Influence of leg axis alignment on MRI T2* mapping of the knee in young professional soccer players

BACKGROUND

Investigation of the association between leg axis alignment and biochemical MRI in young professional soccer players in order to identify a potential influence of the leg axis on cartilage regions at risk.

METHODS

Sixteen professional soccer players (21 ± 3 years) underwent static and dynamic leg axis analysis via radiation free DIERS formetric 4 D as well as 3-T MRI examination of both knees. Quantitative T2* mapping of the knee cartilage was performed and T2* values were evaluated as 144 regions of interest. Subgroup analysis was performed in players with severe varus alignment (> 6°).

RESULTS

Analysis of the leg axis geometry revealed a mean static alignment of 6.6° ± 2.5 varus and a mean dynamic alignment of 5.1° ± 2.6 varus. Quantitative T2* mapping showed significantly increased T2* values in the superficial cartilage layer compared to the deeper region (p < 0.001) as well as a significant increase in relaxation times in the femoral cartilage from anterior to intermediate to posterior (p < 0.001). Combination of both methods revealed a significant correlation for the degree of varus alignment and the femoral, posterior, deep region of the medial knee compartment (r = 0.4; p = 0.03). If severe varus alignment was present this region showed a significant increase in relaxation time compared to players with a less pronounced leg axis deviation (p = 0.003).

CONCLUSION

This study demonstrates that varus alignment in young soccer players is associated with elevated T2* relaxation times in the deep cartilage layer of the medial, posterior, femoral compartment and might therefore be a contributing factor in the early pathogenesis of manifest cartilage lesions. Therefore, these findings should be considered in the development of preventive training programs.

Acromioclavicular joint suture button repair leads to coracoclavicular tunnel widening

PURPOSE

Biomechanical evaluation of three different suture button devices used in acromioclavicular joint repair and analysis of their effect on post-testing tunnel widening.

METHODS

Eighteen human shoulder girdles were assigned into three groups with a similar mean bone mineral density. Three different single-tunnel acromioclavicular repair devices were tested: (1) AC TightRope^® with FiberWire; (2) AC Dog Bone^™ Button with FiberTape; (3) Low Profile AC Repair System. Biomechanical testing was performed simulating the complex movement of the distal clavicle as follows. A vertical load of 80 N was applied continuously. The rotation of the clavicle about its long axis was set at 10° anterior and 30° posterior for 2500 cycles at 0.25 Hz. The horizontal translation of the clavicle was set at 6 mm medial and 6 mm lateral for 10,000 cycles at 1 Hz. The coracoclavicular distance was measured before and after testing. After testing, each sample underwent micro-CT analysis. Following 3D reconstruction, the area of the bone tunnels was measured at five defined cross sections.

RESULTS

In TightRope^® and Dog Bone^™ groups, all samples completed testing, whereas in the Low Profile group, three out of six samples showed system failure. The mean absolute difference of coracoclavicular distance after testing was significantly greater in the Low Profile group compared to TightRope^® and Dog Bone^™ groups (4.3 ± 1.3 mm vs 1.9 ± 0.7 mm vs 1.9 ± 0.8 mm; p = 0.001). Micro-CT analysis of the specimens demonstrated significant tunnel widening in the inferior clavicular and superior coracoid regions in all three groups (p < 0.05).

CONCLUSION

Significant tunnel widening can be observed for all devices and is primarily found in the inferior parts of the clavicle and superior parts of the coracoid. The Low Profile AC Repair System showed inferior biomechanical properties compared to the AC TightRope^® and AC Dog Bone^™ devices. Therefore, clinicians should carefully select the type of acromioclavicular repair device used and need to consider tunnel widening as a complication.

Validation of an electrocardiographic algorithm for the detection of cardiac amyloidosis

Background

Despite new therapies, diagnosis of cardiac amyloidosis (CA) is often delayed. We recently developed a simple electrocardiographic (ECG) algorithm to suspect CA without the aid of advanced imaging modalities (Figure).

Methods

The aim of this study was to validate the algorithms' usefulness in clinical practice. ECG readings from patients with CA, heart failure with preserved ejection fraction (HFpEF), and hypertrophic cardiomyopathy (HCMP) were analyzed in a blinded fashion.

Results

884 patients were included. Patients with pacemakers were excluded, leaving 827 ECGs (237 CA, 407 HFpEF, 183 HCMP) for final analysis. A characteristic pattern defined by the algorithm was visually perceptible in 165 ECGs (69.6%) of the amyloidosis patients vs. 114 (28%) of HFpEF vs. 22 (12.0%) of HCMP patients (p&lt;0.001). The area under the curve (AUC) for the detection CA was 0.75 with a sensitivity of 69.6% and a specificity of 76.9% (Figure). Binary logistic regression analysis revealed that the presence of a distinctive pattern increased the probability of CA with an odds ratio of 7.66 (CI: 5.47–10.72; p&lt;0.001).

Conclusion

This easy-to-use ECG algorithm has proven helpful to suspect CA. Our tool may significantly improve the treatment of heart failure patients by identifying those with amyloidosis-related disease.

Funding Acknowledgement

Type of funding sources: None.

The impact of left atrial mechanics on adverse events and clinical outcome after cardiac surgery

Objectives

Post-operative atrial fibrillation (POAF) represents a common complication after cardiac surgery that is associated with unfavorable clinical outcome. Identifying patients at risk for POAF is crucial but challenging.

Purpose

This study aimed to investigate the prognostic potential of speckle tracking echocardiography (STE) on POAF and fatal adverse events from a long-term perspective.

Methods

A total of 124 patients undergoing elective cardiac surgery were prospectively enrolled and underwent pre-operative STE. Patients were followed prospectively for the occurrence of POAF within the entire hospitalization and reaching the secondary endpoints cardiovascular (CV) and all-cause mortality.

Results

Within the study population 43.5% (n=53) of enrolled individuals developed POAF. After a median follow-up of 3.9 years, 25 (20.2%) patients died. We observed that patients presenting with POAF had lower global peak atrial longitudinal strain (PALS) values compared to the non-POAF arm (POAF: 14.8% [95% CI: 10.9–17.8] vs. non-POAF: 19.4% [95% CI: 14.8–23.5], p&lt;0.001). Moreover, global PALS was a strong and independent predictor for POAF (adjusted Odds Ratio per 1-SD: 0.37 [95% CI: 0.22–0.65], p&lt;0.001), and independently associated with mortality (adjusted Hazard Ratio per 1-SD: 0.63 [95% CI: 0.40–0.99], p=0.048). CART analysis revealed a cut-off value of &lt;17% global PALS as high-risk for both POAF and mortality.

Conclusion

Global PALS is associated with the development of POAF following surgery in an unselected patient population undergoing CABG and/or valve surgery. Since patients with global PALS &lt;17% face a poor long-term prognosis, routine assessment of global PALS needs to be considered in terms of proper secondary prevention in the era of personalized medicine.

Funding Acknowledgement

Type of funding sources: None.

Two-dimensional speckle-tracking echocardiography in tafamidis-treated patients with transthyretin amyloid cardiomyopathy: a glimmer of hope for viable therapy monitoring?

Background

Treatment with Tafamidis in patients with transthyretin amyloid cardiomyopathy (ATTR-CM) has been shown to have beneficial effects on the left ventricle (LV), as assessed by cardiac magnetic resonance (CMR) imaging. Although CMR represents the gold standard among imaging modalities, its limited availability in clinical practice makes it unfeasible for routine therapy monitoring.

Purpose

We aimed to determine Tafamidis-induced changes using two-dimensional (2D) speckle-tracking echocardiography and to identify echocardiographic imaging parameters that could be used for specific therapy monitoring.

Methods

We subjected a series of ATTR-CM patients to transthoracic echocardiography (TTE) at baseline and follow-up and compared patients treated with Tafamidis 61mg once daily (QD) to a historical control cohort treated with Tafamidis 20mg QD, as well as to a historical treatment-naïve control cohort reflecting the natural course of the disease.

Results

Patients receiving Tafamidis 61mg (n=62) or 20mg (n=21) QD showed stable measurements at follow-up [61mg: 8.5 months, 20mg: 7.0 months] in LV global longitudinal strain (GLS) (61mg: −11.75% vs. −11.58%, p=0.534; 20mg: −10.61% vs. −10.12%, p=0.309), right ventricular (RV) GLS (61mg: −14.18% vs. −13.72%, p=0.377; 20mg: −14.53% vs. −13.99%, p=0.452) and left atrial (LA) reservoir strain (LASr; 61mg: 8.80% vs. 9.42%, p=0.283; 20mg: 8.23% vs. 8.67%, p=0.589), whereas treatment-naïve ATTR-CM patients (n=54) had clear signs of disease progression at the end of the observation period [10.5 months; LV-GLS: −11.71% vs. −10.59%, p=0.001; RV-GLS: −14.36% vs. −12.99%, p=0.038; LASr: 10.67% vs. 8.41%, p=0.005]. Between-group comparison at follow-up revealed beneficial effects of Tafamidis 61mg on LASr (p=0.003), LV-GLS (p=0.030) and interventricular septum (IVS) thickness (p=0.006), resulting in clinical benefits (six-minute walk distance (6-MWD): p=0.006, NT-proBNP: p≤0.001), while patients treated with Tafamidis 20mg QD showed positive effects on LASr (p=0.039) but no differences in LV-GLS (p=0.274), IVS thickness (p=0.068) and clinical status (6-MWD: p=0.124, NT-proBNP: p=0.053) compared to the natural course.

Conclusion

Treatment with Tafamidis 61mg in ATTR-CM patients delays the increase in IVS thickness and the deterioration of LA and LV longitudinal function, resulting in significant clinical benefits compared with natural history. Serial TTE with 2D speckle-tracking imaging may be appropriate for disease-specific therapy monitoring.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): This work was supported by the Pfizer Inc. However, Pfizer Inc. did not have influence on study design, data processing, or statistical analysis.

Quantification of myocardial amyloid deposition in tafamidis-treated patients with transthyretin amyloid cardiomyopathy

Background

Tafamidis kinetically stabilizes the tetrameric form of transthyretin (TTR) and thus may halt disease progression in patients with transthyretin amyloid cardiomyopathy (ATTR-CM). However, the effect of tafamidis treatment on the progression of myocardial amyloid deposition is still unclear.

Methods

In our explorative analysis, we aimed to investigate the treatment effect of tafamidis on myocardial amyloid deposition measured by myocardial standardized uptake value (SUV) peak and SUV retention index using quantitative single-photon emission computed tomography/computed tomography (SPECT/CT) of the thorax and to observe its association with clinical parameters. Therefore, clinical, laboratory, imaging, and SPECT/CT examinations were performed in twenty consecutive ATTR-CM patients who started treatment with tafamidis 61mg, both at baseline and at a median of 6 months.

Results

Main results are summarized in Table 1. In brief, we observed a significant reduction of mean myocardial SUV peak (baseline: 15.50 vs. follow-up: 11.61, p&lt;0.001) and mean SUV retention index (5.64 vs. 3.58, p=0.001) after treatment with tafamidis (Figure 1A). Interestingly, a higher percentage decrease in the SUV retention index is more likely to be associated with clinical benefit, with a threshold of −30% distinguishing between patients who respond clinically (n=12) and those who do not (n=8, Figure 1B). Clinical response is demonstrated by improvement in exertional dyspnea (NYHA class III: 83.3% vs. 41.7%, p=0.047) and mean functional capacity as measured by 6-minute walk distance (349.5m vs. 356.7m, p=0.736). Cardiac biomarkers analysis showed a clear reduction in median NT-proBNP levels in the responder cohort (2765.0 pg/mL vs. 1904.0 pg/mL, p=0.041) compared to an increase in the non-responder cohort (1825.0 pg/mL vs. 1944.0 pg/mL, p=0.208; cohort comparison: p=0.026, Figure 1C). Echocardiographic findings revealed improvement in mean left ventricular (LV) strain (−12.0% vs. −13.5%, p=0.049) and mean LV ejection fraction (LVEF, 48.5% vs. 52.7%, p=0.287) in the responder cohort, while significant deterioration in mean LV function (LV strain: −13.9 vs. −10.5, p=0.035; LVEF: 53.2% vs. 46.5%, p=0.012) was observed in the non-responder cohort, with an additional substantial deterioration in right ventricular (RV) function as measured by tricuspid annular plane systolic excursion (TAPSE, mean, BL: 19.2mm vs. FU: 12.6mm, p=0.037) in those patients. These results are consistent with changes in the LV and RV function in cardiac magnetic resonance imaging parameters in each of the two cohorts.

Conclusion

Treatment with tafamidis in patients with ATTR-CM results in a significant reduction in myocardial amyloid deposition as measured by the SUV retention index, with a threshold of −30% distinguishing patients who respond clinically from those who do not. However, a larger patient sample is needed to verify these results.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): Pfizer Inc.

A machine learning-derived electrocardiographic algorithm for the detection of cardiac amyloidosis

Background

The diagnosis of cardiac amyloidosis (CA) requires advanced imaging techniques. Typical surface ECG patterns have been described, but their diagnostic value is limited.

Purpose

The aim of this study was to perform a comprehensive electrophysiological characterization in CA patients and to develop a robust, easy-to-use diagnostic tool.

Methods

First, we applied electrocardiographic imaging (ECGI) to generate detailed electroanatomical maps in CA patients and controls. Then, a machine learning approach was used to generate a surface ECG-based diagnostic algorithm from the complex dataset.

Results

Areas of low voltage were localized in the basal inferior regions of both ventricles and the remaining right ventricular segments in CA. The earliest epicardial breakthrough of myocardial activation was visualized in the right ventricle. Potential maps showed an accelerated and diffuse propagation pattern. We correlated the results from ECGI with 12-lead ECG recordings. Ventricular activation correlated best with R-peak timing in leads V1 to V3. Epicardial voltage showed a strong positive correlation with R-peak amplitude in inferior leads II, III, aVF. Ten blinded cardiologists were then asked to identify CA patients by analyzing 12-lead ECGs before and after training for the defined ECG patterns. Training resulted in significant improvements in the detection rate of CA with an AUC of 0.69 before and 0.97 after training (Figure).

Conclusion

Using a machine learning approach, a robust ECG-based tool was developed to detect CA from detailed electroanatomical mapping of CA patients. The developed tool proved to be a simple and reliable diagnostic tool to suspect CA without the aid of advanced imaging modalities.

Funding Acknowledgement

Type of funding sources: None.

Cardiac imaging in tafamidis-treatment patients with transthyretin amyloid cardiomyopathy

Background

Tafamidis kinetically stabilizes the tetrameric form of transthyretin (TTR) and thus may halt disease progression in patients with transthyretin amyloid cardiomyopathy (ATTR-CM). In our explorative analysis, we aimed to investigate the treatment effect on functional capacity and cardiac biomarkers as well as cardiac function and structure using echocardiography and cardiac magnetic resonance imaging (CMR), and to compare patients treated with tafamidis with an untreated control cohort.

Methods

Consecutive ATTR-CM patients either received tafamidis 61mg (n=64) or tafamidis 20mg (n=23) or were assigned to an untreated control cohort (n=54) reflecting the natural history of the disease. Subsequently, we performed clinical, laboratory, echocardiography and CMR follow-up at a median of 9 to 12.5 months.

Results

Main results are summarized in Table 1. In brief, we observed evidence of improvement in functional capacity as measured by the 6-minute walk distance (6MWD) in tafamidis 61mg treated patients (baseline: 377.1m vs. follow-up: 383.2m, p=0.678) compared to a significant decline in mean 6MWD in untreated patients (388.1m vs. 336.4m, p=0.002; cohort comparison: p=0.005). Analysis of cardiac biomarkers revealed evidence of therapeutic response by a decrease in median NT-proBNP levels in patients treated with tafamidis 61mg (2633.0pg/mL vs. 2244.0pg/mL, p=0.366), whereas a significant increase was observed in untreated patients (2798.0pg/mL vs. 3422.0pg/mL, p&lt;0.001; cohort comparison: p&lt;0.001). Echocardiographic findings revealed evidence of approximate stabilization in mean left ventricular (LV) strain (−11.75% vs. −11.58%, p=0.534) and mean right ventricular (RV) strain (−14.18% vs. −13.72, p=0.377) in the tafamidis 61mg treatment cohort compared to significant deterioration of mean LV longitudinal function (−11.71% vs. −10.59%, p=0.001) and mean RV longitudinal function (−14.36% vs. −12.99%, p=0.038) in the untreated cohort (cohort comparison: p=0.030 and p=0.269). Furthermore, cardiac structural assessment by CMR showed a significant increase in mean LV mass (199.1g vs. 214.3g, p=0.040) and mean extracellular volume (50.52% vs. 55.96%, p=0.026) in untreated patients, suggesting increased progression of myocardial amyloid deposition.

Conclusion

Treatment with tafamidis in patients with ATTR-CM results in significant improvements in functional capacity and cardiac biomarkers, and shows marked benefits in functional as well as structural imaging parameters compared with an untreated control cohort.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): Pfizer Inc.

Recurrent heart failure hospitalizations in patients with preserved ejection fraction: predictors and outcome

Background

Heart failure with preserved ejection fraction (HFpEF) is the most common form of HF and its prevalence is approaching epidemic proportions. Current treatment strategies aim to improve clinical status and reduce mortality rates. Episodes of acute HF are one of the main reasons for hospitalization in people over 65 years; however, they have not been well studied in HFpEF patients yet.

Objectives

The aim of this study was to investigate the impact of recurrent HF hospitalizations on long-term outcomes and to find predictors for subsequent events.

Methods

Between December 2010 and December 2019, 422 patients with confirmed HFpEF were enrolled in this study and prospectively followed.

Results

During follow-up, 190 HFpEF patients (45%) experienced HF hospitalizations with a median frequency of 2 (IQR: 1–4). Those presenting with acute HF had higher body mass indices (p=0.018), worse performance in 6-minute walking tests (p&lt;0.001), higher levels of N-terminal prohormone of brain natriuretic peptide (NT-proBNP, p&lt;0.001) and, compared to stable patients, a larger proportion suffered from New York Heart Association functional class ≥ III (p&lt;0.001). Furthermore, baseline left ventricular diastolic dysfunction (early mitral inflow velocity/early diastolic mitral annular velocity; p=0.002) as well as right ventricular afterload (pulmonary artery wedge pressure; p&lt;0.001) were more pronounced in patients with acute deteriorations. Over the observation period 107 patients (25%) died. Kaplan-Meier curves revealed increasingly worse survival in patients with recurrent HF events (5-years survival: 1 HF event 66% vs ≥3 HF events 53%, p&lt;0.001; Figure). Time to last hospitalization was a strong predictor of survival with an adjusted HR of 2.5 (95% CI 1.63–3.98; p&lt;0.001) in multivariate Cox regression analysis. Predictors of recurrent HF hospitalization were 6-minute walking distance (OR: 0.07, CI 0.02–0.22; p=0.001), systolic pulmonary artery pressure (OR: 1.05, CI 1.03–1.07, p=0.001) and NT-pro BNP (OR: 4.92, CI: 2.68–9.04, p=0.001)

Conclusions

HFpEF patients experiencing recurrent HF hospitalizations have worse long-term outcome. Intensive efforts should be made to maintain HFpEF patients compensated over time.

Funding Acknowledgement

Type of funding sources: Private company. Main funding source(s): Novartis

Cardiac remodeling in ambitious endurance-trained amateur athletes older than 50 years

Background

Data on cardiac remodeling in athletes are conflicting and rarely available in veteran athletes.

Purpose

Therefore, in this study we aimed to define echocardiographic features of healthy endurance athletes older than 50 years.

Methods

We included probands aged &gt;50 years, who have performed endurance sports at 70% of their maximum heart rate for at least 1 hour 3 times per week over the previous 5 years.

Results

Between November 2018 and May 2019, 69 probands were recruited. Median age was 57 years (IQR 52–64) and 26% were female. Groups were formed according to the median training time of 6 hours per week: (A) 45 probands with ≥6 hours (IQR 6–10) and (B) 24 probands with &lt;6 hours (IQR 3.5–5). Age, sex and body mass index were similar. Group A demonstrated slightly smaller right atrial and left atrial (LA) diameters when compared to probands in B (53mm (IQR 49–55) vs. 53mm (IQR 51–58), p=0.045, and 52mm (IQR 49–55) vs. 53mm (IQR 52–58), p=0.039, respectively) and showed preserved diastolic function (p=0.026) with lower E/E' ratio (7 (IQR 6–9) vs. 9 (IQR 7–10), p=0.039). Interventricular septal thickness and relative wall thickness ratio were similar.

Global right and left ventricular (LV) strain were similar, but LA reservoir strain was higher in group A than in B (27% (IQR 22–34) vs. 20% (IQR 15–29), p=0.016).

Conclusion

Endurance training in healthy athletes older than 50 years does not result in relevant chamber dilatation or LV hypertrophy and a weekly training duration of ≥6 hours seems beneficial to preserve diastolic function associated with an increased LA reservoir function.

Funding Acknowledgement

Type of funding source: None

A novel SGLT-2 score to identify HFpEF patients who may benefit from SGLT-2 inhibitors

Background

Established heart failure (HF) treatments have shown no effects in HF and preserved ejection fraction (HFpEF). Subgroup analyses of the HFpEF populations suggest that certain patients benefit from HF treatments. This underlines the importance of individualized therapy regimens in HFpEF. Sodium-glucose transporter 2 (SGLT-2) inhibitors are emerging as a promising treatment of HF. The mechanisms leading to improved outcomes include 1) treatment of diabetes, 2) osmodiuresis preventing volume overload, 3) enhancement of the cardio protective Angiotensin (Ang) 1–7 pathway, instead of Ang II. We aimed to characterize patients by factors which are modified by SGLT-2 inhibitors to identify individuals who may benefit from these drugs.

Methods

HFpEF patients were included in a single center registry. Baseline evaluation included assessment of HbA1c, fluid status measured by body composition monitor and plasma angiotensin concentration. A “SGLT-2 score” with a maximum of 3 points was calculated using the following parameters: 1) HbA1c &gt;6.5%, 2) overhydration, defined as a fluid overload of &gt;1,5L and 3) plasma renin activity (PRA) levels above the median as a parameter of over-all RAS activity. Primary outcome was defined as all-cause death or HF hospitalization. All parameters used in the “SGLT-2 score” were independently predictive for the chosen endpoint. Kaplan Meier analyses was used to show the association between the score and outcomes.

Results

90 patients were included in this registry. Median HbA1c was 6.0%, median fluid status was 1.2L and the median Ang II levels in the “high PRA-group” were 5.35.1 pmol/L. After a mean follow up time of 44.0±38.7 months, 60 patients (66.6%) reached the endpoint. Kaplan Meier analysis showed an association between SGLT-2 score and outcome (p=0.003).

Conclusion

Patients with HbA1c &gt;6.5%, overhydration and high RAS activity have poor outcomes. We propose the future use of this score to identify a subgroup of HFpEF patients who may benefit from SGLT-2 inhibitors.

Kaplan Meier analysis

Funding Acknowledgement

Type of funding source: None

P4572Lipoprotein(a) is not associated with survival after acute coronary syndromes

Aims

Lipoprotein(a) [Lp(a)] is associated with coronary artery disease in population studies, however studies on its predictive value in patients with established cardiovascular disease, in particular after acute coronary syndromes (ACS), are conflicting. The aim of this study was to investigate whether Lp(a) is associated with survival after ACS.

Methods and results

We analyzed 4469 consecutive patients that underwent coronary angiography for ACS. Lp(a) measurement at time of ACS was available in 1245 patients and median follow-up for cardiovascular and total mortality was 5.0 (IQR 3.2–8.0) years. 655 (52.6%) presented with ST-segment elevation myocardial infarction (STEMI), 424 (34.1%) with Non-ST-segment elevation myocardial infarction (NSTEMI) and 166 (13.3%) underwent coronary angiography for unstable angina. Cardiovascular mortality was 9.1% and total mortality was 15.7%. Patients were stratified into four groups to their Lp(a) levels. (≤15 mg/dL, >15–30 mg/dL, >30–60 mg/dL, and >60 mg/dL). Multivessel disease was significantly more common in patients with Lp(a) >60 mg/dL (p<0.05). Increased levels of Lp(a) were not associated with cardiovascular mortality (HR compared with Lp(a) ≤15 mg/dL were 1.2, 1.2, and 1.0, respectively; p=0.69) and not with total mortality (HR compared with Lp(a) ≤15 mg/dL were 1.2, 1.2, and 1.2, respectively; p=0.46).

Central Figure

Conclusion

Lp(a) levels at time of ACS were neither associated with cardiovascular nor with total mortality. Although Lp(a) has been shown to be associated with incidence of coronary artery disease, this study does not support any role of Lp(a) as a risk factor after ACS. This should be taken into account for development of outcome studies for agents targeting Lp(a) plasma levels.

P2726Extremely boosted prediction of cardiac amyloidosis by routine laboratory paramaters

Background/Introduction

Cardiac amyloidosis (CA) is a rare and complex condition with poor prognosis. Novel therapies have been shown to improve outcome, however, most of the affected individuals remain undiagnosed, mainly due to a lack in awareness among clinicians. One approach to overcome this issue is to use automated diagnostic algorithms that act based on routinely available laboratory results.

Purpose

We tested the performance of flexible machine learning and traditional statistical prediction models for non-invasive CA diagnosis based on routinely collected laboratory parameters. Since laboratory routines vary between hospitals or other health care providers, special attention has been taken to adaptive and dynamic parameter selection, and to dealing with the frequent occurrence of missing values.

Methods

Our cohort consisted of 376 clinically accepted patients with various types of heart failure. Of these, 69 were diagnosed with CA via endomyocardial biopsy (positives), and 307 had unrelated cardiac disorders (negatives). A total of 63 routine laboratory parameters were collected from these patients, with a high incidence of missing values (on average 60% of patients for each parameter). We tested the performance of two prediction models: logistic regression, and extreme gradient boosting with regression trees. To deal with missing values we adopted two strategies: a) finding an optimal overlap of parameters and deleting all patients with missing values (reduction of parameters and samples), and b) retaining all features and imputing missing values with parameter-wise means. To fairly assess the performance of prediction models we employed a 10-fold cross validation (stratified to preserve sample class ratio). Finally, area under curve for receiver-operator characteristic (ROC AUC) was used as our final performance measure.

Results

A complex machine learning model based on forests of regression trees proved to be the most performant (ROC AUC 0.94±4%) and robust to missing values. The best regression model was obtained with the 25 most frequent variables and patient deletion in case of missing values (ROC AUC 0.82±0.8%). While progressive inclusion of predictor variables worsened the performance of the logistic regression, it increased that of the machine learning approach.

Conclusions

Extreme gradient boosting of regression trees by routine laboratory parameters achieved staggering accuracy results for the automated diagnosis of CA. Our data suggest that implementations of such algorithms as independent interpreters of routine laboratory results may help to establish or suggest the diagnosis of CA in patients with heart failure symptoms, even in the absence of specialized experts.

P5627Automated detection of calcified plaques in coronary optical coherence tomography images using image segmentation based on machine learning

Background and aim

Automated image recognition based on machine learning methods was proven to be feasible in several medical imaging applications recently. Beside image classification methods to categorize input images for example into healthy or suspicious, image segmentation allows accurate localization of pathologies and thereby facilitates a wide area of applications. Because of the unique composition of every machine learning problem the applicability of image segmentation methods for detecting coronary pathologies in optical coherence tomography images remains unclear. Furthermore, the prediction accuracy of deep learning methods usually depends on vast amounts of training data which are often not available for particular medical questions. Therefore special strategies need to be applied to achieve satisfying results with smaller training datasets. We aimed to investigate the applicability of machine learning methods for plaque detection in coronary OCT images, especially considering the challenge of a small training dataset.

Methods

Originating from a dataset of 104 OCT frames containing calcified plaques, we performed image preprocessing using a custom build OCT image processing software to crop the luminal part as well as the areas outside the circular OCT signal to reduce entropy. Furthermore, plaques were identified and marked by an experienced OCT analyst, drawing plaque-enclosing polygonal masks using the same software. We also performed common image augmentation strategies, primarily applying rotation and zoom operations. Subsequently, we split the samples randomly into training, validation and test datasets (80:10:10%). To train the segmentation model, we fed the training and validation samples into an U-Net Convolutional Neuronal Network implementation with domain-specific adaptions using the RMSprop optimizer based on the publicly available PyTorch library.

Results

After 50 training epochs, we could achieve a prediction accuracy of 74.4% with the current configuration measured by the Sørensen–Dice coefficient comparing the similarity of the predicted plaque masks with the ground truth samples (figure 1 illustrates an exemplary comparison between predicted and ground truth plaque masks).

Exemplary projection of a predicted mask

Conclusion

We were able to show that image segmentation based on machine learning strategies is a feasible way for automated plaque detection in coronary OCT imaging even based on small training datasets. Larger training datasets are necessary to raise prediction accuracy.