A Stacked Generalization of 3D Orthogonal Deep Learning Convolutional Neural Networks for Improved Detection of White Matter Hyperintensities in 3D FLAIR Images

BACKGROUND AND PURPOSE:Accurate and reliable detection of white matter hyperintensities and their volume quantification can provide valuable clinical information to assess neurologic disease progression. In this work, a stacked generalization ensemble of orthogonal 3D convolutional neural networks, StackGen-Net, is explored for improving automated detection of white matter hyperintensities in 3D T2-FLAIR images.MATERIALS AND METHODS:Individual convolutional neural networks in StackGen-Net were trained on 2.5D patches from orthogonal reformatting of 3D-FLAIR (n = 21) to yield white matter hyperintensity posteriors. A meta convolutional neural network was trained to learn the functional mapping from orthogonal white matter hyperintensity posteriors to the final white matter hyperintensity prediction. The impact of training data and architecture choices on white matter hyperintensity segmentation performance was systematically evaluated on a test cohort (n = 9). The segmentation performance of StackGen-Net was compared with state-of-the-art convolutional neural network techniques on an independent test cohort from the Alzheimer’s Disease Neuroimaging Initiative-3 (n = 20).RESULTS:StackGen-Net outperformed individual convolutional neural networks in the ensemble and their combination using averaging or majority voting. In a comparison with state-of-the-art white matter hyperintensity segmentation techniques, StackGen-Net achieved a significantly higher Dice score (0.76 [SD, 0.08], F1-lesion (0.74 [SD, 0.13]), and area under precision-recall curve (0.84 [SD, 0.09]), and the lowest absolute volume difference (13.3% [SD, 9.1%]). StackGen-Net performance in Dice scores (median = 0.74) did not significantly differ (P = .22) from interobserver (median = 0.73) variability between 2 experienced neuroradiologists. We found no significant difference (P = .15) in white matter hyperintensity lesion volumes from StackGen-Net predictions and ground truth annotations.CONCLUSIONS:A stacked generalization of convolutional neural networks, utilizing multiplanar lesion information using 2.5D spatial context, greatly improved the segmentation performance of StackGen-Net compared with traditional ensemble techniques and some state-of-the-art deep learning models for 3D-FLAIR.

White matter hyperintensities (WMHs) correspond to pathologic features of axonal degeneration, demyelination, and gliosis observed within cerebral white matter.¹ Clinically, the extent of WMHs in the brain has been associated with cognitive impairment, Alzheimer’s disease and vascular dementia, and increased risk of stroke.^2,3 The detection and quantification of WMH volumes to monitor lesion burden evolution and its correlation with clinical outcomes have been of interest in clinical research.^4,5 Although the extent of WMHs can be visually scored,⁶ the categoric nature of such scoring systems makes quantitative evaluation of disease progression difficult. Manually segmenting WMHs is tedious, prone to inter- and intraobserver variability, and is, in most cases, impractical. Thus, there is an increased interest in developing fast, accurate, and reliable computer-aided automated techniques for WMH segmentation.Convolutional neural network (CNN)-based approaches have been successful in several semantic segmentation tasks in medical imaging.⁷ Recent works have proposed using deep learning–based methods for segmenting WMHs using 2D-FLAIR images.^8-11 More recently, a WMH segmentation challenge¹² was also organized (http://wmh.isi.uu.nl/) to facilitate comparison of automated segmentation of WMHs of presumed vascular origin in 2D multislice T2-FLAIR images. Architectures that used an ensemble of separately trained CNNs showed promising results in this challenge, with 3 of the top 5 winners using ensemble-based techniques.¹²Conventional 2D-FLAIR images are typically acquired with thick slices (3–4 mm) and possible slice gaps. Partial volume effects from a thick slice are likely to affect the detection of smaller lesions, both in-plane and out-of-plane. 3D-FLAIR images, with isotropic resolution, have been shown to achieve higher resolution and contrast-to-noise ratio¹³ and have shown promising results in MS lesion detection using 3D CNNs.¹⁴ Additionally, the isotropic resolution enables viewing and evaluation of the images in multiple planes. This multiplanar reformatting of 3D-FLAIR without the use of interpolating kernels is only possible due to the isotropic nature of the acquisition. Network architectures that use information from the 3 orthogonal views have been explored in recent works for CNN-based segmentation of 3D MR imaging data.¹⁵ The use of data from multiple planes allows more spatial context during training without the computational burden associated with full 3D training.¹⁶ The use of 3 orthogonal views simultaneously mirrors how humans approach this segmentation task.Ensembles of CNNs have been shown to average away the variances in the solution and the choice of model- and configuration-specific behaviors of CNNs.¹⁷ Traditionally, the solutions from these separately trained CNNs are combined by averaging or using a majority consensus. In this work, we propose the use of a stacked generalization framework (StackGen-Net) for combining multiplanar lesion information from 3D CNN ensembles to improve the detection of WMH lesions in 3D-FLAIR. A stacked generalization¹⁸ framework learns to combine solutions from individual CNNs in the ensemble. We systematically evaluated the performance of this framework and compared it with traditional ensemble techniques, such as averaging or majority voting, and state-of-the-art deep learning techniques.     

Chronic synovitis and HLA B27 in patients with severe haemophilia

Chronic synovitis affects about 10% of patients with severe haemophilia in India. This disease has some features in common with ankylosing spondylitis, which has been linked to HLA B27. We therefore aimed to test whether there is an association between HLA B27 and chronic synovitis. We studied 473 patients with severe haemophilia (33 of whom had chronic synovitis), and 1175 healthy controls using a standard serological technique and the reverse line strip assay. 64% (21 of 33) of patients with haemophilia and chronic synovitis were positive for HLA B27, compared with 5% (23 of 440) of those with severe haemophilia, but not chronic synovitis (odds ratio 31.6 [95% CI 9.28-39.38], p<0.0001), and 9% (100 of 1175) of healthy controls (18.81 [9.6-27.7], p<0.0001). We conclude that there is a strong association between HLA B27 and chronic synovitis in Indian patients with severe haemophilia and screening in this population could allow treatment and prevention of the complication.     

Natural History After Acute Necrotizing Pancreatitis: a Large US Tertiary Care Experience

Background

Most studies of acute necrotizing pancreatitis (ANP) focus on short-term outcomes. We evaluated long-term survival and outcomes following ANP.

Methods

Patients treated for ANP at the University of Pittsburgh Medical Center from 2001 to 2008 were studied. Data on presentation and course during initial hospitalization and follow-up (median 34 months) was extracted.

Results

Mean age of patients (n?=?167) was 53?±?16 years; 70 % were male, 94 % white, 71 % transfers, 52 % biliary etiology, and 78 % had first-attack of acute pancreatitis. Majority had severe disease with high Acute Physiology and Chronic Health Evaluation II (APACHE-II) score (median 11), length of stay (median 26 days), intensive care unit (ICU) admission (87 %), presence of systemic inflammatory response syndrome (SIRS) (90 %), persistent organ failure (60 %), and infected necrosis (50 %). Intervention was needed in 74 %. Eighteen (10.8 %) patients died during index hospitalization, 9 (5.4 %) during the first year, and 13 (7.8 %) after 1 year. Median survival was significantly shorter when compared with age- and sex-matched US general population (9.1 vs. 26.1 years, p?<?0.001). Increasing age (HR 1.05), persistent organ failure (HR 4.5), and >50 % necrosis (HR 3.8) were independent predictors of death at 1 year. In eligible patients, new-onset diabetes, oral pancreatic enzyme replacement therapy, and disability were noted in 45, 25, and 53 %, respectively.

Conclusion

ANP significantly impacts long-term survival. A high proportion of patients develop functional derangement and disability following ANP.

     

Lithium and valproic acid treatments reduce PKC activation and receptor-G protein coupling in platelets of bipolar manic patients

Dysregulated protein kinase C (PKC) distribution and activation, and abnormal receptor-G protein coupling, have been implicated in the pathophysiology of bipolar affective disorder (BD). The therapeutic effectiveness of lithium has also been correlated with its ability to reduce PKC activation and G protein-mediated signaling. We examine the cellular distribution and activation of PKC and receptor-G protein coupling in blood platelets from normal controls, patients with BD mania or schizophrenia during treatment-free state, and after lithium or valproic acid administration. PKC activity was measured under basal and 50 nM phorbol 12-myristate, 13-acetate (PMA), 1 microM serotonin or 0.5 U/ml thrombin-stimulated conditions. The coupling of G proteins to serotonin or thrombin receptors were assessed by serotonin or thrombin-mediated [35S]GTPgammaS binding to membrane Galpha proteins. The results demonstrate that membrane-associated PKC activity and stimulus-induced PKC translocation are increased in BD manic, whereas stimulus-elicited PKC translocation is attenuated in schizophrenic patients. Lithium and valproic acid treatments attenuated the stimulus-induced PKC translocations to a similar degree and decreased PKC activity in both cytosolic and membranous fractions after two weeks of drug administration. An increase in 5-HT or thrombin stimulated [35S]GTPgammaS binding to Galpha proteins was detected in BD manic but not in schizophrenic patients although basal [35S]GTPgammaS binding was not different across the diagnostic groups. Lithium and valproic acid treatments similarly reduced receptor-G protein coupling with comparable time courses. Thus, increased membrane-associated PKC, cytosol to membrane PKC translocation and receptor-G protein coupling in platelets of BD manic patients were alleviated by lithium or valproic acid treatments.     

Transport of Amino Acid Esters and the Amino-Acid-Based Prodrug Valganciclovir by the Amino Acid Transporter ATB0,+

PURPOSE: The purpose of this study was to analyze the transport of amino acid esters and the amino-acid-based prodrug valganciclovir by the Na(+)/Cl(-)-coupled amino acid transporter ATB(0,+). METHODS: The interaction of amino acid esters and valganciclovir with the cloned rat ATB(0,+) was evaluated in a mammalian cell expression system and in the Xenopus oocyte expression system. RESULTS: In mammalian cells, expression of ATB(0,+) induced glycine uptake. This uptake was inhibited by valine and its methyl, butyl, and benzyl esters. The benzyl esters of other neutral amino acids were also effective inhibitors. Valganciclovir, the valyl ester of ganciclovir, was also found to inhibit ATB(0,+)-mediated glycine uptake competitively. Exposure of ATB(0,+)-expressing oocytes to glycine induced inward currents. Exposure to different valyl esters (methyl, butyl, and benzyl), benzyl esters of various neutral amino acids, and valganciclovir also induced inward currents in these oocytes. The current induced by valganciclovir was saturable with a K0.5 value of 3.1+/-0.7 mM and was obligatorily dependent on Na+ and Cl-. The Na+:Cl-:valganciclovir stoichiometry was 2 or 3:1:1. CONCLUSIONS: Amino acid esters and the amino-acid-based prodrug valganciclovir are transported by ATB(0,+). This shows that ATB(0,+) can serve as an effective delivery system for amino acid-based prodrugs.     

Sperm characteristics and accessory sex gland functions in HIV-infected men

A comparative study was carried out in the andrology clinic, Parirenyatwa Hospital, Harare, Zimbabwe, to investigate the sperm characteristics and accessory sex gland functions in HIV-infected individuals. Sixty-two patients with infertility problems who attended the clinic were requested to donate semen and blood after consent was obtained. HIV antibodies in paired semen and blood samples, sperm morphology, sperm count, sperm motility, seminal leucocytes, seminal fructose, seminal neutral alpha-glucosidase, and citric acid were analyzed. Nine out of 31 blood samples tested positive, while 21 out of 62 semen samples were positive for HIV. Leucocytospermia was associated with HIV-seropositive men (p < .01). The accessory sex gland function, as evaluated by biochemical markers, was not affected in HIV-seropositive men. HIV causes impairment of sperm motility by activating seminal leucocytes, which in turn induce oxidative stress on the sperm. Leucocytospermia is almost always present in HIV-seropositive men.     

A classification scheme for ventricular arrhythmias using wavelets analysis

Identification and classification of ventricular arrhythmias such as rhythmic ventricular tachycardia (VT) and disorganized ventricular fibrillation (VF) are vital tasks in guiding implantable devices to deliver appropriate therapy in preventing sudden cardiac deaths. Recent studies have shown VF can exhibit strong regional organizations, which makes the overlap zone between the fast paced rhythmic VT and VF even more ambiguous. Considering that implantable cardioverter-defibrillator (ICD) are primarily rate dependent detectors of arrhythmias and that there may be patients who suffer from arrhythmias that fall in the overlap zone, it is essential to identify the degree of affinity of the arrhythmia toward VT or organized/disorganized VF. The method proposed in this work better categorizes the overlap zone using Wavelet analysis of surface ECGs. Sixty-three surface ECG signal segments from the MIT-BIH database were used to classify between VT, organized VF (OVF), and disorganized VF (DVF). A two-level binary classifier was used to first extract VT with an overall accuracy of 93.7 % and then the separation between OVF and DVF with an accuracy of 80.0 %. The proposed approach could assist clinicians to provide optimal therapeutic solutions for patients in the overlap zone of VT and VF.     

Electrophysiological characterization and modeling of the structure activity relationship of the human concentrative nucleoside transporter 3 (hCNT3)

We characterized the electrophysiology, kinetics, and quantitative structure-activity relationship (QSAR) of the human concentrative nucleoside transporter 3 (hCNT3) expressed in Xenopus laevis oocytes by measuring substrate-induced inward currents using a two-microelectrode voltage-clamp system. At membrane potentials between -30 and -150 mV, sodium activation of gemcitabine transport was sigmoidal, with a K0.5 of 8.5+/-0.3 mM for Na+ and a Hill coefficient of 2.2+/-0.25 independent of membrane potential. We measured the Imax and K0.5 for substrate at -50 mV for the nucleoside analog drugs gemcitabine (638+/-58 nA, 59.7+/-17.5 microM), ribavirin (546+/-37 nA, 61.0+/-13.2 microM), AZT (420+/-4 nA, 310+/-9 microM), and 3-deazauridine (506+/-30 nA, 50.8+/-9.90 microM). K0.5 and Imax for substrate were dependent on membrane potential (both increasing as the membrane became more hyperpolarized) for all four drugs. hCNT3 also exhibited pre-steady-state currents. The quantitative structure-activity relationship (QSAR) was examined using comparative molecular field analysis and comparative molecular similarity indices analysis of the inward currents induced by 27 nucleoside analogs with substitutions at both the ribose and the nucleobase. Two statistically significant QSAR models identified electrostatic interaction as the major force in hCNT3 transport and attributed a critical role to the 3'-hydroxyl position of hCNT3 substrates. Steric hindrance at the 3-position and positive charge at the 5-position of the pyrimidine ring were favorable for transport. Two hCNT3 pharmacophore models revealed the minimal features required for hCNT3 transport as two hydrogen bond acceptors at 3'-OH and 5'-O and the hydrophobic center occupied by the base ring.     

Feature analysis of pathological speech signals using local discriminant bases technique

Speech is an integral part of the human communication system. Various pathological conditions affect the vocal functions, inducing speech disorders. Acoustic parameters of speech are commonly used for the assessment of speech disorders and for monitoring the progress of the patient over the course of therapy. In the last two decades, signal-processing techniques have been successfully applied in screening speech disorders. In the paper, a novel approach is proposed to classify pathological speech signals using a local discriminant bases (LDB) algorithm and wavelet packet decompositions. The focus of the paper was to demonstrate the significance of identifying the signal subspaces that contribute to the discriminatory characteristics of normal and pathological speech signals in a computationally efficient way. Features were extracted from target subspaces for classification, and time-frequency decomposition was used to eliminate the need for segmentation of the speech signals. The technique was tested with a database of 212 speech signals (51 normal and 161 pathological) using the Daubechies wavelet (db4). Classification accuracies up to 96% were achieved for a two-group classification as normal and pathological speech signals, and 74% was achieved for a four-group classification as male normal, female normal, male pathological and female pathological signals.