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.

Protection of C3H/He mice from experimental Borrelia burgdorferi infection by immunization with a 110-kilodalton fusion protein.

A 110-kDa Borrelia burgdorferi fusion protein, Escherichia coli expressing the fusion protein, transformed E. coli lacking the fusion protein insert, and lyophilized whole B. burgdorferi bacteria were compared for immunogenicity in C3H/He mice. Immunized mice were challenged with a variety of isolates from the United States or the European isolate P/Gau 3 weeks following the last inoculation. An average of 76.7% of the mice immunized with 25 micrograms of lyophilized whole B. burgdorferi cells were protected from infection, while 60% of the mice immunized with the 110-kDa fusion protein were protected. Whole E. coli bacteria expressing the fusion protein protected 57.7% of immunized mice against experimental challenge. Lower levels of protection occurred in mice challenged with the European isolate than in those challenged with isolates originating from the United States. These results demonstrate the potential of the 110-kDa fusion protein for use as a component of a subunit vaccine for prevention of Lyme borreliosis.

Class specific antibody response to influenza A H1N1 infection in swine.

Early detection of swine influenza A outbreaks is essential to understand the true cause and effect relationship that exists between this disease and other serious respiratory or herd health problems. Enzyme-linked immunosorbent assays (ELISAs) for the early detection of H1N1 subtype specific serum IgM, IgG and secretory IgA were compared to direct virus detection in in embryonated eggs. Elevated levels of H1 hemagglutinin (HA) specific IgM and IgG were detected as early as 3 days post experimental infection with a field strain of swine influenza A (H1N1). Influenza specific IgA in nasal mucous samples was detected on day 4 post infection (PI). This compared favorably with egg inoculation methods which detected virus 2-4 days PI. Identification of elevated H1 HA specific IgM in test herds could signify a recent influenza outbreak. Alternatively, ELISA analysis of nasal mucous samples for H1 HA specific IgA could provide a noninvasive method of obtaining similar information on the influenza specific immune status of the herd.

Nerve conduction and aldose reductase inhibition during 5 years of diabetes or galactosaemia in dogs.

To evaluate the role of excessive polyol pathway activity in the pathogenesis of nerve disorders in diabetes mellitus, nerve conduction velocity was measured in motor nerves of diabetic dogs given an aldose reductase inhibitor (Sorbinil) or placebo, and also in non-diabetic dogs made experimentally galactosaemic. The nerve conduction velocity slowly declined in the diabetic placebo group, becoming significantly less than normal by the fifth year of the study, and the decline was prevented by administration of the aldose reductase inhibitor. Non-diabetic dogs made galactosaemic by consuming a 30% galactose diet developed erythrocyte and nerve polyol concentrations many times greater than that of diabetic or normal animals, but the nerve conduction velocity remained normal throughout 5 years of study. These results in dogs suggest that aldose reductase inhibitors may prevent defective nerve conduction in long-term diabetes, and raise the possibility that excessive accumulation of polyol itself is not sufficient to produce the nerve defect in the absence of excessive polyol utilization.

Nerve conduction velocity in dogs is reduced by diabetes and not by galactosemia.

To evaluate the role of hyperglycemia and excessive polyol pathway activity in the pathogenesis of nerve disorders in diabetes, motor nerve conduction velocity (MNCV) was measured in dogs alloxan diabetic or experimentally galactosemic for 5 years. Diabetic dogs in poor glycemic control showed a progressive decline of MNCV from baseline values. Diabetic dogs that had been randomly assigned to good glycemic control retained normal MNCV. Nondiabetic dogs made galactosemic by a 30% galactose diet developed erythrocyte polyol concentrations many-fold greater than in diabetic animals, but the MNCV remained unchanged and comparable to that of normal dogs. Nerve polyol levels, when compared in short-term diabetic dogs or dogs galactose-fed 2 to 4 months, were elevated at least as much by the galactose-rich diet as by diabetes. Thus, in dogs, excessive tissue polyol accumulation is associated with subnormal MNCV in diabetes, but not in experimental galactosemia.