Diagnostic Accuracy and Failure Mode Analysis of a Deep Learning Algorithm for the Detection of Cervical Spine Fractures.

BACKGROUND AND PURPOSE: Artificial intelligence decision support systems are a rapidly growing class of tools to help manage ever-increasing imaging volumes. The aim of this study was to evaluate the performance of an artificial intelligence decision support system, Aidoc, for the detection of cervical spinal fractures on noncontrast cervical spine CT scans and to conduct a failure mode analysis to identify areas of poor performance. MATERIALS AND METHODS: This retrospective study included 1904 emergent noncontrast cervical spine CT scans of adult patients (60 [SD, 22] years, 50.3% men). The presence of cervical spinal fracture was determined by Aidoc and an attending neuroradiologist; discrepancies were independently adjudicated. Algorithm performance was assessed by calculation of the diagnostic accuracy, and a failure mode analysis was performed. RESULTS: Aidoc and the neuroradiologist's interpretation were concordant in 91.5% of cases. Aidoc correctly identified 67 of 122 fractures (54.9%) with 106 false-positive flagged studies. Diagnostic performance was calculated as the following: sensitivity, 54.9% (95% CI, 45.7%-63.9%); specificity, 94.1% (95% CI, 92.9%-95.1%); positive predictive value, 38.7% (95% CI, 33.1%-44.7%); and negative predictive value, 96.8% (95% CI, 96.2%-97.4%). Worsened performance was observed in the detection of chronic fractures; differences in diagnostic performance were not altered by study indication or patient characteristics. CONCLUSIONS: We observed poor diagnostic accuracy of an artificial intelligence decision support system for the detection of cervical spine fractures. Many similar algorithms have also received little or no external validation, and this study raises concerns about their generalizability, utility, and rapid pace of deployment. Further rigorous evaluations are needed to understand the weaknesses of these tools before widespread implementation.

Rapid diffusion of magic-size islands by combined glide and vacancy mechanism.

Using molecular dynamics, nudged elastic band, and embedded atom methods, we show that certain 2D Ag islands undergo extremely rapid one-dimensional diffusion on Cu(001) surfaces. Indeed, below 300 K, hopping rates for "magic-size" islands are orders of magnitude faster than hopping rates for single Ag adatoms. This rapid diffusion requires both the c(10 x 2) hexagonally packed superstructure typical of Ag on Cu(001) and appropriate "magic sizes" for the islands. The novel highly cooperative diffusion mechanism presented here couples vacancy diffusion with simultaneous core glide.

Direct transformation of vacancy voids to stacking fault tetrahedra.

Defect accumulation is the principal factor leading to the swelling and embrittlement of materials during irradiation. It is commonly assumed that, once defect clusters nucleate, their structure remains essentially constant while they grow in size. Here, we describe a new mechanism, discovered during accelerated molecular dynamics simulations of vacancy clusters in fcc metals, that involves the direct transformation of a vacancy void to a stacking fault tetrahedron (SFT) through a series of 3D structures. This mechanism is in contrast with the collapse to a 2D Frank loop which then transforms to an SFT. The kinetics of this mechanism are characterized by an extremely large rate prefactor, tens of orders of magnitude larger than is typical of atomic processes in fcc metals.

Closing the gap between experiment and theory: crystal growth by temperature accelerated dynamics.

We present atomistic simulations of crystal growth where realistic experimental deposition rates are reproduced, without needing any a priori information on the relevant diffusion processes. Using the temperature accelerated dynamics method, we simulate the deposition of 4 monolayers (ML) of Ag/Ag(100) at the rate of 0.075 ML/s, thus obtaining a boost of several orders of magnitude with respect to ordinary molecular dynamics. In the temperature range analyzed (0-70 K), steering and activated mechanisms compete in determining the surface roughness.

Structure and mobility of defects formed from collision cascades in MgO.

We study radiation-damage events in MgO on experimental time scales by augmenting molecular dynamics cascade simulations with temperature accelerated dynamics, molecular statics, and density functional theory. At 400 eV, vacancies and mono- and di-interstitials form, but often annihilate within milliseconds. At 2 and 5 keV, larger clusters can form and persist. While vacancies are immobile, interstitials aggregate into clusters (In) with surprising properties; e.g., an I4 is immobile, but an impinging I2 can create a metastable I6 that diffuses on the nanosecond time scale but is stable for years.