RESUMO
Heart auscultation is a simple and inexpensive first-line diagnostic test for the early screening of heart abnormalities. A phonocardiogram (PCG) is a digital recording of an analog heart sound acquired using an electronic stethoscope. A computerized algorithm for PCG analysis can aid in detecting abnormal signal patterns and support the clinical use of auscultation. It is important to detect fundamental components, such as the first and second heart sounds (S1 and S2), to accurately diagnose heart abnormalities. In this study, we developed a fully convolutional hybrid fusion network to identify S1 and S2 locations in PCG. It enables timewise, high-level feature fusion from dimensionally heterogeneous features: 1D envelope and 2D spectral features. For the fusion of heterogeneous features, we proposed a novel convolutional multimodal factorized bilinear pooling approach that enables high-level fusion without temporal distortion. We experimentally demonstrated the benefits of the comprehensive interpretation of heterogeneous features, with the proposed method outperforming other state-of-the-art PCG segmentation methods. To the best of our knowledge, this is the first study to interpret heterogeneous features through a high level of feature fusion in PCG analysis.
RESUMO
BACKGROUND: Radiomics is expected to identify imaging features beyond the human eye. We investigated whether radiomics can identify coronary segments that will develop new atherosclerotic plaques on coronary computed tomography angiography (CCTA). METHODS: From a prospective multinational registry of patients with serial CCTA studies at ≥ 2-year intervals, segments without identifiable coronary plaque at baseline were selected and radiomic features were extracted. Cox models using clinical risk factors (Model 1), radiomic features (Model 2) and both clinical risk factors and radiomic features (Model 3) were constructed to predict the development of a coronary plaque, defined as total PV â≥ â1 âmm3, at follow-up CCTA in each segment. RESULTS: In total, 9583 normal coronary segments were identified from 1162 patients (60.3 â± â9.2 years, 55.7% male) and divided 8:2 into training and test sets. At follow-up CCTA, 9.8% of the segments developed new coronary plaque. The predictive power of Models 1 and 2 was not different in both the training and test sets (C-index [95% confidence interval (CI)] of Model 1 vs. Model 2: 0.701 [0.690-0.712] vs. 0.699 [0.0.688-0.710] and 0.696 [0.671-0.725] vs. 0.0.691 [0.667-0.715], respectively, all p â> â0.05). The addition of radiomic features to clinical risk factors improved the predictive power of the Cox model in both the training and test sets (C-index [95% CI] of Model 3: 0.772 [0.762-0.781] and 0.767 [0.751-0.787], respectively, all p â< â00.0001 compared to Models 1 and 2). CONCLUSION: Radiomic features can improve the identification of segments that would develop new coronary atherosclerotic plaque. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov NCT0280341.