Artificial Intelligence Applied to Chest X-ray: A Reliable Tool to Assess the Differential Diagnosis of Lung Pneumonia in the Emergency Department.

BACKGROUND: Considering the large number of patients with pulmonary symptoms admitted to the emergency department daily, it is essential to diagnose them correctly. It is necessary to quickly solve the differential diagnosis between COVID-19 and typical bacterial pneumonia to address them with the best management possible. In this setting, an artificial intelligence (AI) system can help radiologists detect pneumonia more quickly. METHODS: We aimed to test the diagnostic performance of an AI system in detecting COVID-19 pneumonia and typical bacterial pneumonia in patients who underwent a chest X-ray (CXR) and were admitted to the emergency department. The final dataset was composed of three sub-datasets: the first included all patients positive for COVID-19 pneumonia (n = 1140, namely "COVID-19+"), the second one included all patients with typical bacterial pneumonia (n = 500, "pneumonia+"), and the third one was composed of healthy subjects (n = 1000). Two radiologists were blinded to demographic, clinical, and laboratory data. The developed AI system was used to evaluate all CXRs randomly and was asked to classify them into three classes. Cohen's κ was used for interrater reliability analysis. The AI system's diagnostic accuracy was evaluated using a confusion matrix, and 95%CIs were reported as appropriate. RESULTS: The interrater reliability analysis between the most experienced radiologist and the AI system reported an almost perfect agreement for COVID-19+ (κ = 0.822) and pneumonia+ (κ = 0.913). We found 96% sensitivity (95% CIs = 94.9-96.9) and 79.8% specificity (76.4-82.9) for the radiologist and 94.7% sensitivity (93.4-95.8) and 80.2% specificity (76.9-83.2) for the AI system in the detection of COVID-19+. Moreover, we found 97.9% sensitivity (98-99.3) and 88% specificity (83.5-91.7) for the radiologist and 97.5% sensitivity (96.5-98.3) and 83.9% specificity (79-87.9) for the AI system in the detection of pneumonia+ patients. Finally, the AI system reached an accuracy of 93.8%, with a misclassification rate of 6.2% and weighted-F1 of 93.8% in detecting COVID+, pneumonia+, and healthy subjects. CONCLUSIONS: The AI system demonstrated excellent diagnostic performance in identifying COVID-19 and typical bacterial pneumonia in CXRs acquired in the emergency setting.

Accelerated molecular dynamics simulation of thin-film growth with the bond-boost method.

We review the bond-boost method for accelerated molecular dynamics (MD) simulation and we demonstrate its application to kinetic phenomena relevant to thin-film growth. To illustrate various aspects of the method, three case studies are presented. We first illustrate aspects of the bond-boost method in studies of the diffusion of Cu atoms on Cu(001). In these studies, Cu interactions are described using a semi-empirical embedded-atom method potential. We recently extended the bond-boost method to perform accelerated ab initio MD (AIMD) simulations and we present results from preliminary studies in which we applied the bond-boost method in AIMD to uncover diffusion mechanisms of Al adatoms on Al(110). Finally, a problem inherent to many rare-event simulation methods is the 'small-barrier problem', in which the system resides in a group of states connected by small energy barriers and separated from the rest of phase space by large barriers. We developed the state-bridging bond-boost method to address this problem and we discuss its application for studying the diffusion of Co clusters on Cu(001). We discuss the outlook for future applications of the bond-boost method in materials simulation.

Multiple-time scale accelerated molecular dynamics: addressing the small-barrier problem.

We present a method for accelerated molecular-dynamics simulation in systems with rare-event dynamics that span a wide range of time scales. Using a variant of hyperdynamics, we detect, on the fly, groups of recurrent states connected by small energy barriers and we modify the potential-energy surface locally to consolidate them into large, coarse states. In this way, fast motion between recurrent states is treated within an equilibrium formalism and dynamics can be simulated over the longer time scale of the slow events. We apply the method to simulate cluster diffusion and the initial growth of Co on Cu(001),where time scales spanning more than 6 orders of magnitude are present, and show that the method correctly follows the slow events, so that much larger times can be simulated than with accelerated molecular dynamics alone.

Thermal desorption of large molecules from solid surfaces.

We use molecular-dynamics simulations and importance sampling to obtain transition-state-theory rate constants for thermal desorption of an n-alkane series from Au(111). We find that the binding of a large molecule to a solid surface involves different types of local minima. The preexponential factors increase with increasing chain length and can be substantially larger than typical estimates for small molecules. Our results match recent experimental studies and indicate that a proper treatment of conformational isomerism and entropy, heretofore not found in coarse-grained models, is essential to quantitatively describe the thermal desorption of large molecules from solid surfaces.