|
|
Schaffter, T. et al, Albiol, F., & Caballero, L. (2020). Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms. JAMA Netw. Open, 3(3), e200265–15pp.
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.
|
|
|
|
Assam, I., Vijande, J., Ballester, F., Perez-Calatayud, J., Poppe, B., & Siebert, F. A. (2022). Evaluation of dosimetric effects of metallic artifact reduction and tissue assignment on Monte Carlo dose calculations for I-125 prostate implants. Med. Phys., 49, 6195–6208.
Abstract: Purpose Monte Carlo (MC) simulation studies, aimed at evaluating the magnitude of tissue heterogeneity in I-125 prostate permanent seed implant brachytherapy (BT), customarily use clinical post-implant CT images to generate a virtual representation of a realistic patient model (virtual patient model). Metallic artifact reduction (MAR) techniques and tissue assignment schemes (TAS) are implemented on the post-implant CT images to mollify metallic artifacts due to BT seeds and to assign tissue types to the voxels corresponding to the bright seed spots and streaking artifacts, respectively. The objective of this study is to assess the combined influence of MAR and TAS on MC absorbed dose calculations in post-implant CT-based phantoms. The virtual patient models used for I-125 prostate implant MC absorbed dose calculations in this study are derived from the CT images of an external radiotherapy prostate patient without BT seeds and prostatic calcifications, thus averting the need to implement MAR and TAS. Methods The geometry of the IsoSeed I25.S17plus source is validated by comparing the MC calculated results of the TG-43 parameters for the line source approximation with the TG-43U1S2 consensus data. Four MC absorbed dose calculations are performed in two virtual patient models using the egs_brachy MC code: (1) TG-43-based D-w,w-TG(43), (2) D-w,D-w-MBDC that accounts for interseed scattering and attenuation (ISA), (3) D-m,D-m that examines ISA and tissue heterogeneity by scoring absorbed dose in tissue, and (4) D-w,D-m that unlike D-m,D-m scores absorbed dose in water. The MC absorbed doses (1) and (2) are simulated in a TG-43 patient phantom derived by assigning the densities of every voxel to 1.00 g cm(-3) (water), whereas MC absorbed doses (3) and (4) are scored in the TG-186 patient phantom generated by mapping the mass density of each voxel to tissue according to a CT calibration curve. The MC absorbed doses calculated in this study are compared with VariSeed v8.0 calculated absorbed doses. To evaluate the dosimetric effect of MAR and TAS, the MC absorbed doses of this work (independent of MAR and TAS) are compared to the MC absorbed doses of different I-125 source models from previous studies that were calculated with different MC codes using post-implant CT-based phantoms generated by implementing MAR and TAS on post-implant CT images. Results The very good agreement of TG-43 parameters of this study and the published consensus data within 3% validates the geometry of the IsoSeed I25.S17plus source. For the clinical studies, the TG-43-based calculations show a D-90 overestimation of more than 4% compared to the more realistic MC methods due to ISA and tissue composition. The results of this work generally show few discrepancies with the post-implant CT-based dosimetry studies with respect to the D-90 absorbed dose metric parameter. These discrepancies are mainly Type B uncertainties due to the different I-125 source models and MC codes. Conclusions The implementation of MAR and TAS on post-implant CT images have no dosimetric effect on the I-125 prostate MC absorbed dose calculation in post-implant CT-based phantoms.
|
|
|
|
Garani, R., & Palomares-Ruiz, S. (2022). Evaporation of dark matter from celestial bodies. J. Cosmol. Astropart. Phys., 05(5), 042–53pp.
Abstract: Scatterings of galactic dark matter (DM) particles with the constituents of celestial bodies could result in their accumulation within these objects. Nevertheless, the finite temperature of the medium sets a minimum mass, the evaporation mass, that DM particles must have in order to remain trapped. DM particles below this mass are very likely to scatter to speeds higher than the escape velocity, so they would be kicked out of the capturing object and escape. Here, we compute the DM evaporation mass for all spherical celestial bodies in hydrostatic equilibrium, spanning the mass range [10(-)(10) – 10(2)] M-circle dot, for constant scattering cross sections and s-wave annihilations. We illustrate the critical importance of the exponential tail of the evaporation rate, which has not always been appreciated in recent literature, and obtain a robust result: for the geometric value of the scattering cross section and for interactions with nucleons, at the local galactic position, the DM evaporation mass for all spherical celestial bodies in hydrostatic equilibrium is approximately given by E-c/T-chi similar to 30, where E-c is the escape energy of DM particles at the core of the object and T-chi is their temperature. In that case, the minimum value of the DM evaporation mass is obtained for super-Jupiters and brown dwarfs, m(ev)(ap) similar or equal to 0.7 GeV. For other values of the scattering cross section, the DM evaporation mass only varies by a factor smaller than three within the range 10(-41) cm(2) <= sigma(p) <= 10(-31) cm(2), where sigma(p) is the spin-independent DM-nucleon scattering cross section. Its dependence on parameters such as the galactic DM density and velocity, or the scattering and annihilation cross sections is only logarithmic, and details on the density and temperature profiles of celestial bodies have also a small impact.
|
|
|
|
Otten, S., Caron, S., de Swart, W., van Beekveld, M., Hendriks, L., van Leeuwen, C., et al. (2021). Event generation and statistical sampling for physics with deep generative models and a density information buffer. Nat. Commun., 12(1), 2985–16pp.
Abstract: Simulating nature and in particular processes in particle physics require expensive computations and sometimes would take much longer than scientists can afford. Here, we explore ways to a solution for this problem by investigating recent advances in generative modeling and present a study for the generation of events from a physical process with deep generative models. The simulation of physical processes requires not only the production of physical events, but to also ensure that these events occur with the correct frequencies. We investigate the feasibility of learning the event generation and the frequency of occurrence with several generative machine learning models to produce events like Monte Carlo generators. We study three processes: a simple two-body decay, the processes e(+)e(-)-> Z -> l(+)l(-) and pp -> tt<mml:mo><overbar></mml:mover> including the decay of the top quarks and a simulation of the detector response. By buffering density information of encoded Monte Carlo events given the encoder of a Variational Autoencoder we are able to construct a prior for the sampling of new events from the decoder that yields distributions that are in very good agreement with real Monte Carlo events and are generated several orders of magnitude faster. Applications of this work include generic density estimation and sampling, targeted event generation via a principal component analysis of encoded ground truth data, anomaly detection and more efficient importance sampling, e.g., for the phase space integration of matrix elements in quantum field theories. Here, the authors report buffered-density variational autoencoders for the generation of physical events. This method is computationally less expensive over other traditional methods and beyond accelerating the data generation process, it can help to steer the generation and to detect anomalies.
|
|
|
|
KM3NeT Collaboration(Aiello, S. et al), Alves Garre, S., Calvo, D., Carretero, V., Colomer, M., Corredoira, I., et al. (2020). Event reconstruction for KM3NeT/ORCA using convolutional neural networks. J. Instrum., 15(10), P10005–39pp.
Abstract: The KM3NeT research infrastructure is currently under construction at two locations in the Mediterranean Sea. The KM3NeT/ORCA water-Cherenkov neutrino detector off the French coast will instrument several megatons of seawater with photosensors. Its main objective is the determination of the neutrino mass ordering. This work aims at demonstrating the general applicability of deep convolutional neural networks to neutrino telescopes, using simulated datasets for the KM3NeT/ORCA detector as an example. To this end, the networks are employed to achieve reconstruction and classification tasks that constitute an alternative to the analysis pipeline presented for KM3NeT/ORCA in the KM3NeT Letter of Intent. They are used to infer event reconstruction estimates for the energy, the direction, and the interaction point of incident neutrinos. The spatial distribution of Cherenkov light generated by charged particles induced in neutrino interactions is classified as shower- or track-like, and the main background processes associated with the detection of atmospheric neutrinos are recognized. Performance comparisons to machine-learning classification and maximum-likelihood reconstruction algorithms previously developed for KM3NeT/ORCA are provided. It is shown that this application of deep convolutional neural networks to simulated datasets for a large-volume neutrino telescope yields competitive reconstruction results and performance improvements with respect to classical approaches.
|
|
|
|
LHCb Collaboration(Aaij, R. et al), Jashal, B. K., Martinez-Vidal, F., Oyanguren, A., Remon Alepuz, C., & Ruiz Vidal, J. (2022). Evidence for a New Structure in the J/psi p and J/psi(p)over-bar Systems in B-s(0) -> J/psi p(p)over-bar Decays. Phys. Rev. Lett., 128(6), 062001–11pp.
Abstract: An amplitude analysis of flavor-untagged B-s(0) -> J=psi p (p) over bar decays is performed using a sample of 797 +/- 31 decays reconstructed with the LHCb detector. The data, collected in proton-proton collisions between 2011 and 2018, correspond to an integrated luminosity of 9 fb(-1). Evidence for a new structure in the J=psi p and J=psi(p) over bar systems with a mass of 4337(-4-2)(+7+2) MeV and a width of 29(-12-14)(+26+14) MeV is found, where the first uncertainty is statistical and the second systematic, with a significance in the range of 3.1 to 3.7 sigma, depending on the assigned J(P) hypothesis.
|
|
|
|
ATLAS Collaboration(Aad, G. et al), Alvarez Piqueras, D., Aparisi Pozo, J. A., Bailey, A. J., Cabrera Urban, S., Castillo, F. L., et al. (2020). Evidence for electroweak production of two jets in association with a Z gamma pair in pp collisions at root S=13 TeV with the ATLAS detector. Phys. Lett. B, 803, 135341–23pp.
Abstract: Evidence for electroweak production of two jets in association with a Z gamma pair in root s = 13 TeV proton-proton collisions at the Large Hadron Collider is presented. The analysis uses data collected by the ATLAS detector in 2015 and 2016 that corresponds to an integrated luminosity of 36.1 fb(-1). Events that contain a Z boson candidate decaying leptonically into either e(+)e(-) or mu(+)mu(-), a photon, and two jets are selected. The electroweak component is measured with observed and expected significances of 4.1 standard deviations. The fiducial cross-section for electroweak production is measured to be sigma(Z gamma jj-Ew) = 7.8 +/- 2.0 fb, in good agreement with the Standard Model prediction.
|
|
|
|
AGATA Collaboration(Liu, X. et al), Gadea, A., Jurado, M., Domingo-Pardo, C., Huyuk, T., & Perez-Vidal, R. M. (2021). Evidence for enhanced neutron-proton correlations from the level structure of the N = Z+1 nucleus Tc-87(43)44. Phys. Rev. C, 104(2), L021302–5pp.
Abstract: The low-lying excited states in the neutron-deficient N = Z + 1 nucleus (87)(43)Tcc(44) have been studied via the fusion-evaporation reaction Fe-54(Ar-36, 2n1p)Tc-87 at the Grand Accelerateur National d'Ions Lourds (GANIL), France. The AGATA spectrometer was used in conjunction with the auxiliary NEDA, Neutron Wall, and DIAMANT detector arrays to measure coincident prompt gamma rays, neutrons, and charged particles emitted in the reaction. A level scheme of Tc-87 from the (9/2(g.s.)(+)) state to the (33/2(1)(+)) state was established based on six mutually coincident gamma-ray transitions. The constructed level structure exhibits a rotational behavior with a sharp backbending at (h) over bar omega approximate to 0.50 MeV. A decrease in alignment frequency and increase in alignment sharpness in the odd-mass isotonic chains around N = 44 is proposed as an effect of the enhanced isoscalar neutron-proton interactions in odd-mass nuclei when approaching the N = Z line.
|
|
|
|
ATLAS Collaboration(Aad, G. et al), Aparisi Pozo, J. A., Bailey, A. J., Cabrera Urban, S., Cardillo, F., Castillo, F. L., et al. (2021). Evidence for Higgs boson decays to a low-mass dilepton system and a photon in pp collisions at root s=13 TeV with the ATLAS detector. Phys. Lett. B, 819, 136412–22pp.
Abstract: A search for the Higgs boson decaying into a photon and a pair of electrons or muons with an invariant mass m(ll) < 30 GeV is presented. The analysis is performed using 139 fb(-1) of proton-proton collision data, produced by the LHC at a centre-of-mass energy of 13 TeV and collected by the ATLAS experiment. Evidence for the H -> ll(gamma) process is found with a significance of 3.2 over the background-only hypothesis, compared to an expected significance of 2.1 for the Standard Model prediction. The best-fit value of the signal-strength parameter, defined as the ratio of the observed signal yield to the one expected in the Standard Model, is μ= 1.5 +/- 0.5. The Higgs boson production cross-section times the H -> ll(gamma) branching ratio for m(ll) < 30 GeV is determined to be 8.7(-2.7)(+2.8) fb.
|
|
|
|
LHCb Collaboration(Aaij, R. et al), Jaimes Elles, S. J., Jashal, B. K., Martinez-Vidal, F., Oyanguren, A., Rebollo De Miguel, M., et al. (2023). Evidence for Modification of b Quark Hadronization in High-Multiplicity pp Collisions at √s=13 TeV. Phys. Rev. Lett., 131(6), 061901–11pp.
Abstract: The production rate of B-s(0) mesons relative to B-0 mesons is measured by the LHCb experiment in pp collisions at a center-of-mass energy root s = 13 TeV over the forward rapidity interval 2 < y < 4.5 as a function of the charged particle multiplicity measured in the event. Evidence at the 3.4 sigma level is found for an increase of the ratio of B-s(0) to B-0 cross sections with multiplicity at transverse momenta below 6 GeV=c, with no significant multiplicity dependence at higher transverse momentum. Comparison with data from e(+)e(-) collisions implies that the density of the hadronic medium may affect the production rates of B mesons. This is qualitatively consistent with the emergence of quark coalescence as an additional hadronization mechanism in high-multiplicity collisions.
|
|
