Chances and challenges of machine learning-based disease classification in genetic association studies illustrated on age-related macular degeneration.

Imaging technology and machine learning algorithms for disease classification set the stage for high-throughput phenotyping and promising new avenues for genome-wide association studies (GWAS). Despite emerging algorithms, there has been no successful application in GWAS so far. We establish machine learning-based phenotyping in genetic association analysis as misclassification problem. To evaluate chances and challenges, we performed a GWAS based on automatically classified age-related macular degeneration (AMD) in UK Biobank (images from 135,500 eyes; 68,400 persons). We quantified misclassification of automatically derived AMD in internal validation data (4,001 eyes; 2,013 persons) and developed a maximum likelihood approach (MLA) to account for it when estimating genetic association. We demonstrate that our MLA guards against bias and artifacts in simulation studies. By combining a GWAS on automatically derived AMD and our MLA in UK Biobank data, we were able to dissect true association (ARMS2/HTRA1, CFH) from artifacts (near HERC2) and identified eye color as associated with the misclassification. On this example, we provide a proof-of-concept that a GWAS using machine learning-derived disease classification yields relevant results and that misclassification needs to be considered in analysis. These findings generalize to other phenotypes and emphasize the utility of genetic data for understanding misclassification structure of machine learning algorithms.

Aortic Atherosclerosis Determines Increased Retrograde Blood Flow as a Potential Mechanism of Retrograde Embolic Stroke.

BACKGROUND: Retrograde brain embolization from complex plaques of the proximal descending aorta (DAo) has been identified as a new potential mechanism of stroke. Our purpose was to identify predictors of increased retrograde aortic blood flow indicating an elevated risk of brain embolization from the DAo. METHODS: A total of 485 patients with acute ischemic stroke were prospectively included and underwent transesophageal echocardiography. Blood flow velocities in the proximal DAo were studied using 2D pulse-wave Doppler ultrasound. Velocity-time integrals (VTI) were calculated for antegrade and retrograde velocity directions. The ratio (VTIretrograde/VTIantegrade) was used to estimate retrograde flow extent. Associations between patient demographics, cardiovascular risk factors, echocardiographic parameters, and VTIratio were analyzed using multivariate linear regression. RESULTS: Retrograde blood flow in the DAo occurred in all patients. Velocity profiles in the proximal DAo were as follows (mean ± SD): VTIantegrade = 21.1 ± 6.5, VTIretrograde = 11.0 ± 3.6, and VTIratio = 0.54 ± 0.16. Diameter (r = 0.25, p < 0.001), presence of complex plaques (r = 0.12, p = 0.007), and reduced strain of the DAo (r = -0.23, p < 0.001) had significant partial effects in a predictor model based on predefined variables, which predicted 26% (adjusted R2 = 0.26) of the variance in VTIratio. A unit increase in the DAo diameter was associated with a 2% increase in VTIratio (95% CI 1-2.8%, p < 0.001). Presence of complex plaques increased VTIratio by 7% (95% CI 2-13%, p = 0.007) and an increase in strain by 0.1 indicated a decrease in VTIratio by about 11% (95% CI 6.2-15.5%, p < 0.001). Complex atheroma was found in the proximal DAo of 79 subjects, of which 40 (50.6%) had a VTIratio above average (VTIratio ≥0.54) compared to 87 of 261 (33.3%) patients without any complex plaques (p < 0.001). Twenty-five of 79 (31.7%) patients with complex DAo plaques had a VTIratio ≥0.60, which indicates a high likelihood of retrograde pathline length of ≥3 cm and thus increased risk of retrograde cerebral embolization. Stroke etiology of those 25 patients was determined in 13 and cryptogenic in 12 cases. CONCLUSIONS: Retrograde blood flow in the DAo was found in all stroke patients. However, it increased further in patients with concomitant complex plaques, low strain, and/or large aortic diameter, that is, in those with atherosclerosis of the DAo. Accordingly, such patients may be predisposed to retrograde embolization in case of occurrence of a complex plaque in proximity to a brain-supplying artery.

Remifentanil for abdominal surgery is associated with unexpectedly unfavorable outcomes.

Insufficient perioperative pain treatment is known as a highly predictive risk factor for the development of chronic postoperative pain. Remifentanil is an ultrashort-acting opioid that provides quick and efficient analgesia but is associated with the induction of opioid-induced hyperalgesia. Despite these well-known characteristics, this substance is being increasingly used in anesthesia and in a variety of medical fields, such as intensive-care medicine and obstetrics. The aim of our study was to reveal whether remifentanil influences postoperative pain, the requirement for postoperative analgesics, and requirement of antiemetics (as indirect indicator of postoperative nausea and vomiting), as well as the effects on time to extubation and length of stay in the postanesthesia care unit in daily clinical routine. From an electronic medical records database of 55,693 anesthesias, we analyzed data from all patients receiving intraabdominal surgery (visceral, gynecological, and urological) under general anesthesia or combined general-epidural anesthesia by propensity score matching. The administration of remifentanil was associated with higher postoperative pain scores despite a higher requirement of postoperative analgesics. Additional epidural analgesia was not able to avoid this finding. The intraoperative use of remifentanil is associated with a deterioration of pain levels and postoperative analgesic requirement, wherefore the potential benefit of this substance seems to be outweighed by its potential disadvantages. Especially in operative procedures in which high postoperative pain scores are expected, the unreflective use should be critically questioned.

Rupture of an aneurysm of the noncoronary sinus of Valsalva into the right atrium.

A 32-year-old woman presented to the emergency department after the sudden onset of palpitations, dyspnoea and left-sided tinnitus. Echocardiography revealed a ruptured aneurysm of the noncoronary sinus of Valsalva into the right atrium. Due to deterioration of right heart failure with rapidly progressive decline of systemic blood pressure, the patient underwent immediate surgery with a patch repair of the ruptured aneurysm. A few days later, the patient was discharged home in good condition.

Measurement of cardiac valve and aortic blood flow velocities in stroke patients: a comparison of 4D flow MRI and echocardiography.

4D flow MRI is an emerging technique that allows quantification of 3D blood flow in vivo. However, comparisons with methods of blood velocity quantification used in clinical routine are sparse. Therefore, we compared velocity quantification using 4D flow MRI with transthoracic and transesophageal echocardiography at the mitral and aortic valves and the aorta. Forty-eight stroke patients (age 67.3 ± 15.0 years) were examined by 4D flow MRI. Blood flow velocities were assessed using standardized 2D analysis planes positioned in the mitral valve (MV), aortic valve (AV), ascending aorta (AAo), and descending aorta (DAo) and were compared with echocardiography. MRI showed moderate-high correlations of systolic velocity values for the MV (r = 0.67, p < 0.001), AV (r = 0.77, p < 0.001), AAo (r = 0.93, p < 0.001), and DAo (r = 0.76, p < 0.001) along with moderate-high intraclass-correlation-coefficients: MV 0.79 (95% CI 0.62, 0.88), AV 0.86 (95% CI 0.75, 0.92), AAo 0.96 (95% CI 0.93, 0.98), and DAo 0.83 (95% CI 0.70, 0.90). However, MRI underestimated absolute systolic blood flow velocities compared with echocardiography by 8.6% for the MV (p = 0.07), 3.1% for the AV (p = 0.48), 10.7% for the AAo (p = 0.09), and 15.0% for the DAo (p = 0.01). Blood flow velocities obtained using 4D flow MRI and echocardiography at the MV, AV, and the ascending and DAo showed moderate to high correlations. Underestimation of absolute velocity values by MRI was low. Thus, 4D flow MRI seems ideally suited to comprehensively assess cardiac and aortic pathologies and related hemodynamic changes in future studies.

A molecular intravascular ultrasound contrast agent allows detection of activated platelets on the surface of symptomatic human plaques.

BACKGROUND AND AIMS: Activated platelets are amongst the most attractive imaging targets in atherosclerosis due to their important role in early processes of atherogenesis and thrombus formation. We developed a molecular intravascular ultrasound (IVUS) approach to detect activated platelets ex vivo on the surface of human plaques, using an IVUS system applied in clinical routine. METHODS: Human carotid endarterectomy specimens were obtained directly from the operating room and exposed to artificial arterial flow conditions for incubation with the contrast agent. This consists of microbubbles (MB), which are linked to an antibody against the ligand induced binding site (LIBS) of the activated platelet glycoprotein IIb/IIa, and a sialyl Lewis polymer (SL), which mediates binding to selectins (LIBS-SL-MB). IVUS was performed pre and post incubation with LIBS-SL-MB and after rinsing with PBS. In comparison, IVUS was performed pre and post incubation with MBs linked to an unspecific control antibody and a dysfunctional polymer (control-MB). All imaging results were correlated to histology findings. RESULTS: IVUS imaging showed a high signal enhancement after administration of LIBS-SL-MB. After rinsing with PBS, the signal enhancement remained stable. Immunofluorescence and immunohistochemistry confirmed significant binding of microbubbles to thrombi on the plaque surface. Moreover, thrombus size and number of bound MBs correlated well. CONCLUSIONS: LIBS-SL-MB allows ex vivo IVUS imaging of even small numbers of activated platelets on the surface of human carotid endarterectomy specimens. This diagnostic approach could deliver valuable additional information for risk stratification of atherosclerotic plaques, especially since we apply a clinically well-established IVUS imaging system.

An ultrasound contrast agent targeted to P-selectin detects activated platelets at supra-arterial shear flow conditions.

OBJECTIVES: To evaluate targeting of a microbubble contrast agent to platelets under high shear flow using the natural selectin ligand sialyl Lewis. MATERIALS AND METHODS: Biotinylated polyacrylamide Sialyl Lewis or biotinylated carbohydrate-free polymer (used as a control) were attached to biotinylated microbubbles via a streptavidin linker. Activated human platelets were isolated and attached to fibrinogen-coated culture dishes. Fibrinogen-coated dishes without platelets or platelet dishes blocked by an anti-P-selectin antibody served as negative control substrates. Dishes coated by recombinant P-selectin served as a positive control substrate. Microbubble adhesion was assessed by microscopy in an inverted parallel plate flow chamber, with wall shear stress values of 40, 30, 20, 10, and 5 dynes/cm2. The ratio of binding and passing microbubbles was defined as capture efficiency. RESULTS: There was no significant difference between the groups regarding the number of microbubbles in the fluid flow at each shear rate. Sialyl Lewis-targeted microbubbles were binding and slowly rolling on the surface of activated platelets and P-selectin-coated dishes at all the flow conditions including 40 dynes/cm2. Capture efficiency of targeted microbubbles to activated platelets and recombinant P-selectin decreased with increasing shear flow: at 5 dynes/cm2, capture efficiency was 16.11% on activated platelets versus 21.83% on P-selectin, and, at 40 dynes/cm2, adhesion efficiency was still 3.4% in both groups. There was neither significant adhesion of Sialyl Lewis-targeted microbubbles to control substrates, nor adhesion of control microbubbles to activated platelets or to recombinant P-selectin. CONCLUSIONS: Microbubble targeting using sialyl Lewis, a fast-binding ligand to P-selectin, is a promising strategy for the design of ultrasound contrast binding to activated platelets under high shear stress conditions.