Semi-automated echocardiographic quantification of right ventricular size and function

Measuring the accuracy of cardiac output using POCUS: the introduction of artificial intelligence into routine care

Background

Shock management requires quick and reliable means to monitor the hemodynamic effects of fluid resuscitation. Point-of-care ultrasound (POCUS) is a relatively quick and non-invasive imaging technique capable of capturing cardiac output (CO) variations in acute settings. However, POCUS is plagued by variable operator skill and interpretation. Artificial intelligence may assist healthcare professionals obtain more objective and precise measurements during ultrasound imaging, thus increasing usability among users with varying experience. In this feasibility study, we compared the performance of novice POCUS users in measuring CO with manual techniques to a novel automation-assisted technique that provides real-time feedback to correct image acquisition for optimal aortic outflow velocity measurement.

Methods

28 junior critical care trainees with limited experience in POCUS performed manual and automation-assisted CO measurements on a single healthy volunteer. CO measurements were obtained using left ventricular outflow tract (LVOT) velocity time integral (VTI) and LVOT diameter. Measurements obtained by study subjects were compared to those taken by board-certified echocardiographers. Comparative analyses were performed using Spearman’s rank correlation and Bland–Altman matched-pairs analysis.

Results

Adequate image acquisition was 100% feasible. The correlation between manual and automated VTI values was not significant (p = 0.11) and means from both groups underestimated the mean values obtained by board-certified echocardiographers. Automated measurements of VTI in the trainee cohort were found to have more reproducibility, narrower measurement range (6.2 vs. 10.3 cm), and reduced standard deviation (1.98 vs. 2.33 cm) compared to manual measurements. The coefficient of variation across raters was 11.5%, 13.6% and 15.4% for board-certified echocardiographers, automated, and manual VTI tracing, respectively.

Conclusions

Our study demonstrates that novel automation-assisted VTI is feasible and can decrease variability while increasing precision in CO measurement. These results support the use of artificial intelligence-augmented image acquisition in routine critical care ultrasound and may have a role for evaluating the response of CO to hemodynamic interventions. Further investigations into artificial intelligence-assisted ultrasound systems in clinical settings are warranted.

Automatic cardiac evaluations using a deep video object segmentation network

BACKGROUND

Accurate cardiac volume and function assessment have valuable and significant diagnostic implications for patients suffering from ventricular dysfunction and cardiovascular disease. This study has focused on finding a reliable assistant to help physicians have more reliable and accurate cardiac measurements using a deep neural network. EchoRCNN is a semi-automated neural network for echocardiography sequence segmentation using a combination of mask region-based convolutional neural network image segmentation structure with reference-guided mask propagation video object segmentation network.

RESULTS

The proposed method accurately segments the left and right ventricle regions in four-chamber view echocardiography series with a dice similarity coefficient of 94.03% and 94.97%, respectively. Further post-processing procedures on the segmented left and right ventricle regions resulted in a mean absolute error of 3.13% and 2.03% for ejection fraction and fractional area change parameters, respectively.

CONCLUSION

This study has achieved excellent performance on the left and right ventricle segmentation, leading to more accurate estimations of vital cardiac parameters such as ejection fraction and fractional area change parameters in the left and right ventricle functionalities, respectively. The results represent that our method can predict an assured, accurate, and reliable cardiac function diagnosis in clinical screenings.

Comparison of semi-automated versus manual quantitative right ventricular assessment in tetralogy of Fallot

BACKGROUND

Tetralogy of Fallot is a congenital heart defect diagnosed in infancy. Assessment of right ventricular size and function is important for evaluation of patients with tetralogy of Fallot, but these quantitative measures are challenging by echocardiography. This study evaluates a semi-automated software (EchoInsight®, Epsilon Imaging) by comparing its measures to manual measures in children with tetralogy of Fallot.

METHODS

Echocardiographic measurements were performed using manual techniques and semi-automated software. Right ventricular measurements included end-diastolic and end-systolic area, fractional area change, chamber dimensions, and tricuspid annular plane systolic excursion. Reliability, correlation, and agreement between manual and semi-automated measures were assessed.

RESULTS

Echocardiograms for 46 patients were analysed. Intra- and inter-observer reliabilities for semi-automated measures were good with intraclass correlation coefficients all over 0.95 and 0.85, respectively. There was high correlation between manual and semi-automated methods for areas and dimensions (r = 0.91-0.98). Tricuspid annular plane systolic excursion measures and fractional area change also correlated, albeit less strongly. The semi-automated measurements of end-systolic and end-diastolic area were a 20 and 47% higher than manual methods, respectively.The semi-automated method yielded a relative 52% lower fractional area change compared to the manual method.

CONCLUSIONS

The semi-automated software generates quantitative right ventricular measures in children with tetralogy of Fallot with good reliability and good correlation with manual methods for all measures, but with significant difference between manual and semi-automated techniques for area and functional measures. The specific right ventricular geometry in tetralogy of Fallot children may be why, compared to normal anatomy, greater differences were observed between the two techniques.

The Year in Perioperative Echocardiography: Selected Highlights from 2020

This article is the fifth of an annual series reviewing the research highlights of the year pertaining to the subspecialty of perioperative echocardiography for the Journal of Cardiothoracic and Vascular Anesthesia. The authors thank Editor-in-Chief Dr. Kaplan and the editorial board for the opportunity to continue this series. In most cases, these will be research articles that are targeted at the perioperative echocardiography diagnosis and treatment of patients after cardiothoracic surgery; but in some cases, these articles will target the use of perioperative echocardiography in general.

Artificial Intelligence in Echocardiography for Anesthesiologists

Echocardiography is a unique diagnostic tool for intraoperative monitoring and assessment of patients with cardiovascular diseases. However, there are high levels of interoperator variations in echocardiography interpretations that could lead to inaccurate diagnosis and incorrect treatment. Furthermore, anesthesiologists are faced with the additional challenge to interpret echocardiography and make decisions in a limited timeframe from these complex data. The need for an automated, less operator-dependent process that enhances speed and accuracy of echocardiography analysis is crucial for anesthesiologists. Artificial intelligence is playing an increasingly important role in the medical field and could help anesthesiologists analyze complex echocardiographic data while adding increased accuracy and consistency to interpretation. This review aims to summarize practical use of artificial intelligence in echocardiography and discusses potential limitations and challenges in the future for anesthesiologists.

Comparison of Semi-Automated Versus Manual Quantitative Right Ventricular Assessment in Hypoplastic Left Heart Syndrome

Quantitative echocardiographic evaluation is important for systemic right ventricles, but its asymmetric shape makes this challenging and time consuming when performed manually. Semi-automated software could make these quantitative measures easier to accomplish in the clinical setting. We hypothesized that semi-automated software would approximate manual measures of right ventricular size and function. Children with hypoplastic left heart who had echocardiograms were prospectively identified. These measurements were performed using manual and semi-automated techniques: end-diastolic and end-systolic area, fractional area change (FAC), dimensions (longitudinal, basal and mid-cavitary diameters), and tricuspid annular plane systolic excursion (TAPSE). Agreement between measures was evaluated. Sixty-three echocardiograms were analyzed. Intra- and inter-observer reliability was acceptable and similar between methods except that inter-observer reliability for the manual method was superior for TAPSE. Correlation between methods was high (r > 0.9, p < 0.001) for most of the measures. Correlation for FAC was r = 0.79, and for TAPSE the correlation was r = 0.61 (both p < 0.001). The percent relative difference between manual and semi-automated methods was less than 6% for most measures. End-systolic area and FAC had a relative difference of 10% and 11% respectively. The only measure with substantial bias between the manual and semi-automated methods was TAPSE which had a relative difference of 52%. EchoInsight® semi-automated software provides similar measures of right ventricular dimensions and FAC in patients with hypoplastic left heart compared to manual measures. Measures of TAPSE do not correlate well between manual and semi-automated methods. Further research is warranted on the use of semi-automated analyses in this patient population.

Artificial intelligence in cardiovascular imaging: state of the art and implications for the imaging cardiologist

Healthcare, conceivably more than any other area of human endeavour, has the greatest potential to be affected by artificial intelligence (AI). This potential has been shown by several reports that demonstrate equal or superhuman performance in medical tasks that aim to improve efficiency, diagnosis and prognosis. This review focuses on the state of the art of AI applications in cardiovascular imaging. It provides an overview of the current applications and studies performed, including the potential value, implications, limitations and future directions of AI in cardiovascular imaging.It is envisioned that AI will dramatically change the way doctors practise medicine. In the short term, it will assist physicians with easy tasks, such as automating measurements, making predictions based on big data, and putting clinical findings into an evidence-based context. In the long term, AI will not only assist doctors, it has the potential to significantly improve access to health and well-being data for patients and their caretakers. This empowers patients. From a physician's perspective, reliable AI assistance will be available to support clinical decision-making. Although cardiovascular studies implementing AI are increasing in number, the applications have only just started to penetrate contemporary clinical care.

Reduced Right Ventricular Fractional Area Change, Strain, and Strain Rate before Bidirectional Cavopulmonary Anastomosis is Associated with Medium-Term Mortality for Children with Hypoplastic Left Heart Syndrome

BACKGROUND

Ventricular dysfunction is associated with increased morbidity and mortality in children with hypoplastic left heart syndrome. The aim of this study was to assess the diagnostic performance of conventional and speckle-tracking echocardiographic measures of right ventricular (RV) function before bidirectional cavopulmonary anastomosis palliation in predicting death or need for heart transplantation (HTx).

METHODS

RV fractional area change (RVFAC) and longitudinal and circumferential strain and strain rate (SR) were measured in 64 prospectively recruited patients with hypoplastic left heart syndrome from echocardiograms obtained before bidirectional cavopulmonary anastomosis surgery. The composite end point of death or HTx was examined. Receiver operating characteristic analysis was performed, and cutoff values optimizing sensitivity and specificity were derived.

RESULTS

At a median follow-up of 5.0 years (interquartile range, 2.8-6.4 years), 13 patients meeting the composite end point had lower longitudinal strain and SR, circumferential SR, and RVFAC compared with survivors (n = 51). The conventional cutoff of RVFAC < 35% was specific for death or HTx (86%) but had poor sensitivity (46%), with an area under the curve of 0.73. Speckle-tracking echocardiographic variables showed similar areas under the curve (range, 0.69-0.79), with negative predictive values >90%. Addition of speckle-tracking echocardiographic variables to RVFAC < 35% showed no added benefit. However, in a subpopulation of patients with RVFAC ≥ 35% (n = 44), those meeting the composite end point (n = 7) had lower longitudinal SR (median, -1.0 1/sec [interquartile range, -0.8 to -1.1 1/sec] vs -1.21/sec [interquartile range, -1.0 to -1.3 1/sec], P = .03). Interobserver reproducibility was superior for longitudinal strain and SR (intraclass correlation coefficient > 0.92) compared with RVFAC (intraclass correlation coefficient = 0.75).

CONCLUSIONS

Children with hypoplastic left heart syndrome with normal RVFAC and ventricular deformation before bidirectional cavopulmonary anastomosis have a low likelihood of death or HTx in the medium term. In the presence of reduced RVFAC, speckle-tracking echocardiography does not provide additional prognostic value. However, in patients with "normal" RVFAC, it may have a role in improving outcome prediction and warrants further investigation.