Coronary Artery Disease Classification with Different Lesion Degree Ranges based on Deep Learning

TL;DR

This work tackles binary classification of invasive coronary angiography (ICA) patches into lesion vs non-lesion while systematically varying the lesion-degree thresholds. It uses an annotated CADICA dataset with seven severity levels from less than twenty percent to one hundred percent, converts frames to 32x32 patches, and evaluates five pretrained CNNs (DenseNet-201, MobileNet-V2, NasNet-Mobile, ResNet-18, ResNet-50) under four experimental strategies including data augmentation. Key findings show that high-severity lesions ($100\%$ and $99\%$) are detected with high F-measure up to about 0.93 and AUC up to about 0.98, whereas including milder degrees reduces performance by roughly 15–25% in F-measure and lowers AUC; DenseNet-201 and NasNet-Mobile perform best overall. These results inform CAD system design and highlight the need for balanced data across lesion severities to achieve robust clinical performance, with DenseNet-201 emerging as a strong generalist classifier.

Abstract

Invasive Coronary Angiography (ICA) images are considered the gold standard for assessing the state of the coronary arteries. Deep learning classification methods are widely used and well-developed in different areas where medical imaging evaluation has an essential impact due to the development of computer-aided diagnosis systems that can support physicians in their clinical procedures. In this paper, a new performance analysis of deep learning methods for binary ICA classification with different lesion degrees is reported. To reach this goal, an annotated dataset of ICA images that contains the ground truth, the location of lesions and seven possible severity degrees ranging between 0% and 100% was employed. The ICA images were divided into 'lesion' or 'non-lesion' patches. We aim to study how binary classification performance is affected by the different lesion degrees considered in the positive class. Therefore, five known convolutional neural network architectures were trained with different input images where different lesion degree ranges were gradually incorporated until considering the seven lesion degrees. Besides, four types of experiments with and without data augmentation were designed, whose F-measure and Area Under Curve (AUC) were computed. Reported results achieved an F-measure and AUC of 92.7% and 98.1%, respectively. However, lesion classification is highly affected by the degree of the lesion intended to classify, with 15% less accuracy when <99% lesion patches are present.

Figures (4)

• Figure 1: Samples of the eight lesion ranges in which patches are categorized.
• Figure 2: Examples of the modifications applied to the training sets to augment data.
• Figure 3: Highest F-measure and AUC obtained in the test set with 5-fold cross-validation and the number of patches used as the train set.
• Figure 4: Ranking of methods considering F-measure obtained for each positive class established.