Abstract: Importance Mammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives. Objective To evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms. Design, Setting, and Participants In this diagnostic accuracy study conducted between September 2016 and November 2017, an international, crowdsourced challenge was hosted to foster AI algorithm development focused on interpreting screening mammography. More than 1100 participants comprising 126 teams from 44 countries participated. Analysis began November 18, 2016. Main Outcomes and Measurements Algorithms used images alone (challenge 1) or combined images, previous examinations (if available), and clinical and demographic risk factor data (challenge 2) and output a score that translated to cancer yes/no within 12 months. Algorithm accuracy for breast cancer detection was evaluated using area under the curve and algorithm specificity compared with radiologists' specificity with radiologists' sensitivity set at 85.9% (United States) and 83.9% (Sweden). An ensemble method aggregating top-performing AI algorithms and radiologists' recall assessment was developed and evaluated. Results Overall, 144231 screening mammograms from 85580 US women (952 cancer positive <= 12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166578 examinations from 68008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists' sensitivity, lower than community-practice radiologists' specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity. Conclusions and Relevance While no single AI algorithm outperformed radiologists, an ensemble of AI algorithms combined with radiologist assessment in a single-reader screening environment improved overall accuracy. This study underscores the potential of using machine learning methods for enhancing mammography screening interpretation. Question How do deep learning algorithms perform compared with radiologists in screening mammography interpretation? Findings In this diagnostic accuracy study using 144231 screening mammograms from 85580 women from the United States and 166578 screening mammograms from 68008 women from Sweden, no single artificial intelligence algorithm outperformed US community radiologist benchmarks; including clinical data and prior mammograms did not improve artificial intelligence performance. However, combining best-performing artificial intelligence algorithms with single-radiologist assessment demonstrated increased specificity. Meaning Integrating artificial intelligence to mammography interpretation in single-radiologist settings could yield significant performance improvements, with the potential to reduce health care system expenditures and address resource scarcity experienced in population-based screening programs. This diagnostic accuracy study evaluates whether artificial intelligence can overcome human mammography interpretation limits with a rigorous, unbiased evaluation of machine learning algorithms.

Abstract: We study the predictions for CP phases and absolute neutrino mass scale for broad classes of models with a U(2)-like flavor symmetry. For this purpose we consider the same special textures in neutrino and charged lepton mass matrices that are successful in the quark sector. While in the neutrino sector the U(2) structure enforces two texture zeros, the contribution of the charged lepton sector to the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix can be parametrized by two rotation angles. Restricting to the cases where at least one of these angles is small, we obtain three representative scenarios. In all scenarios we obtain a narrow prediction for the sum of neutrino masses in the range of 60-75 meV, possibly in the reach of upcoming galaxy survey experiments. All scenarios can be excluded if near-future experimental date provide evidence for either neutrinoless double-beta decay or inverted neutrino mass ordering.

Abstract: The integrated fiducial cross-section and unfolded differential jet mass spectrum of high transverse momentum Z -> b (b) over bar decays are measured in Z gamma events in proton-proton collisions at root s = 13 TeV. The data analysed were collected between 2015 and 2016 with the ATLAS detector at the Large Hadron Collider and correspond to an integrated luminosity of 36.1 fb(-1). Photons are required to have a transverse momentum p(T) > 175 GeV. The Z -> b (b) over bar decay is reconstructed using a jet with p(T) > 200 GeV, found with the anti-k(t) R = 1.0 jet algorithm, and groomed to remove soft and wide-angle radiation and to mitigate contributions from the underlying event and additional proton-proton collisions. Two different but related measurements are performed using two jet grooming definitions for reconstructing the Z -> b (b) over bar decay: trimming and soft drop. These algorithms differ in their experimental and phenomenological implications regarding jet mass reconstruction and theoretical precision. To identify Zbosons, b-tagged R = 0.2 track-jets matched to the groomed large-R calorimeter jet are used as a proxy for the b-quarks. The signal yield is determined from fits of the data-driven background templates to the different jet mass distributions for the two grooming methods. Integrated fiducial cross-sections and unfolded jet mass spectra for each grooming method are compared with leading-order theoretical predictions. The results are found to be in good agreement with Standard Model expectations within the current statistical and systematic uncertainties.

Abstract: We analyze the decays of the theoretically predicted lowest bottomonium hybrid H(1P) to open bottom two-meson states. We do it by embedding a quark pair creation model into the Born-Oppenheimer framework which allows for a unified, QCD-motivated description of bottomonium hybrids as well as bottomonium. A new 1P1 decay model for H(1P) comes out. The same analysis applied to bottomonium leads naturally to the well-known 3 P0 decay model. We show that H(1P) and the theoretically predicted bottomonium state Upsilon (5S), whose calculated masses are close to each other, have very different widths for such decays. A comparison with data from Upsilon (10860), an experimental resonance whose mass is similar to that of Upsilon (5S) and H(1P), is carried out. Neither a Upsilon (5S) nor a H(1P) assignment can explain the measured decay widths. However, a Upsilon (5S)-H(1P) mixing may give account of them supporting previous analyses of dipion decays of Upsilon (10860) and suggesting a possible experimental evidence of H(1P).

Abstract: We study the energy distribution and equation of state of the universe between the end of inflation and the onset of radiation domination (RD), considering observationally consistent single-field inflationary scenarios, with a potential 'flattening' at large field values, and a monomial shape V(phi) proportional to vertical bar phi vertical bar(p) around the origin. As a proxy for (p)reheating, we include a quadratic interaction g(2)phi X-2(2) between the inflaton phi and a light scalar 'daughter' field X, with g(2) > 0. We capture the non-perturbative and non-linear nature of the system dynamics with lattice simulations, obtaining that: i) the final energy transferred to X depends only on p, not on g(2); ii) the final transfer of energy is always negligible for 2 <= p < 4, and of order similar to 50% for p >= 4; iii) the system goes at late times to matter-domination for p = 2, and always to RD for p > 2. In the latter case we calculate the number of e-folds until RD, significantly reducing the uncertainty in the inflationary observables Tl-s and r.