The effect of rotational angle and experimental parameters on the diffraction patterns and micro-structural information obtained from q-space diffusion NMR: implication for diffusion in white matter fibers

Diffusion NMR may provide, under certain experimental conditions, micro-structural information about confined compartments totally non-invasively. The influence of the rotational angle, the pulse gradient length and the diffusion time on the diffusion diffraction patterns and q-space displacement distribution profiles was evaluated for ensembles of long cylinders having a diameter of 9 and 20 microm. It was found that the diffraction patterns are sensitive to the rotational angle (alpha) and are observed only when diffusion is measured nearly perpendicular to the long axis of the cylinders i.e., when alpha= 90 degrees +/- 5 degrees under our experimental conditions. More importantly, we also found that the structural information extracted from the displacement distribution profiles and from the diffraction patterns are very similar and in good agreement with the experimental values for cylinders of 20 microm or even 9 microm, when data is acquired with parameters that satisfy the short gradient pulse (SGP) approximation (i.e., delta -->0) and the long diffusion time limit. Since these experimental conditions are hardly met in in vitro diffusion MRI of excised organs, and cannot be met in clinical MRI scanners, we evaluated the effect of the pulse gradient duration and the diffusion time on the structural information extracted from q-space diffusion MR experiments. Indeed it was found that, as expected, accurate structural information, and diffraction patterns are observed when Delta is large enough so that the spins reach the cylinders' boundaries. In addition, it was found that large delta results in extraction of a compartment size, which is smaller than the real one. The relevance of these results to q-space MRI of neuronal tissues and fiber tracking is discussed.     

Coherence resonance near a Hopf bifurcation

We report on the observation of coherence resonance for a semiconductor laser with short optical feedback close to Hopf bifurcations. Noise-induced self-pulsations are documented by distinct Lorentzian-like features in the power spectrum. The character of coherence is critically related to the type of the bifurcation. In the supercritical case, spectral width and height of the peak are monotonic functions of the noise level. In contrast, for the subcritical bifurcation, the width exhibits a minimum, translating into resonance behavior of the correlation time in the pulsation transients. A theoretical analysis based on the generic model of a self-sustained oscillator demonstrates that these observations are of general nature and are related to the fact that the damping depends qualitatively different on the noise intensity for the subcritical and supercritical case.     

White matter changes in multiple sclerosis: correlation of q-space diffusion MRI and 1H MRS

OBJECTIVE: To explore the diagnostic usefulness of high b-value diffusion magnetic resonance brain imaging ("q-space" imaging) in multiple sclerosis (MS). More specifically, we aimed at evaluating the ability of this methodology to identify tissue damage in the so-called normal-appearing white matter (NAWM). DESIGN: In this study we examined the correlation between q-space diffusion imaging and magnetic resonance spectroscopy (MRS)-based two-dimensional 1H chemical shift imaging. Eight MS patients with different degree of disease severity and seven healthy subjects were scanned in a 1.5-T magnetic resonance imaging (MRI) scanner. The MRI protocol included diffusion tensor imaging (DTI) (with bmax of 1000 s/mm2), high b-value diffusion-weighted imaging (with bmax of 14,000 s/mm2) and 2D chemical shift imaging. The high b-value data set was analyzed using the q-space methodology to produce apparent displacement and probability maps. RESULTS: We found that the q-space diffusion displacement and probability image intensities correlated well with N-acetylaspartate levels (r=.61 and .54, respectively). Furthermore, NAWM that was abnormal on MRS was also found to be abnormal using q-space diffusion imaging. In these areas, the q-space displacement values increased from 3.8+/-0.2 to 4.6+/-0.6 microm (P<.02), the q-space probability values decreased from 7.4+/-0.3 to 6.8+/-0.3 (P<.002), while DTI revealed only a small, but still significant, reduction in fractional anisotropy values from 0.40+/-0.02 to 0.37+/-0.02 (P<.05). CONCLUSION: High b-value diffusion imaging can detect tissue damage in the NAWM of MS patients. Despite the theoretical limitation of this method, in practice it provides additional information which is clinically relevant for detection of tissue damage not seen in conventional imaging techniques.     

Chiral analog gauge theories on the lattice

We discuss a class of lattice gauge theories with fermions that have properties in common with continuum chiral gauge theories. The symmetries we gauge have often been mistaken for chiral symmetries in the literature. We show that in the continuum limit they converge to ordinary vector-like symmetries, but that at strong coupling they behave like chiral symmetries. We find lattice analogs of the technicolor mechanism and of the generation of composite massless fermions in chiral gauge theories.     

Arbitrarily large neighborly families of symmetric convex polytopes

A family of convex d-polytopes in E _d is called neighborly if every two of them have a (d–1)-dimensional intersection. Settling an old problem of B. Grünbaum, we show that there exist arbitrarily large neighborly families of centrally (or any other prescribed type of) symmetric convex d-poliytopes in E _d ,for all d3; moreover, they can all be congruent, if d4.A version of this paper has been written while the author visited R. K. Guy in Calgary, Alberta, Canada, in the summer of 1981; the author wishes to thank Louise and Richard Guy for their warm hospitality.     

2H NMR measurements of the molecular orientational ordering in a reentrant liquid-crystalline mixture

Deuterium NMR measurements of the molecular orientational ordering (nematic order) are reported for a reentrant binary mixture of some alkoxy-eyanobiphenyls (nOCB) to which small amounts of perdeuterated p-xylene have been added. The results indicate that the degree of orientational order is enhanced at the smectic A to reentrant nematic phase transition. This effect is shown to be in agreement with the predictions of a Landau-type theory of the reentrant phase transition where-by this phase transition is explained as the result of a coupling between the smectic and the nematic order parameters.     

Enzymatic kinetic resolution of piperidine atropisomers: synthesis of a key intermediate of the farnesyl protein transferase inhibitor, SCH66336

The resolution of secondary amines via enzyme-catalyzed acylation is a relatively rare process. The kinetic resolution of a series of intermediates of SCH66336 (1), by either enzymatic acylation of the pendant piperidine (4, 5) or hydrolysis of the corresponding carbamate 3, was investigated. In the case of 4, the molecule exists as a pair of enantiomers due to atropisomerism about the exocyclic double bond. The enzymatic acylation of (+/-)-4 was optimized in terms of acylating agent, solvent, and moisture content. The use of lipase, Toyobo LIP-300, and trifluoroethyl isobutyrate as acylating agent resulted in isobutyrylation of the (+)-enantiomer, which is easily separated from the unwanted (-)-4. Hydrolysis of the isobutyramide 6c yielded the desired (+)-4 in high enantiomeric excess. (-)-4 may be recovered from the resolution step, racemized, and resubjected to enzymatic acylation to increase material throughput.     

DNN Intellectual Property Extraction Using Composite Data

As state-of-the-art deep neural networks are being deployed at the core level of increasingly large numbers of AI-based products and services, the incentive for “copying them” (i.e., their intellectual property, manifested through the knowledge that is encapsulated in them) either by adversaries or commercial competitors is expected to considerably increase over time. The most efficient way to extract or steal knowledge from such networks is by querying them using a large dataset of random samples and recording their output, which is followed by the training of a student network, aiming to eventually mimic these outputs, without making any assumption about the original networks. The most effective way to protect against such a mimicking attack is to answer queries with the classification result only, omitting confidence values associated with the softmax layer. In this paper, we present a novel method for generating composite images for attacking a mentor neural network using a student model. Our method assumes no information regarding the mentor’s training dataset, architecture, or weights. Furthermore, assuming no information regarding the mentor’s softmax output values, our method successfully mimics the given neural network and is capable of stealing large portions (and sometimes all) of its encapsulated knowledge. Our student model achieved 99% relative accuracy to the protected mentor model on the Cifar-10 test set. In addition, we demonstrate that our student network (which copies the mentor) is impervious to watermarking protection methods and thus would evade being detected as a stolen model by existing dedicated techniques. Our results imply that all current neural networks are vulnerable to mimicking attacks, even if they do not divulge anything but the most basic required output, and that the student model that mimics them cannot be easily detected using currently available techniques.