Classification of Echocardiogram View using A Convolutional Neural Network

The standard views in echocardiography capture distinct slices of the heart which can be used to assess cardiac function. Determining the view of a given echocardiogram is the first step for analysis. To automate this step, a deep network of the ResNet-18 architecture was used to classify between six standard views. The network parameters were pre-trained with the ImageNet database and prediction quality was assessed with a visualization tool known as gradient-weighted class activation mapping (Grad-CAM). The network was able to distinguish between three parasternal short axis views and three apical views to ~99\% accuracy. 10-fold cross validation showed a 97\%-98\% accuracy for the apical view subcategories (which included apical two-, three-, and four- chamber views). Grad-CAM images of these views highlighted features that were similar to those used by experts in manual classification. Parasternal short axis subcategories (which included apex level, mitral valve level, and papillary muscle level) had accuracies of 54\%-73\%. Grad-CAM images illustrate that the network classifies most parasternal short axis views as belonging to the papillary muscle level. Likely more images and incorporating time-dependent features would increase the parasternal short axis view accuracy. Overall, a convolutional neural network can be used to reliably classify echocardiogram views.

Transthoracic echocardiogram by protocol apical-subcostal in prone position during invasive mechanical ventilation

Funding Acknowledgements Type of funding sources: None. Introduction During COVID-19 pandemic reports increased of transthoracic echocardiography (TTE) in prone and invasive mechanical ventilation (IMV), mostly describing four and five chamber apical views, and inferior vena cava (IVC) by lateral IVC window (from right side of the patient). Purpose Evaluate quality of images obtained with protocol of apical-subcostal TTE in patients with IMV and prone position. Methods Prospective study, between August and December 2020, in adults who required prone position during IMV. After placing the patient in the prone position, left arm was extended overhead, and a pillow was placed only under the left hemithorax to elevate and facilitate the apical window and space below the patient for subcostal window. Operator stands on the left side of the patient and takes images with the transducer in your right hand, starting with apical window and subsequently subcostal window. Apical views were apical four, two and three-chambers, to obtain function parameter of right and left ventricle (LV), evaluation of aortic valve, mitral and tricuspid valve. Subcostal views were four cardiac chambers and IVC, to obtain qualitative ventricle function, presence of pericardial effusion, and volume status and estimation of pulmonary pressure (together to peak systolic tricuspid pressure gradient from apical views). The images were acquired by cardiologist, then were saved, and finally evaluated by two echocardiography cardiologist experts. Results 16 ETTs were performed. Male gender and obesity predominate. Positive end-expiratory pressure average was 10.8 cm of water. One patient cannot be assessed by absence of acoustic window and the rest (n = 15) were analyzed with experts. It was achieved a four-chamber apical view in 100% (n = 15), a two-chambers apical view in 60% (n = 9) and a three-chambers apical view in 100% (n = 15). It was possible to assess global function of the LV in 100% (n = 15), LV segmental function in 53% (n = 8), LV outflow tract velocity time integral in 100% (n = 15) and tricuspid annular plane systolic excursion in 100% (n = 15). Pulsed wave doppler of mitral valve in 100% (n = 15) and tissue doppler of lateral mitral valve annulus 100% (n = 15). Continuous wave doppler of aortic valve in 100% (n = 15) and tricuspid valve in 93% (n = 14). Subcostal four-chamber 80% (n = 12), presence pericardial effusion 100% (n = 15) and IVC 93% (n = 14). Non-complications associated with obtaining the position. Conclusions ETT in the prone position during IMV was possible and interpretable images were achieved. The position described allows assessment by apical and subcostal views at the same time and position of the operator and the patient. In addition, it was a safe technique, and the position was easy to be incorporated by the health team. Limitations were obtaining the two-chamber apical view and evaluation of segmental alterations of the LV. Better validation requires a larger sample.

Regional Strain Pattern Index—A Novel Technique to Predict CRT Response

Background: Cardiac resynchronization therapy (CRT) improves outcome in patients with heart failure (HF) however approximately 30% of patients still remain non-responsive. We propose a novel index—Regional Strain Pattern Index (RSPI)—to prospectively evaluate response to CRT. Methods: Echocardiography was performed in 49 patients with HF (66.5 ± 10 years, LVEF 24.9 ± 6.4%, QRS width 173.1 ± 19.1 ms) two times: before CRT implantation and 15 ± 7 months after. At baseline, dyssynchrony was assessed including RSPI and strain pattern. RSPI was calculated from all three apical views across 12 segments as the sum of dyssynchronous components. From every apical view, presence of four components were assessed: (1) contraction of the early-activated wall; (2) prestretching of the late activated wall; (3) contraction of the early-activated wall in the first 70% of the systolic ejection phase; (4) peak contraction of the late-activated wall after aortic valve closure. Each component scored 1 point, thus the maximum was 12 points. Results: Responders reached higher mean RSPI values than non-responders (5.86 ± 2.9 vs. 4.08 ± 2.4; p = 0.044). In logistic regression analysis value of RSPI ≥ 7 points was a predictor of favorable CRT effect (OR: 12; 95% CI = 1.33–108.17; p = 0.004). Conclusions: RSPI could be a valuable predictor of positive outcome in HF patients treated with CRT.

Does automated left ventricular ejection fraction assessment available on handheld ultrasound devices represent improvement in the diagnostic accuracy?

Funding Acknowledgements Type of funding sources: None. Purpose To assess the accuracy of an algorithm for automated measurement of the left ventricular ejection fraction (LVEF) available on handheld ultrasound device (HUD). Methods 112 patients admitted to the cardiology department, who were referred for the conventional echocardiographic examination, underwent additional assessment performed with HUD (Vscan Extend, GE Vingmed Ultrasound, Horten, Norway). In each case 4 – chamber apical view was obtained and LVEF was calculated by means of the LVivo software. Imaging quality was assessed in a 4-grade scale. Subsequently, during the examination performed with the use of the stationary echocardiograph the three-dimensional (3D) measurement of LVEF was recorded. Results Ultimately 96 (53 men, mean age 63 ± 11) patients were enrolled into the study group In the remaining 16 cases (14%) 3D image quality was not sufficient to allow the calculation of the LVEF. LVivo software was unsuccessful in calculating LVEF in all these 16 patients and in additional 20 patients, who remained in the study group due to satisfactory 3D image quality. The quality of images acquired with the use of HUD was assessed as optimal in 25 (26%) patients, good in 37 (39%), acceptable in 24 (25%), poor in 10 (10%). The average LVEF value was 46%±14 with the 3D LVQ measurements and 48%±14 using the LVivo software. The correlation coefficient between the LVEF values obtained with the two methods was r = 0,92; (P < 0,0001). Using paired samples t-test we found that the difference between these two techniques was not significant (mean difference 4,5± 3,4%; P = 0,35). LVivo software EF assessment is based on a single apical view and for this reason we have assumed that the differences in EF can be larger in patients with regional wall motion abnormalities, in whom LVEF values derived from different apical views can significantly vary. For this reason the group of patients with history of myocardial infarction (40pts, 42%) was analysed separately and we found that the difference between LVivo and 3D LVEF was also not statistically significant (mean difference 6,1± 3,3%; P= 0,14). The correlation coefficient equalled r = 0,78; (P < 0,0001). Conclusion The LVivo software despite its limitations is capable of the accurate LVEF measurement when the acquired views are of at least good imaging quality. Such expanded capabilities of HUDs can potentially lead to the overall improvements of the diagnostic accuracy of the ultrasonographic examinations, particularly when in hands of the non-expert echocardiographers.

The Apical View Sign for Easier Identification of Left Juxtaposition of the Right Atrial Appendage

Left juxtaposition of the right atrial appendage (RAA) is a rare congenital cardiac malformation, where both atrial appendages are located leftward of the great arteries. Despite well-described echocardiographic markers, this diagnosis remains challenging to establish, especially for the novice imager. We describe the apical view sign in three patients with juxtaposed RAA. We noted that the normal left atrial appendage was visualized from the apical four-chamber or equivalent coronal view at the level of the atrioventricular valves.

Kudoa ajurutellus n. sp. (Multivalvulida: Kudoidae), a parasite of the skeletal musculature of the Bressou sea catfish, Aspistor quadriscutis, in northeastern of the State of Pará

Kudoa ajurutellus n. sp. (Multivalvulida: Kudoidae) is described as a parasite of the Bressou catfish, Aspistor quadriscutis (Siluriformes) collected at Ajuruteua beach, northeastern Pará, Brazil. The new species is described based on the morphology of the spores and the sequence of the 18S rDNA. This parasite was found only in the skeletal muscular tissue of fish hosts. In the apical view, the spores were pseudo-quadrangular in shape, with rounded borders, and four symmetrical capsules. The spores are 6.63±0.53 μm in length and 8.16±0.75 μm in width. The polar capsules are piriform, 3.45±0.30 μm in length and 1.87±0.15 μm in width. The phylogenetic analysis based on the sequence of the 18S rDNA gene indicated that the new species is clearly distinct from all other Kudoa species, and that the degree of differentiation is constant with the existence of a new species of the genus Kudoa. This is the first marine Kudoa species recorded in northeastern of the State of Pará.

P990 Multi-plane echocardiographic analysis of right ventricular function in patients with systemic and subpulmonic physiology

Background Right ventricular (RV) function is recognized as a prognostic factor in congenital heart disease (CHD). Accurate echocardiographic parameters to assess systolic function in systemic RV (sRV) lacking. We previously introduced a novel four-view approach with different RV walls visualized in their long axis from one apical view using 2D-multi-plane transthoracic echocardiographic (TTE) (iRotate). Aims To extensively evaluate RV systolic function using iRotate echocardiography in CHD patients with systemic RV compared with a whole spectrum of CHD patients with abnormally loaded subpulmonic RV. Methods and Results Thirty CHD patients with sRV and 112 age, gender and BSA matched patients with abnormally loaded subpulmonic RV were recruited from the outpatient clinic. All subjects underwent complete TTE with evaluation of TAPSE, TDI S’ and peak systolic global longitudinal RV strain (RV-GLS) from the RV walls using the four-view iRotate model. The feasibility of TAPSE and TDI S’ ranged between 94% and 100%. The feasibility of RV-GLS in CHD was 98%, 69%, 87% and 72% respectively in the lateral, anterior, inferior and inferior coronal view walls. All echocardiographic parameters were significantly lower in sRV compared to versus subpulmonic RV cohort (p < 0.001) (Table). Conclusion This study provides for the first time an extensive RV specific analysis of the systemic RV. The feasibility of all RV parameters in the four-view iRotate model is excellent in CHD and represents a reproducible, easily applicable and complete RV assessment in daily practice. Systolic function is significantly reduced in systemic RV compared to subpulmonic RV physiology. Abstract P990 Figure.

P1534 Clinical feasibility of 4D myocardial speckle tracking strain analysis in soccer referee: comparison to 2D strain

Funding Acknowledgements no OnBehalf no Background Quantitative analysis of deformation parameters can provide evidence about the presence of disease . Speckle tracking (ST)strain is recently largely used in sports medicine, as an usefull tool to add value to the traditional echo information .Especially 4D echo-strain is emerging in this context despite not yet completely explored. The study aim to verify the feasibility of the 4D strain analysis , compared to the 2D ST strain data, in a class of international trained soccer referee. Materials and methods : a group of 25 soccer referee ( aged 45,5 ± 5) underwent to an EMT, conducted at the 85% of their effort to obtain the eligibility. An 2D echocardiographic evaluation was also performed with the measurements of the systolic-diastolic parameters . A specific investigation for the 2D and 4D strain analysis by X-Strain-Esaote software, was possible from the 2C,3C;4C apical view chambers, acquired as cine loop. The study was completed by the study of the circumferential strain calculated from short axis view, from the images of the basal and medium LV segments. Results : all 2D echo parameters were normal (LVDD: 50,56 ± 2 mm,68; LVSD: 35,4 ± 2,3 mm; EF: 65,76 ±1,78%; E/A:1,73 ± 0,35 ; RV:23,71 ± 1 mm) The 2D GLS ( -22,29 ± 2,64); Circumferential strain( Basal -22,29 ± 5,3; Apical -25,66 ± 5,8 ) were in the normal range. No statistical differences ( P:NS) were observed comparing the mean GLS(21,60 ± 2,68) to the LV 4D strain( -21,24 ± 2,75) Conclusions : 4D strain completes and strengthens the role of the 2D strain in evaluating the myocardial function contributing to exclude eventual undiscovered disease. Especially in athletes, often checked in a short time the combination of the two methods should be widely proposed .More studies will be necessary in future to confirm this hypothesis. Abstract P1534 Figure. 4D strain

554 Atrial mechanical dispersion in patients with atrial fibrillation undergoing catheter ablation: a strain study

Background Atrial mechanical dispersion (MD) might increase the risk of atrial fibrillation (AF). However, the data on atrial mechanical dispersion in patients with AF undergoing catheter ablation (CA) are scarce. Purpose To describe effects of CA on inter- and left intra-atrial MD in patients with different sub-types of AF undergoing CA. Methods We prospectively enrolled 138 symptomatic patients (age: 63 ± 21 years, 32% females) with paroxysmal AF undergoing first (81%) or redo (19%) CA during sinus rhythm, and 20 individuals (age: 66 ± 23 years, 20% females) with long-standing persistent AF undergoing first CA during AF. All patients had normal (≥50%) left ventricular ejection fraction. Control group consisted of 23 healthy controls. The atrial strain and strain rate (SR) were assessed using the two-dimensional speckle tracking echocardiography as average of segmental values in all apical views for LA and in four chamber (4CH) apical view for RA. We quantified inter-atrial MD as the standard deviation of time from the onset of the P wave to peak negative strain curves (Figure 1), and left intra-atrial MD as the standard deviation of time from the onset of QRS wave to peak positive strain curves (Figure 2) of all LA segmental components in 4CH apical view after setting adjusting the reference frame to coincide with the onset of the QRS. Results Figure 3 shows a time course of reservoir LA strain and left intra-atrial MD in patients with paroxysmal AF undergoing the First-CA (3A) versus the Redo-CA (3B), and in patients with long-standing persistent AF (3C). At 1-day pre-ablation, patients with long-standing persistent AF showed significantly lower reservoir strain of both atria and higher left intra-atrial MD as compared with both paroxysmal AF groups and controls (all p < 0.01). The Redo-CA versus the First-CA group showed significantly lower LA reservoir strain (p < 0.01) while left intra-atrial MD was similar. RA reservoir strain and inter-atrial MD was similar between groups with paroxysmal AF. At 1-day post-ablation, we observed a significant deterioration of reservoir LA strain and left intra-atrial MD only in the First-CA group. After 3- month follow-up, left intra-atrial synchrony was significantly improved compared with pre-ablation in both groups of paroxysmal AF undergoing the first CA and long-standing persistent AF (p < 0.05). The reservoir LA strain showed partial recovery to pre-ablation values in paroxysmal AF undergoing first-CA while it showed a continuous improvement in long-standing persistent AF. In contrast, patients with paroxysmal AF undergoing redo-CA did not show significant improvement in any of these indices. Inter-atrial MD and RA strain did not show significant improvement in either group. Conclusion Atrial strain and MD shows distinct behavior in patients with different sub-types of AF post CA. Atrial MD may provide a complimentary information to strain when assessing LA function. Abstract 554 Figure.