CCSI: Continual Class-Specific Impression for data-free class incremental learning.

In real-world clinical settings, traditional deep learning-based classification methods struggle with diagnosing newly introduced disease types because they require samples from all disease classes for offline training. Class incremental learning offers a promising solution by adapting a deep network trained on specific disease classes to handle new diseases. However, catastrophic forgetting occurs, decreasing the performance of earlier classes when adapting the model to new data. Prior proposed methodologies to overcome this require perpetual storage of previous samples, posing potential practical concerns regarding privacy and storage regulations in healthcare. To this end, we propose a novel data-free class incremental learning framework that utilizes data synthesis on learned classes instead of data storage from previous classes. Our key contributions include acquiring synthetic data known as Continual Class-Specific Impression (CCSI) for previously inaccessible trained classes and presenting a methodology to effectively utilize this data for updating networks when introducing new classes. We obtain CCSI by employing data inversion over gradients of the trained classification model on previous classes starting from the mean image of each class inspired by common landmarks shared among medical images and utilizing continual normalization layers statistics as a regularizer in this pixel-wise optimization process. Subsequently, we update the network by combining the synthesized data with new class data and incorporate several losses, including an intra-domain contrastive loss to generalize the deep network trained on the synthesized data to real data, a margin loss to increase separation among previous classes and new ones, and a cosine-normalized cross-entropy loss to alleviate the adverse effects of imbalanced distributions in training data. Extensive experiments show that the proposed framework achieves state-of-the-art performance on four of the public MedMNIST datasets and in-house echocardiography cine series, with an improvement in classification accuracy of up to 51% compared to baseline data-free methods. Our code is available at https://github.com/ubc-tea/Continual-Impression-CCSI.

Validation of machine learning models for estimation of left ventricular ejection fraction on point-of-care ultrasound: insights on features that impact performance.

BACKGROUND: Machine learning (ML) algorithms can accurately estimate left ventricular ejection fraction (LVEF) from echocardiography, but their performance on cardiac point-of-care ultrasound (POCUS) is not well understood. OBJECTIVES: We evaluate the performance of an ML model for estimation of LVEF on cardiac POCUS compared with Level III echocardiographers' interpretation and formal echo reported LVEF. METHODS: Clinicians at a tertiary care heart failure clinic prospectively scanned 138 participants using hand-carried devices. Video data were analyzed offline by an ML model for LVEF. We compared the ML model's performance with Level III echocardiographers' interpretation and echo reported LVEF. RESULTS: There were 138 participants scanned, yielding 1257 videos. The ML model generated LVEF predictions on 341 videos. We observed a good intraclass correlation (ICC) between the ML model's predictions and the reference standards (ICC = 0.77-0.84). When comparing LVEF estimates for randomized single POCUS videos, the ICC between the ML model and Level III echocardiographers' estimates was 0.772, and it was 0.778 for videos where quantitative LVEF was feasible. When the Level III echocardiographer reviewed all POCUS videos for a participant, the ICC improved to 0.794 and 0.843 when only accounting for studies that could be segmented. The ML model's LVEF estimates also correlated well with LVEF derived from formal echocardiogram reports (ICC = 0.798). CONCLUSION: Our results suggest that clinician-driven cardiac POCUS produces ML model LVEF estimates that correlate well with expert interpretation and echo reported LVEF.

Automated Atrial Fibrillation Diagnosis by Echocardiography without ECG: Accuracy and Applications of a New Deep Learning Approach.

BACKGROUND: Automated rhythm detection on echocardiography through artificial intelligence (AI) has yet to be fully realized. We propose an AI model trained to identify atrial fibrillation (AF) using apical 4-chamber (AP4) cines without requiring electrocardiogram (ECG) data. METHODS: Transthoracic echocardiography studies of consecutive patients ≥ 18 years old at our tertiary care centre were retrospectively reviewed for AF and sinus rhythm. The study was first interpreted by level III-trained echocardiography cardiologists as the gold standard for rhythm diagnosis based on ECG rhythm strip and imaging assessment, which was also verified with a 12-lead ECG around the time of the study. AP4 cines with three cardiac cycles were then extracted from these studies with the rhythm strip and Doppler information removed and introduced to the deep learning model ResNet(2+1)D with an 80:10:10 training-validation-test split ratio. RESULTS: 634 patient studies (1205 cines) were included. After training, the AI model achieved high accuracy on validation for detection of both AF and sinus rhythm (mean F1-score = 0.92; AUROC = 0.95). Performance was consistent on the test dataset (mean F1-score = 0.94, AUROC = 0.98) when using the cardiologist's assessment of the ECG rhythm strip as the gold standard, who had access to the full study and external ECG data, while the AI model did not. CONCLUSIONS: AF detection by AI on echocardiography without ECG appears accurate when compared to an echocardiography cardiologist's assessment of the ECG rhythm strip as the gold standard. This has potential clinical implications in point-of-care ultrasound and stroke risk stratification.

Cardiac Phenotyping of SARS-CoV-2 in British Columbia: A Prospective Echo Study With Strain Imaging.

BACKGROUND: There is limited data on the residual echocardiographic findings including strain analysis among post-coronavirus disease (COVID) patients. The aim of our study is to prospectively phenotype post-COVID patients. METHODS: All patients discharged following acute COVID infection were systematically followed in the post-COVID-19 Recovery Clinic at Vancouver General Hospital and St. Paul's Hospital. At 4-18 weeks post diagnosis, patients underwent comprehensive echocardiographic assessment. Left ventricular ejection fraction (LVEF) was assessed by 3D, 2D Biplane Simpson's, or visual estimate. LV global longitudinal strain (GLS) was measured using a vendor-independent 2D speckle-tracking software (TomTec). RESULTS: A total of 127 patients (53% female, mean age 58 years) were included in our analyses. At baseline, cardiac conditions were present in 58% of the patients (15% coronary artery disease, 4% heart failure, 44% hypertension, 10% atrial fibrillation) while the remainder were free of cardiac conditions. COVID-19 serious complications were present in 79% of the patients (76% pneumonia, 37% intensive care unit admission, 21% intubation, 1% myocarditis). Normal LVEF was seen in 96% of the cohort and 97% had normal right ventricular systolic function. A high proportion (53%) had abnormal LV GLS defined as < 18%. Average LV GLS of septal and inferior segments were lower compared to that of other segments. Among patients without pre-existing cardiac conditions, LVEF was abnormal in only 1.9%, but LV GLS was abnormal in 46% of the patients. CONCLUSIONS: Most post-COVID patients had normal LVEF at 4-18 weeks post diagnosis, but over half had abnormal LV GLS.

Machine learning derived echocardiographic image quality in patients with left ventricular systolic dysfunction: insights on the echo views of greatest image quality.

We sought to determine the cardiac ultrasound view of greatest quality using a machine learning (ML) approach on a cohort of transthoracic echocardiograms (TTE) with abnormal left ventricular (LV) systolic function. We utilize an ML model to determine the TTE view of highest quality when scanned by sonographers. A random sample of TTEs with reported LV dysfunction from 09/25/2017-01/15/2019 were downloaded from the regional database. Component video files were analyzed using ML models that jointly classified view and image quality. The model consisted of convolutional layers for extracting spatial features and Long Short-term Memory units to temporally aggregate the frame-wise spatial embeddings. We report the view-specific quality scores for each TTE. Pair-wise comparisons amongst views were performed with Wilcoxon signed-rank test. Of 1,145 TTEs analyzed by the ML model, 74.5% were from males and mean LV ejection fraction was 43.1 ± 9.9%. Maximum quality score was best for the apical 4 chamber (AP4) view (70.6 ± 13.9%, p<0.001 compared to all other views) and worst for the apical 2 chamber (AP2) view (60.4 ± 15.4%, p<0.001 for all views except parasternal short-axis view at mitral/papillary muscle level, PSAX M/PM). In TTEs scanned by professional sonographers, the view with greatest ML-derived quality was the AP4 view.

Quantification of pleural effusions by two-dimensional transthoracic echocardiography.

PURPOSE: There is lack of validated methods for quantifying the size of pleural effusion from standard transthoracic (TTE) windows. The purpose of this study is to determine whether pleural effusion (Peff) measured from routine two-dimensional (2D) TTE views correlate with chest radiograph (CXR). MATERIALS AND METHODS: We retrospectively identified all inpatients who underwent a TTE and CXR within 2 days in a large tertiary care center. Peff was measured on TTE from parasternal long axis (PLAX), apical four-chamber (A4C), and subcostal views and on CXR. Logistic regression models were used determine optimal cut points to predict moderate or greater Peff. RESULTS: In 200 patients (mean age 69.3 ± 14.3 years, 49.5% female), we found statistically significant associations between Peff size assessed by all TTE views and CXR, with weak to moderate correlation (PLAX length: 0.21 (95% CI [0.05, 0.35]); PLAX depth: 0.21 (95% CI [0.05, 0.35]); A4C left: 0.31 (95% CI [0.13, 0.46]); A4C right: 0.39 (95% CI [0.17, 0.57]); subcostal: 0.38 (95% CI [0.07, 0.61]). The best TTE thresholds for predicting moderate or greater left-sided Peff on CXR was PLAX length left > = 8.6 cm (sensitivity 78%, specificity 54%, PPV 26%, and NPV 92%). The best TTE thresholds for predicting moderate or greater right-sided Peff on CXR was A4C right > = 2.6 cm (sensitivity 87%, specificity 60%, PPV 37%, and NPV 94%). CONCLUSIONS: We identified statistically significant associations with Peff size measured on TTE and CXR. The predictive ability of TTE to identify moderate or large pleural effusion is limited.

Temporal Changes of Cardiac Structure, Function, and Mechanics During Sub-acute Cervical and Thoracolumbar Spinal Cord Injury in Humans: A Case-Series.

Individuals with cervical spinal cord injury (SCI) experience deleterious changes in cardiac structure and function. However, knowledge on when cardiac alterations occur and whether this is dependent upon neurological level of injury remains to be determined. Transthoracic echocardiography was used to assess left ventricular structure, function, and mechanics in 10 male individuals (median age 34 years, lower and upper quartiles 32-50) with cervical (n = 5, c-SCI) or thoracolumbar (n = 5, tl-SCI) motor-complete SCI at 3- and 6-months post-injury. Compared to the 3-month assessment, individuals with c-SCI displayed structural, functional, and mechanical changes during the 6-month assessment, including significant reductions in end diastolic volume [121 mL (104-139) vs. 101 mL (99-133), P = 0.043], stroke volume [75 mL (61-85) vs. 60 mL (58-80), P = 0.042], myocardial contractile velocity (S') [0.11 m/s (0.10-0.13) vs. 0.09 m/s (0.08-0.10), P = 0.043], and peak diastolic longitudinal strain rate [1.29°/s (1.23-1.34) vs. 1.07°/s (0.95-1.15), P = 0.043], and increased early diastolic filling over early myocardial relaxation velocity (E/E') ratio [5.64 (4.71-7.72) vs. 7.48 (6.42-8.42), P = 0.043]. These indices did not significantly change in individuals with tl-SCI between time points. Ejection fraction was different between individuals with c-SCI and tl-SCI at 3 [61% (57-63) vs. 54% (52-55), P < 0.01] and 6 months [58% (57-62) vs. 55% (52-56), P < 0.01], though values were considered normal. These results demonstrate that individuals with c-SCI exhibit significant reductions in cardiac function from 3 to 6 months post-injury, whereas individuals with tl-SCI do not, suggesting the need for early rehabilitation to minimize cardiac consequences in this specific population.

A Novel Continuous Left Ventricular Diastolic Function Score Using Machine Learning.

BACKGROUND: Unlike left ventricular (LV) ejection fraction, which provides a precise, reliable, and prognostically valuable measure of systolic function, there is no single analogous measure of LV diastolic function. OBJECTIVES: We aimed to develop a continuous score to grade LV diastolic function using machine learning modeling of echocardiographic data. METHODS: Consecutive echo studies performed at a tertiary-care center between February 1, 2010, and March 31, 2016, were assessed, excluding studies containing features that would interfere with diastolic function assessment as well as studies in which 1 or more parameters within the contemporary diastolic function assessment algorithm were not reported. Diastolic function was graded based on 2016 American Society of Echocardiography (ASE)/European Association of Cardiovascular Imaging (EACVI) guidelines, excluding indeterminate studies. Machine learning models were trained (support vector machine [SVM], decision tree [DT], XGBoost [XGB], and dense neural network [DNN]) to classify studies within the training set by diastolic dysfunction severity, blinded to the ASE/EACVI classification. The DNN model was retrained to generate a regression model (R-DNN) to predict a continuous LV diastolic function score. RESULTS: A total of 28,986 studies were included; 23,188 studies were used to train the models, and 5,798 studies were used for validation. The models were able to reclassify studies with high agreement to the ASE/EACVI algorithm (SVM, 83%; DT, 100%; XGB, 100%; DNN, 98%). The continuous diastolic function score corresponded well with ASE/EACVI guidelines, with scores of 1.00 ± 0.01 for studies with normal function and 0.74 ± 0.05, 0.51 ± 0.06, and 0.27 ± 0.11 for mild, moderate, and severe diastolic dysfunction, respectively (mean ± 1 SD). A score of <0.91 predicted abnormal diastolic function (area under the receiver operator curve = 0.99), while a score of <0.65 predicted elevated filling pressure (area under the receiver operator curve = 0.99). CONCLUSIONS: Machine learning can assimilate echocardiographic data and generate an automated continuous diastolic function score that corresponds well with current diastolic function grading recommendations.

Incidentally identified left ventricular apical aneurysm in a patient with Fabry disease.

Fabry disease is a rare X-linked lysosomal storage disorder caused by a deficiency in the lysosomal enzyme, galactosidase A, that can result in a progressive increase in the left ventricle (LV) wall thickness from glycosphingolipid deposition leading to myocardial fibrosis, conduction abnormalities, arrhythmias, and heart failure. We present a case of a patient with advanced Fabry cardiomyopathy, in whom a small LV apical aneurysm was incidentally discovered on abdominal imaging, which could have easily evaded detection on standard transthoracic echocardiography. The LV apex should be thoroughly interrogated in patients with Fabry cardiomyopathy, as the finding of LV aneurysm could have important management implications with respect to the prevention of stroke and sudden cardiac death.

Spinal cord injury impairs cardiac function due to impaired bulbospinal sympathetic control.

Spinal cord injury chronically alters cardiac structure and function and is associated with increased odds for cardiovascular disease. Here, we investigate the cardiac consequences of spinal cord injury on the acute-to-chronic continuum, and the contribution of altered bulbospinal sympathetic control to the decline in cardiac function following spinal cord injury. By combining experimental rat models of spinal cord injury with prospective clinical studies, we demonstrate that spinal cord injury causes a rapid and sustained reduction in left ventricular contractile function that precedes structural changes. In rodents, we experimentally demonstrate that this decline in left ventricular contractile function following spinal cord injury is underpinned by interrupted bulbospinal sympathetic control. In humans, we find that activation of the sympathetic circuitry below the level of spinal cord injury causes an immediate increase in systolic function. Our findings highlight the importance for early interventions to mitigate the cardiac functional decline following spinal cord injury.

Handheld Echocardiography in a Clinical Practice Scenario: Concordances Compared to Standard Echocardiographic Reports.

BACKGROUND: The purpose of this study was to assess the utility of a handheld device (HH) used during common daily practice and its agreement with the results of a standard echocardiography study (STD) performed by experienced sonographers and echocardiographer. METHODS: A prospective follow-up was conducted in an adult outpatient echocardiography clinic. Experienced sonographers performed the STD and an experienced echocardiographer performed the HH. STD included 2-dimensional images, Doppler and hemodynamics analysis. Hemodynamic assessment was not performed with the HH device because the HH does not include such technology. The images were interpreted by blinded echocardiographers, and the agreement between the reports was analyzed. RESULTS: A total of 108 patients were included; and the concordance for left ventricle (LV) ejection fraction (EF), wall motion score index, LV and right ventricle (RV) function, RV size, and mitral and aortic stenosis was excellent with κ values greater than 0.80. Wall motion abnormalities had good concordance (κ value 0.78). The agreement for LV hypertrophy, mitral and aortic regurgitation was moderate, and tricuspid and pulmonary regurgitation agreements were low (κ values of 0.26 and 0.25, respectively). CONCLUSIONS: In a daily practice scenario with experienced hands, HH demonstrated good correlation for most echocardiography indications, such as ventricular size and function assessment and stenosis valve lesion analyses.

U-LanD: Uncertainty-Driven Video Landmark Detection.

This paper presents U-LanD, a framework for automatic detection of landmarks on key frames of the video by leveraging the uncertainty of landmark prediction. We tackle a specifically challenging problem, where training labels are noisy and highly sparse. U-LanD builds upon a pivotal observation: a deep Bayesian landmark detector solely trained on key video frames, has significantly lower predictive uncertainty on those frames vs. other frames in videos. We use this observation as an unsupervised signal to automatically recognize key frames on which we detect landmarks. As a test-bed for our framework, we use ultrasound imaging videos of the heart, where sparse and noisy clinical labels are only available for a single frame in each video. Using data from 4,493 patients, we demonstrate that U-LanD can exceedingly outperform the state-of-the-art non-Bayesian counterpart by a noticeable absolute margin of 42% in R2 score, with almost no overhead imposed on the model size.

Relationship between enlarged cardiac silhouette on chest X-ray and left ventricular size on transthoracic echocardiography.

The diagnostic accuracy of the cardiothoracic ratio on chest X-ray to detect left ventricular (LV) enlargement has not been well defined despite its traditional association with cardiomegaly. We aimed to determine whether the cardiothoracic ratio can accurately predict LV enlargement based on indexed linear measurements of the LV on transthoracic echocardiography (TTE). We included consecutive patients who had a TTE and a posteroanterior chest X-ray performed within 90 days of each other at a tertiary care center. LV size was determined by measuring the LV end-diastolic dimension (LVEDD) and LV end-diastolic dimension indexed (LVEDDI) to body surface area. The cardiothoracic ratio was calculated by dividing the maximum transverse diameter of the cardiac silhouette by the maximum transverse diameter of the right and left lung boundaries. 173 patients were included in the study (mean age 68 ± 15 years, 49.1% female). Mean cardiothoracic ratio was 0.56 ± 0.09, and the mean LVEDD and indexed LVEDDI were of 47 ± 8.6 mm and dimension of 27 ± 4.5 mm/m2 respectively. There was no significant correlation between the cardiothoracic ratio measured on chest X-ray and either the LVEDD or LVEDDI measured on TTE (r = 0.011, p = 0.879; r = 0.122, p = 0.111). The ability of the cardiothoracic ratio to predict LV enlargement (defined as LVEDDI > 30 mm/m2) was not statistically significant. The cardiothoracic ratio on chest X-ray is not a predictor of LV enlargement based on indexed linear measurements of the LV by TTE.

Prevalence of left ventricular systolic dysfunction by single echocardiographic view: towards an evidence-based point of care cardiac ultrasound scanning protocol.

Limited views are often obtained in the setting of cardiac ultrasound, however, the likelihood of missing left ventricular (LV) dysfunction based on a single view is not known. We sought to determine the echo views that were least likely to miss LV systolic dysfunction in consecutive transthoracic echocardiograms (TTEs). Structured data from TTEs performed at 2 hospitals from September 25, 2017, to January 15, 2019, were screened. Studies of interest were those with reported LV dysfunction. Views evaluated were the parasternal long-axis (PLAX), parasternal-short axis at mitral (PSAX M), papillary muscle (PSAX PM), and apical (PSAX A) levels, apical 2 (AP2), apical 3 (AP3), and apical 4 (AP4) chamber views. The probability that a view contained at least 1 abnormal segment was determined and analyzed with McNemar's test for 21 adjusted pair-wise comparisons. There were 4102 TTE studies included for analysis. TTEs on males comprised 72.7% of studies with a mean LV ejection fraction of 42.8 ± 9.7%. The echo view with the greatest likelihood of encompassing an abnormal segment was the AP2 view with a prevalence of 93.4% (p < 0.001, compared to all other views). The PLAX view performed the worst with a prevalence of 82.5% (p < 0.015, compared to all other views). The best parasternal view for the detection of abnormality was the PSAX PM view at 90.4%. In conclusions, a single echo view will contain abnormal segments > 82% of the time in the setting of LV systolic dysfunction, with a prevalence of up to 93.4% in the apical windows.

Fabry Cardiomyopathy: Current Practice and Future Directions.

Fabry disease (FD) is an X-linked lysosomal storage disorder caused by mutations in the galactosidase A (GLA) gene that result in deficient galactosidase A enzyme and subsequent accumulation of glycosphingolipids throughout the body. The result is a multi-system disorder characterized by cutaneous, corneal, cardiac, renal, and neurological manifestations. Increased left ventricular wall thickness represents the predominant cardiac manifestation of FD. As the disease progresses, patients may develop arrhythmias, advanced conduction abnormalities, and heart failure. Cardiac biomarkers, point-of-care dried blood spot testing, and advanced imaging modalities including echocardiography with strain imaging and magnetic resonance imaging (MRI) with T1 mapping now allow us to detect Fabry cardiomyopathy much more effectively than in the past. While enzyme replacement therapy (ERT) has been the mainstay of treatment, several promising therapies are now in development, making early diagnosis of FD even more crucial. Ongoing initiatives involving artificial intelligence (AI)-empowered interpretation of echocardiographic images, point-of-care dried blood spot testing in the echocardiography laboratory, and widespread dissemination of point-of-care ultrasound devices to community practices to promote screening may lead to more timely diagnosis of FD. Fabry disease should no longer be considered a rare, untreatable disease, but one that can be effectively identified and treated at an early stage before the development of irreversible end-organ damage.

The Need for Telemedicine Integration Into Adult Cardiology Training Curricula in Canada.

COVID-19 brought telemedicine to the forefront of clinical cardiology. We aimed to examine the extent of trainees' involvement in and comfort with telemedicine practices in Canada with the use of a web-based self-administered survey. Eighty-six trainees from 12 training programs completed the survey (65% response rate). Results showed that before COVID-19, 39 trainees (45%) had telemedicine exposure, compared with 67 (78%) after COVID-19 (P < 0.001). However, only 44 trainees (51%) reported being comfortable or very comfortable with the use of telemedicine. Of the 67 trainees who were involved in telemedicine, 4 (6%) had full supervision during virtual visits, 13 (19%) had partial supervision, and 50 (75%) had minimal or no supervision. Importantly, 67 trainees (78%) expressed the need for telemedicine-specific training and 64 (74%) were willing to have their virtual visits recorded for the purpose of evaluation and feedback. Furthermore, 47 (55%) felt strongly or very strongly positive about incorporating telemedicine into their future practice. The main perceived barriers to telemedicine use were concerns about patients' engagement, fear of weakening the patient-physician relationship, and unfamiliarity with telemedicine technology. These barriers, together with training in virtual physical examination skills and medicolegal aspects of telemedicine, are addressed in several established internal medicine telemedicine curricula that could be adapted by cardiology programs. In conclusion, while the degree of telemedicine involvement since COVID-19 was high, the trainees' comfort level with telemedicine practice remains suboptimal likely due to lack of training and inadequate staff supervision. Therefore, a cardiology telemedicine curriculum is needed to ensure that trainees are equipped to embrace telemedicine in cardiovascular clinical care.