Random projections on manifolds of symmetric positive definite matrices for image classification

Recent advances suggest that encoding images through Symmetric Positive Definite (SPD) matrices and then interpreting such matrices as points on Riemannian manifolds can lead to increased classification performance. Taking into account manifold geometry is typically done via (1) embedding the manifolds in tangent spaces, or (2) embedding into Reproducing Kernel Hilbert Spaces (RKHS). While embedding into tangent spaces allows the use of existing Euclidean-based learning algorithms, manifold shape is only approximated which can cause loss of discriminatory information. The RKHS approach retains more of the manifold structure, but may require non-trivial effort to kernelise Euclidean-based learning algorithms. In contrast to the above approaches, in this paper we offer a novel solution that allows SPD matrices to be used with unmodified Euclidean-based learning algorithms, with the true manifold shape well-preserved. Specifically, we propose to project SPD matrices using a set of random projection hyperplanes over RKHS into a random projection space, which leads to representing each matrix as a vector of projection coefficients. Experiments on face recognition, person re-identification and texture classification show that the proposed approach outperforms several recent methods, such as Tensor Sparse Coding, Histogram Plus Epitome, Riemannian Locality Preserving Projection and Relational Divergence Classification.

Matching image sets via adaptive multi convex hull

Traditional nearest points methods use all the samples in an image set to construct a single convex or affine hull model for classification. However, strong artificial features and noisy data may be generated from combinations of training samples when significant intra-class variations and/or noise occur in the image set. Existing multi-model approaches extract local models by clustering each image set individually only once, with fixed clusters used for matching with various image sets. This may not be optimal for discrimination, as undesirable environmental conditions (eg. illumination and pose variations) may result in the two closest clusters representing different characteristics of an object (eg. frontal face being compared to non-frontal face). To address the above problem, we propose a novel approach to enhance nearest points based methods by integrating affine/convex hull classification with an adapted multi-model approach. We first extract multiple local convex hulls from a query image set via maximum margin clustering to diminish the artificial variations and constrain the noise in local convex hulls. We then propose adaptive reference clustering (ARC) to constrain the clustering of each gallery image set by forcing the clusters to have resemblance to the clusters in the query image set. By applying ARC, noisy clusters in the query set can be discarded. Experiments on Honda, MoBo and ETH-80 datasets show that the proposed method outperforms single model approaches and other recent techniques, such as Sparse Approximated Nearest Points, Mutual Subspace Method and Manifold Discriminant Analysis.

Formation of neutral C7H2 isomers from pour isomeric C7H2 radical anion precursors in the gas phase

Consideration of theoretical calculations \[E3LYP/aug-cc-pVDZ//B3LYP/6-31G(d)\] of the structures of ten C7H2 neutral isomers and the nine corresponding C7H2 radical anions have led us to synthesize four stable C7H2 radical anions in the ion source of our ZAB 2HF mass spectrometer, and to convert these to C7H2 neutrals. The four radical anion isomers prepared were (i) \[(HC equivalent to C)(2)C=C=C\](-.) \[from the reaction between (HC equivalent to C)(3)COCH3 and HO- \], (ii) \[HC=C=C=C=C=C=CH\](-.) \[from the reaction between HC equivalent to C-C equivalent to C-CD(OH)-C equivalent to CH and HO-\], (iii) \[C=C=C=C=C=C=CH2\](-.) \[from the reaction between DC equivalent to C-C equivalent to C-C equivalent to C-CH2OCH2CH3 and HO-\], and (iv) \[C equivalent to C-CH2-C equivalent to C-C equivalent to C\](-.) \[from the bis desilylation reaction of (CH3)(3)Si-C equivalent to C-CH2-C equivalent to C-C equivalent to C-Si (CH3)(3)With SF6-.\]. The four anions were further characterized by their collisional activation (negative ion) and charge reversal (CR, positive ion) mass spectra. The anions were converted into their corresponding neutrals by charge stripping, and the correspondence between the charge reversal (CR) and neutralization reionization (-NR+) mass spectra of each anion is taken as evidence that within the time frame of the -NR+ experiment (some 10(-6) s), each neutral is stable and undergoes no major rearrangement or interconversion to a more stable isomer. Theory and experiment are in accord for these systems.

Formation of neutral C7H2 isomers from four isomeric C7H2 radical anion precursors in the gas phase

Consideration of theoretical calculations [B3LYP/aug-cc-pVDZ//B3LYP/6-31G(d)] of the structures of ten C7H2 neutral isomers and the nine corresponding C7H2 radical anions have led us to synthesize four stable C7H2 radical anions in the ion source of our ZAB 2HF mass spectrometer, and to convert these to C7H2 neutrals. The four radical anion isomers prepared were (i) [(HC≡C)2C=C=C]-̇ [from the reaction between (HC≡C)3COCH3 and HO- ], (ii) [HC=C=C=C=C=C=CH]-̇ [from the reaction between HC≡C-C≡C- CD(OH)-C≡CH and HO-], (iii) [C=C=C=C=C=C=CH2]-̇ [from the reaction between DC≡C-C≡C- C≡C-CH2OCH2CH3 and HO-], and (iv) [C≡C-CH2-C≡C-C≡C]-̇ [from the bis desilylation reaction of (CH3)3Si-C≡C-CH2-C≡C-C≡C-Si (CH3)3 with SF6 -̇]. The four anions were further characterized by their collisional activation (negative ion) and charge reversal (CR, positive ion) mass spectra. The anions were converted into their corresponding neutrals by charge stripping, and the correspondence between the charge reversal (CR) and neutralization reionization (-NR+) mass spectra of each anion is taken as evidence that within the time frame of the -NR+ experiment (some 10-6 s), each neutral is stable and undergoes no major rearrangement or interconversion to a more stable isomer. Theory and experiment are in accord for these systems.

Improved image set classification via joint sparse approximated nearest subspaces

Existing multi-model approaches for image set classification extract local models by clustering each image set individually only once, with fixed clusters used for matching with other image sets. However, this may result in the two closest clusters to represent different characteristics of an object, due to different undesirable environmental conditions (such as variations in illumination and pose). To address this problem, we propose to constrain the clustering of each query image set by forcing the clusters to have resemblance to the clusters in the gallery image sets. We first define a Frobenius norm distance between subspaces over Grassmann manifolds based on reconstruction error. We then extract local linear subspaces from a gallery image set via sparse representation. For each local linear subspace, we adaptively construct the corresponding closest subspace from the samples of a probe image set by joint sparse representation. We show that by minimising the sparse representation reconstruction error, we approach the nearest point on a Grassmann manifold. Experiments on Honda, ETH-80 and Cambridge-Gesture datasets show that the proposed method consistently outperforms several other recent techniques, such as Affine Hull based Image Set Distance (AHISD), Sparse Approximated Nearest Points (SANP) and Manifold Discriminant Analysis (MDA).

Image-guided sampling reveals increased stroma and lower glandular complexity in mammographically dense breast tissue

Mammographic density (MD) adjusted for age and body mass index (BMI) is a strong heritable breast cancer risk factor; however, its biological basis remains elusive. Previous studies assessed MD-associated histology using random sampling approaches, despite evidence that high and low MD areas exist within a breast and are negatively correlated with respect to one another. We have used an image-guided approach to sample high and low MD tissues from within individual breasts to examine the relationship between histology and degree of MD. Image-guided sampling was performed using two different methodologies on mastectomy tissues (n = 12): (1) sampling of high and low MD regions within a slice guided by bright (high MD) and dark (low MD) areas in a slice X-ray film; (2) sampling of high and low MD regions within a whole breast using a stereotactically guided vacuum-assisted core biopsy technique. Pairwise analysis accounting for potential confounders (i.e. age, BMI, menopausal status, etc.) provides appropriate power for analysis despite the small sample size. High MD tissues had higher stromal (P = 0.002) and lower fat (P = 0.002) compositions, but no evidence of difference in glandular areas (P = 0.084) compared to low MD tissues from the same breast. High MD regions had higher relative gland counts (P = 0.023), and a preponderance of Type I lobules in high MD compared to low MD regions was observed in 58% of subjects (n = 7), but did not achieve significance. These findings clarify the histologic nature of high MD tissue and support hypotheses regarding the biophysical impact of dense connective tissue on mammary malignancy. They also provide important terms of reference for ongoing analyses of the underlying genetics of MD.

Joint 2D and 3D cues for image segmentation towards robotic applications

This thesis investigates the fusion of 3D visual information with 2D image cues to provide 3D semantic maps of large-scale environments in which a robot traverses for robotic applications. A major theme of this thesis was to exploit the availability of 3D information acquired from robot sensors to improve upon 2D object classification alone. The proposed methods have been evaluated on several indoor and outdoor datasets collected from mobile robotic platforms including a quadcopter and ground vehicle covering several kilometres of urban roads.

The heat shock protein 90 inhibitor, 17-allylamino-17- demethoxygeldanamycin, enhances osteoclast formation and potentiates bone metastasis of a human breast cancer cell line

Breast cancer metastasis to the bone occurs frequently, causing numerous complications including severe pain, fracture, hypercalcemia, and paralysis. Despite its prevalence and severity, few effective therapies exist. To address this, we examined whether the heat shock protein 90 (Hsp90) inhibitor, 17-allylamino-17-demethoxygeldanamycin (17-AAG), would be efficacious in inhibiting breast cancer metastasis to bone. Utilizing the human breast cancer subline, MDA-MB-231SA, previously in vivo selected for its enhanced ability to generate osteolytic bone lesions, we determined that 17-AAG potently inhibited its in vitro proliferation and migration. Moreover, 17-AAG significantly reduced MDA-MB-231SA tumor growth in the mammary-fat pad of nude mice. Despite these findings, 17-AAG enhanced the incidence of bone metastasis and osteolytic lesions following intracardiac inoculation in the nude mouse. Consistent with these findings, 17-AAG enhanced osteoclast formation 2- to 4-fold in mouse bone marrow/osteoblast cocultures, receptor activator of nuclear factor κB ligand (BANKL)-stimulated bone marrow, and RAW264.7 cell models of in vitro osteoclastogenesis. Moreover, the drug enhanced osteoclastogenesis in human cord blood progenitor cells, demonstrating that its effects were not limited to mouse models. In addition to 17-AAG, other Hsp90 inhibitors, such as radicicol and herbimycin A, also enhanced osteoclastogenesis. A pro-osteolytic action of 17-AAG independent of tumor presence was also determined in vivo, in which 17-AAG-treated tumor-naive mice had reduced trabecular bone volume with an associated increase in osteoclast number. Thus, HSP90 inhibitors can stimulate osteoclast formation, which may underlie the increased incidence of osteolysis and skeletal tumor incidence causedby 17-AAG in vivo. These data suggest an important contraindication to the Hsp90 targeted cancer therapy currently undergoing clinical trial.

Formation of neutral molecules of potential stellar interest by neutralisation of negative ions in a mass spectrometer - The application of experiment and molecular modelling in concert

This paper is a modified version of a lecture which describes the synthesis, structure and reactivity of some neutral molecules of stellar significance. The neutrals are formed in the collision cell of a mass spectrometer following vertical Franck-Condon one electron oxidation of anions of known bond connectivity. Neutrals are characterised by conversion to positive ions and by extensive theoretical studies at the CCSD(T)/aug-cc-pVDZ//B3LYP/6-31G(d) level of theory. Four systems are considered in detail, viz (i) the formation of linear C-4 and its conversion to the rhombus C-4, (ii) linear C-5 and the atom scrambling of this system when energised, (iii) the stable cumulene oxide CCCCCO, and (iv) the elusive species O2C-CO. This paper is not intended to be a review of interstellar chemistry: examples are selected from our own work in this area. (C) 2002 Elsevier Science Inc. All rights reserved.

Phosphorus-based functional groups as hydrogen bonding templates for rotaxane formation

We report on the use of the hydrogen bond acceptor properties of some phosphorus-containing functional groups for the assembly of a series of [2]rotaxanes. Phosphinamides, and the homologous thio– and selenophosphinamides, act as hydrogen bond acceptors that, in conjunction with an appropriately positioned amide group on the thread, direct the assembly of amide-based macrocycles around the axle to form rotaxanes in up to 60% yields. Employing solely phosphorus-based functional groups as the hydrogen bond accepting groups on the thread, a bis(phosphinamide) template and a phosphine oxide-phosphinamide template afforded the corresponding rotaxanes in 18 and 15 % yields, respectively. X-Ray crystallography of the rotaxanes shows the presence of up to four intercomponent hydrogen bonds between the amide groups of the macrocycle and various hydrogen bond accepting groups on the thread, including rare examples of amide-to-phosphonamide, -thiophosphinamide and -selenophosphinamide groups. With a phosphine oxide-phosphinamide thread, the solid state structure of the rotaxane is remarkable, featuring no direct intercomponent hydrogen bonds but rather a hydrogen bond network involving water molecules that bridge the H-bonding groups of the macrocycle and thread through bifurcated hydrogen bonds. The incorporation of phosphorus-based functional groups into rotaxanes may prove useful for the development of molecular shuttles in which the macrocycle can be used to hinder or expose binding ligating sites for metal-based catalysts.

Improvements to bauxite residue neutralisation efficiency through the use of alternative feed sources

Currently, open circuit Bayer refineries pump seawater directly into their operations to neutralise the caustic fraction of the Bayer residue. The resulting supernatant has a reduced pH and is pumped back to the marine environment. This investigation has assessed modified seawater sources generated from different ion filtration processes to compare their relative capacities to neutralise bauxite residues. An assessment of the chemical stability of the neutralisation products, neutralisation efficiency, discharge water quality, bauxite residue composition, and associated economic benefits have been considered to determine the most preferable seawater filtration process based on implementation costs, savings to operations and environmental benefits. The mechanism of neutralisation for each technology was determined to be predominately due to the formation of Bayer hydrotalcite and calcium carbonate, however variations in neutralisation capacity and efficiencies have been observed. The neutralisation efficiency of each feed source has been found to be dependent on the concentration of magnesium, aluminium, calcium and carbonate. These studies have revealed that multiple neutralisation steps occur throughout the process. Environmental, economic and social advantages and disadvantages of the different filtration technologies have been explored to determine the most sustainable method for the neutralisation of bauxite residues. The relative degree of “green” associated with nanofiltered seawater and reverse osmosis filtered seawater are discussed.

Scleral matrix metalloproteinases, serine proteinase activity and hydrational capacity are increased in myopia induced by retinal image degradation

In the avian model of myopia, retinal image degradation quickly leads to ocular enlargement. We now give evidence that regionally specific changes in ocular size are correlated with both biomechanical indices of scleral remodeling, e.g. hydration capacity and with biochemical changes in proteinase activities. The latter include a 72 kDa matrix metalloproteinase (putatively MMP-2), other gelatin-binding MMPs, an acid pH MMP and a serine protease. Specifically, we have found that increases in scleral hydrational capacity parallel increases in collagen degrading activities. Gelatin zymography reveals that eyes with 7 days of retinal image degradation have elevated levels (1.4-fold) of gelatinolytic activities at 72 and 67 kDa M(r) in equatorial and posterior pole regions of the sclera while, after 14 days of treatment, increases are no longer apparent. Lower M(r) zymographic activities at 50, 46 and 37 kDa M(r) are collectively increased in eyes treated for both 7 and 14 days (1.4- and 2.4-fold respectively) in the equator and posterior pole areas of enlarging eyes. Western blot analyses of scleral extracts with an antibody to human MMP-2 reveals immunoreactive bands at 65, 30 and 25 kDa. Zymograms incubated under slightly acidic conditions reveal that, in enlarging eyes, MMP activities at 25 and 28 kDa M(r) are increased in scleral equator and posterior pole (1.6- and 4.5-fold respectively). A TIMP-like protein is also identified in sclera and cornea by Western blot analysis. Finally, retinal-image degradation also increases (~2.6-fold) the activity of a 23.5 kDa serine proteinase in limbus, equator and posterior pole sclera that is inhibited by aprotinin and soybean trypsin inhibitor. Taken together, these results indicate that eye growth induced by retinal-image degradation involves increases in the activities of multiple scleral proteinases that could modify the biomechanical properties of scleral structural components and contribute to tissue remodeling and growth.

Hyperbranched polymer-gold nanoparticle assemblies : role of polymer architecture in hybrid assembly formation and SERS activity

Plasmonic gold nano-assemblies that self-assemble with the aid of linking molecules or polymers have the potential to yield controlled hierarchies of morphologies and consequently result in materials with tailored optical (e.g. localized surface plasmon resonances (LSPR)) and spectroscopic properties (e.g. surface enhanced Raman scattering (SERS)). Molecular linkers that are structurally well-defined are promising for forming hybrid nano-assemblies which are stable in aqueous solution and are increasingly finding application in nanomedicine. Despite much ongoing research in this field, the precise role of molecular linkers in governing the morphology and properties of the hybrid nano-assemblies remains unclear. Previously we have demonstrated that branched linkers, such as hyperbranched polymers, with specific anchoring end groups can be successfully employed to form assemblies of gold NPs demonstrating near-infrared SPRs and intense SERS scattering. We herein introduce a tailored polymer as a versatile molecular linker, capable of manipulating nano-assembly morphologies and hot-spot density. In addition, this report explores the role of the polymeric linker architecture, specifically the degree of branching of the tailored polymer in determining the formation, morphology and properties of the hybrid nano-assemblies. The degree of branching of the linker polymer, in addition to the concentration and number of anchoring groups, is observed to strongly influence the self-assembly process. The assembly morphology shifts primarily from 1D-like chains to 2D plates and finally to 3D-like globular structures, with increase in degree of branching. Insights have been gained into how the morphology influences the SERS performance of these nano-assemblies with respect to hot-spot density. These findings supplement the understanding of the morphology determining nano-assembly formation and pave the way for the possible application of these nano-assemblies as SERS bio-sensors for medical diagnostics.

University of Glasgow at ImageCLEFPhoto 2009 : optimising similarity and diversity in image retrieval

In this paper we describe the approaches adopted to generate the runs submitted to ImageCLEFPhoto 2009 with an aim to promote document diversity in the rankings. Four of our runs are text based approaches that employ textual statistics extracted from the captions of images, i.e. MMR [1] as a state of the art method for result diversification, two approaches that combine relevance information and clustering techniques, and an instantiation of Quantum Probability Ranking Principle. The fifth run exploits visual features of the provided images to re-rank the initial results by means of Factor Analysis. The results reveal that our methods based on only text captions consistently improve the performance of the respective baselines, while the approach that combines visual features with textual statistics shows lower levels of improvements.

Assessment of angiogenic potential--the use of AIDS-KS cell supernatants as an in vitro model

Metastasis, the passage of primary tumour cells throughout the body via the vascular system and their subsequent proliferation into secondary lesions in distant organs, represents a poor prognosis and therefore an understandably feared event for cancer patients. Despite considerable advances in cancer diagnosis and treatment, most deaths are the result of metastases resistant to conventional treatment [1]. Rather than being a random process, metastasis involves a series of organised steps leading to the growth of a secondary tumour. Malignant tumours stimulate the production of new vessels by the host, and this process is a prerequisite for the increase in size of a new tumour [2]. Angiogenesis, not only permits tumour expansion but also allows the entry of tumour cells into the circulation and is probably the most vital event for the metastatic process [3]. Metastasis and angiogenesis [4] have received much attention in recent years. A biological understanding of both phenomena seems to be an urgent priority towards the search for an effective prevention and treatment of tumour progression. Studies in vitro and in vivo have shown that one of the most important barriers to the passage of malignant cells is the basement membrane. The crossing of such barriers is a vital step in the formation of a metastasis [5].