Abstract 15899: A Hybrid Artificial Intelligence-Intrinsic Frequency Method for Instantaneous Determination of Myocardial Infarct Size
| Autor: | Rashid Alavi, Niema M. Pahlevan, Wangde Dai, Robert A. Kloner |
|---|---|
| Rok vydání: | 2020 |
| Předmět: | |
| Zdroj: | Circulation. 142 |
| ISSN: | 1524-4539 0009-7322 |
| Popis: | Introduction: Intrinsic Frequency (IF) method is a recently developed systems-based method that extracts dynamics information about left ventricle function (LV), arterial dynamics, and the LV-arterial coupling from arterial waveforms. We have recently shown (Alavi et al. Circulation, 140 (2019), A12573-A12573) that IF can detect occurrence of an acute myocardial infarction (MI) using a single carotid pressure waveform. Here, we propose that the myocardial infarct size (area of necrosis over total LV area) can be approximated using a hybrid IF-artificial neural network (ANN) method. Methods: The standard MI model was used in anesthetized Sprague Dawley rats (n=27). The proximal left coronary artery was occluded for 30 minutes to ensure necrosis followed by 3 hours of reperfusion. The left ventricle slices were incubated in triphenyl tetrazolium chloride (TTC) to distinguish the necrotic (white) and the non-necrotic (dark red) areas (Fig.1a), thereby obtaining the size of MI through histopathology. IF parameters were computed from random carotid pressure waveforms 2 hours after the reperfusion. A 3-layer ANN model (4 input, 5 hidden, and 1 output node) was applied on IFs from 22 rats to design the ANN (18 for training, 4 for validation). The model was then tested on 5 different rats with the same MI procedure described above. Results: The results showed a significant correlation (R=0.64, P Conclusions: Our results suggest that a hybrid IF-AI method can predict the anatomic infarct size from an arterial waveform without advanced imaging. This technique is clinically significant since infarct sizes are link to the survival and development of heart failure in MI patients, and IF parameters can be obtained noninvasively from carotid waveforms using arterial tonometry devices or an iPhone. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|