that an IDFT produces a projection with unwanted artifacts. Then one solves the matrix equation that takes the desired projection to the artifact-ridden projection. A third uses the same artifact-ridden projection, but fits the region outside the bandwidth of the sample with as many sinusoidal functions as there are missing data. The coefficients of these functions are estimates of the missing data, and the projection is obtained by transforming the completed FID or subtracting the extrapolation of the fitted curve from the region containing the object. We show that when all three methods are applicable, they theoretically produce the same result. They differ by ease of implementation and possibly by computational errors. They give a result similar to that of the previous method that iteratively corrects the FID and projection after repeated IDFTs and DFTs. We find that one can obtain a projection despite missing a substantial number of data.

Observing curved flow using RUFIS.

The rotating ultra-fast imaging sequence (RUFIS) is used to image spins flowing through the curved portion of a 180 degrees U-tube with a circular cross section and a ratio of inner diameter to curvature diameter of 0.41. A velocity-encoding preparation sequence is used with RUFIS to measure quantitatively the axial velocity distribution for spins under steady-flow conditions. The resulting velocity contours confirm that higher velocity spins migrate toward the outer wall of the curve as the Dean number is increased.

Characterization of flow emerging from a stenosis using MRI.

MRI ultra-fast imaging techniques are used to characterize flow emerging from streamlined and abrupt stenoses inside cylindrical channels. Reattachment lengths of the shear boundary to the channel wall are measured using rotating ultra-fast imaging sequence (RUFIS) in-flow imaging. Velocity profiles of flow are created using velocity (sine and cosine)-encoded RUFIS sequences. The sine-encoded images permit one to identify reverse flow (i.e., eddies) that arise within the region of flow reattachment. The ratios of peak velocities (downstream/upstream of the stenosis) derived from the cosine-encoded images are used to identify the transition from the laminar to the turbulent regimen. Based on these experiments, the transition from the laminar to turbulent regimen occurs at a stenotic Reynolds Number of 350, whereas fully developed turbulence occurs at a stenotic Reynolds Number of 2600. These results are compared with the results from invasive studies.

A programmable pre-emphasis system.

MRI systems often use magnetic field gradient and shim pulse-shaping networks (pre-emphasis) to correct for magnetic field distortions caused by eddy currents. A pre-emphasis system that uses up to 16 fixed resistor-capacitor (RC) time constants per channel with programmable amplitude coefficients is described. The magnetic fields induced by the pre-emphasis RC time constants serve as a set of basis functions for compensating eddy-current fields induced by the gradient set. The resultant time-varying magnetic field gradient accurately reflects the gradient specified by the pulse programmer. Reductions in eddy-current fields are demonstrated for actively shielded and unshielded gradient sets.

Ultra-fast velocity imaging in stenotically produced turbulent jets using RUFIS.

A method for rapidly producing velocity images is presented. This sequence combines a modified bipolar gradient pulse to magnitude encode the velocity with the rotating ultra-fast imaging sequence (RUFIS) to image the encoded spins. Velocity encoding is done in 3 msec, and RUFIS acquires 32 projections in 8 msec. The method is applied to turbulent jets associated with a 75% stenosis in a 15-mm inner diameter glass pipe. Data is acquired upstream and downstream from the stenosis for Reynolds numbers from 560 to 3750. In addition, a robust method of reconstructing the unobserved short time region of a free induction decay is presented and incorporated into the image processing.