Performance Comparison of Individual and Ensemble CNN Models for the Classification of Brain 18F-FDG-PET Scans.

The high-background glucose metabolism of normal gray matter on [18F]-fluoro-2-D-deoxyglucose (FDG) positron emission tomography (PET) of the brain results in a low signal-to-background ratio, potentially increasing the possibility of missing important findings in patients with intracranial malignancies. To explore the strategy of using a deep learning classifier to aid in distinguishing normal versus abnormal findings on PET brain images, this study evaluated the performance of a two-dimensional convolutional neural network (2D-CNN) to classify FDG PET brain scans as normal (N) or abnormal (A). METHODS: Two hundred eighty-nine brain FDG-PET scans (N; n = 150, A; n = 139) resulting in a total of 68,260 images were included. Nine individual 2D-CNN models with three different window settings for axial, coronal, and sagittal axes were trained and validated. The performance of these individual and ensemble models was evaluated and compared using a test dataset. Odds ratio, Akaike's information criterion (AIC), and area under curve (AUC) on receiver-operative-characteristic curve, accuracy, and standard deviation (SD) were calculated. RESULTS: An optimal window setting to classify normal and abnormal scans was different for each axis of the individual models. An ensembled model using different axes with an optimized window setting (window-triad) showed better performance than ensembled models using the same axis and different windows settings (axis-triad). Increase in odds ratio and decrease in SD were observed in both axis-triad and window-triad models compared with individual models, whereas improvements of AUC and AIC were seen in window-triad models. An overall model averaging the probabilities of all individual models showed the best accuracy of 82.0%. CONCLUSIONS: Data ensemble using different window settings and axes was effective to improve 2D-CNN performance parameters for the classification of brain FDG-PET scans. If prospectively validated with a larger cohort of patients, similar models could provide decision support in a clinical setting.

Oxidation and hydration of U3O8 materials following controlled exposure to temperature and humidity.

Chemical signatures correlated with uranium oxide processing are of interest to forensic science for inferring sample provenance. Identification of temporal changes in chemical structures of process uranium materials as a function of controlled temperatures and relative humidities may provide additional information regarding sample history. In this study, a high-purity α-U3O8 sample and three other uranium oxide samples synthesized from reaction routes used in nuclear conversion processes were stored under controlled conditions over 2-3.5 years, and powder X-ray diffraction analysis and X-ray absorption spectroscopy were employed to characterize chemical speciation. Signatures measured from the α-U3O8 sample indicated that the material oxidized and hydrated after storage under high humidity conditions over time. Impurities, such as uranyl fluoride or schoepites, were initially detectable in the other uranium oxide samples. After storage under controlled conditions, the analyses of the samples revealed oxidation over time, although the signature of the uranyl fluoride impurity diminished. The presence of schoepite phases in older uranium oxide material is likely indicative of storage under high humidity and should be taken into account for assessing sample history. The absence of a signature from a chemical impurity, such as uranyl fluoride hydrate, in an older material may not preclude its presence at the initial time of production. LA-UR-15-21495.

Investigation of thorium salts as candidate materials for direct observation of the (229m)Th nuclear transition.

Recent efforts to measure the (229m)Th → (229g)Th nuclear transition sparked interest in understanding the electronic structure of wide-gap thorium salts. Such materials could be used to measure this nuclear transition using optical spectroscopy in solid-state devices. Here, we present screened hybrid density functional theory and many-body G0W0 calculations of two candidate materials, namely, Na2ThF6 and ThF4, for such a measurement. Our results show an electronic gap larger than 10 eV for both materials, suggesting that the internal conversion nuclear de-excitation channel would be suppressed in these materials. We also present results for ThX4 (X = Cl, Br, I), materials with smaller gaps significantly easier to access experimentally.

Seniority and orbital symmetry as tools for establishing a full configuration interaction hierarchy.

We explore the concept of seniority number (defined as the number of unpaired electrons in a determinant) when applied to the problem of electron correlation in atomic and molecular systems. Although seniority is a good quantum number only for certain model Hamiltonians (such as the pairing Hamiltonian), we show that it provides a useful partitioning of the electronic full configuration interaction (FCI) wave function into rapidly convergent Hilbert subspaces whose weight diminishes as its seniority number increases. The primary focus of this study is the adequate description of static correlation effects. The examples considered are the ground states of the helium, beryllium, and neon atoms, the symmetric dissociation of the N(2) and CO(2) molecules, as well as the symmetric dissociation of an H(8) hydrogen chain. It is found that the symmetry constraints that are normally placed on the spatial orbitals greatly affect the convergence rate of the FCI expansion. The energy relevance of the seniority zero sector (determinants with all paired electrons) increases dramatically if orbitals of broken spatial symmetry (as those commonly used for Hubbard Hamiltonian studies) are allowed in the wave function construction.

Constrained-pairing mean-field theory. V. Triplet pairing formalism.

Describing strong (also known as static) correlation caused by degenerate or nearly degenerate orbitals near the Fermi level remains a theoretical challenge, particularly in molecular systems. Constrained-pairing mean-field theory has been quite successful, capturing the effects of static correlation in bond formation and breaking in closed-shell molecular systems by using singlet electron entanglement to model static correlation at mean-field computational cost. This work extends the previous formalism to include triplet pairing. Additionally, a spin orbital extension of the "odd-electron" formalism is presented as a method for understanding electron entanglement in molecules.

Projected quasiparticle theory for molecular electronic structure.

We derive and implement symmetry-projected Hartree-Fock-Bogoliubov (HFB) equations and apply them to the molecular electronic structure problem. All symmetries (particle number, spin, spatial, and complex conjugation) are deliberately broken and restored in a self-consistent variation-after-projection approach. We show that the resulting method yields a comprehensive black-box treatment of static correlations with effective one-electron (mean-field) computational cost. The ensuing wave function is of multireference character and permeates the entire Hilbert space of the problem. The energy expression is different from regular HFB theory but remains a functional of an independent quasiparticle density matrix. All reduced density matrices are expressible as an integration of transition density matrices over a gauge grid. We present several proof-of-principle examples demonstrating the compelling power of projected quasiparticle theory for quantum chemistry.

On Pair Functions for Strong Correlations.

The UHF wave function may be written as a spin-contaminated pair wave function of the APSG form, and the overlap of the α and ß corresponding orbitals of the UHF solution can be taken as a proxy for the strength of the correlation captured by breaking symmetry. We demonstrate this with calculations on one- and two-dimensional hydrogen clusters and make contact with the well studied Hubbard model. The UHF corresponding orbitals pair in a manner that allows a smooth evolution from doubly occupied orbitals at small distance to one in which wave function breaks symmetry, segregating the α and ß electrons onto distinct sublattices at large distances. By performing spin projection on these UHF solutions, we address strong correlations that are difficult to capture at intermediate distances using a single determinant.