Coronary artery decision algorithm trained by two-step machine learning algorithm

Non-invasive fractional flow reserve estimation using deep learning on intermediate left anterior descending coronary artery lesion angiography images

This study aimed to design an end-to-end deep learning model for estimating the value of fractional flow reserve (FFR) using angiography images to classify left anterior descending (LAD) branch angiography images with average stenosis between 50 and 70% into two categories: FFR > 80 and FFR ≤ 80. In this study 3625 images were extracted from 41 patients' angiography films. Nine pre-trained convolutional neural networks (CNN), including DenseNet121, InceptionResNetV2, VGG16, VGG19, ResNet50V2, Xception, MobileNetV3Large, DenseNet201, and DenseNet169, were used to extract the features of images. DenseNet169 indicated higher performance compared to other networks. AUC, Accuracy, Sensitivity, Specificity, Precision, and F1-score of the proposed DenseNet169 network were 0.81, 0.81, 0.86, 0.75, 0.82, and 0.84, respectively. The deep learning-based method proposed in this study can non-invasively and consistently estimate FFR from angiographic images, offering significant clinical potential for diagnosing and treating coronary artery disease by combining anatomical and physiological parameters.

Computational Fractional Flow Reserve From Coronary Computed Tomography Angiography—Optical Coherence Tomography Fusion Images in Assessing Functionally Significant Coronary Stenosis

Background

Coronary computed tomography angiography (CTA) and optical coherence tomography (OCT) provide additional functional information beyond the anatomy by applying computational fluid dynamics (CFD). This study sought to evaluate a novel approach for estimating computational fractional flow reserve (FFR) from coronary CTA-OCT fusion images.

Methods

Among patients who underwent coronary CTA, 148 patients who underwent both pressure wire-based FFR measurement and OCT during angiography to evaluate intermediate stenosis in the left anterior descending artery were included from the prospective registry. Coronary CTA-OCT fusion images were created, and CFD was applied to estimate computational FFR. Based on pressure wire-based FFR as a reference, the diagnostic performance of Fusion-FFR was compared with that of CT-FFR and OCT-FFR.

Results

Fusion-FFR was strongly correlated with FFR (r = 0.836, P < 0.001). Correlation between FFR and Fusion-FFR was stronger than that between FFR and CT-FFR (r = 0.682, P < 0.001; z statistic, 5.42, P < 0.001) and between FFR and OCT-FFR (r = 0.705, P < 0.001; z statistic, 4.38, P < 0.001). Area under the receiver operating characteristics curve to assess functionally significant stenosis was higher for Fusion-FFR than for CT-FFR (0.90 vs. 0.83, P = 0.024) and OCT-FFR (0.90 vs. 0.83, P = 0.043). Fusion-FFR exhibited 84.5% accuracy, 84.6% sensitivity, 84.3% specificity, 80.9% positive predictive value, and 87.5% negative predictive value. Especially accuracy, specificity, and positive predictive value were superior for Fusion-FFR than for CT-FFR (73.0%, P = 0.007; 61.4%, P < 0.001; 64.0%, P < 0.001) and OCT-FFR (75.7%, P = 0.021; 73.5%, P = 0.020; 69.9%, P = 0.012).

Conclusion

CFD-based computational FFR from coronary CTA-OCT fusion images provided more accurate functional information than coronary CTA or OCT alone.

Clinical Trial Registration

[www.ClinicalTrials.gov], identifier [NCT03298282].

Optimization of FFR prediction algorithm for gray zone by hemodynamic features with synthetic model and biometric data

BACKGROUND

Recent attempts on adopting artificial intelligence algorithm on coronary diagnosis had limitations on data quantity and quality. While most of previous studies only used vessel image as input data, flow features and biometric features should be also considered. Moreover, the accuracy should be optimized within gray zone as the purpose is to decide stent insertion with estimated fractional flow reserve.

OBJECTIVES

The main purpose of this study is to develop an artificial intelligence-based coronary vascular diagnosis system focused on performance in the gray zone, from CT image extraction to FFR estimation. Three main issues should be considered for an algorithm to be used for pre-screening: algorithm optimization in the gray zone, minimization of labor during image processing, and consideration of flow and biometric features. This paper introduces a full FFR pre-screening system from automatic image extraction to an algorithm for estimating the FFR value.

METHOD

The main techniques used in this study are an automatic image extraction algorithm, lattice Boltzmann method based computational fluid dynamics analysis of a synthetic model and patient data, and an AI algorithm optimization. For feature extraction, this study focused on an automatic process to reduce manual labor. The algorithm consisted of two steps: the first algorithm calculates flow features from geometrical features, and the second algorithm estimates the FFR value from flow features and patient biometric features. Algorithm selection, outlier elimination, and k-fold selection were included to optimize the algorithm.

CONCLUSION

Eight types of algorithms including two neural network models and six machine learning models were optimized and tested. The random forest model shows the highest performance before optimization, whereas the multilayer perceptron regressor shows the highest gray zone accuracy after optimization.

Diagnosis of obstructive sleep apnea with prediction of flow characteristics according to airway morphology automatically extracted from medical images: Computational fluid dynamics and artificial intelligence approach

BACKGROUND

Obstructive sleep apnea syndrome (OSAS) is being observed in an increasing number of cases. It can be diagnosed using several methods such as polysomnography.

OBJECTIVES

To overcome the challenges of time and cost faced by conventional diagnostic methods, this paper proposes computational fluid dynamics (CFD) and machine-learning approaches that are derived from the upper-airway morphology with automatic segmentation using deep learning.

METHOD

We adopted a 3D UNet deep-learning model to perform medical image segmentation. 3D UNet prevents the feature-extraction loss that may occur by concatenating layers and extracts the anteroposterior coordination and width of the airway morphology. To create flow characteristics of the upper airway training data, we analyzed the changes in flow characteristics according to the upper-airway morphology using CFD. A multivariate Gaussian process regression (MVGPR) model was used to train the flow characteristic values. The trained MVGPR enables the prompt prediction of the aerodynamic features of the upper airway without simulation. Unlike conventional regression methods, MVGPR can be trained by considering the correlation between the flow characteristics. As a diagnostic step, a support vector machine (SVM) with predicted aerodynamic and biometric features was used in this study to classify patients as healthy or suffering from moderate OSAS. SVM is beneficial as it is easy to learn even with a small dataset, and it can diagnose various flow characteristics as factors while enhancing the feature via the kernel function. As the patient dataset is small, the Monte Carlo cross-validation was used to validate the trained model. Furthermore, to overcome the imbalanced data problem, the oversampling method was applied.

RESULT

The segmented upper-airway results of the high-resolution and low-resolution models present overall average dice coefficients of 0.76±0.041 and 0.74±0.052, respectively. Furthermore, the classification accuracy, sensitivity, specificity, and F1-score of the diagnosis algorithm were 81.5%, 89.3%, 86.2%, and 87.6%, respectively.

CONCLUSION

The convenience and accuracy of sleep apnea diagnosis are improved using deep learning and machine learning. Further, the proposed method can aid clinicians in making appropriate decisions to evaluate the possible applications of OSAS.