Coronary angiography image segmentation based on PSPNet

https://doi.org/10.1016/j.cmpb.2020.105897Get rights and content

Highlights

  • l

    The segmentation of coronary angiography images based on PSPNet network was compared with FCN, and analyzed and discussed the experimental results using three evaluation indicators of precision, recall and Fl-score.

  • l

    the segmentation accuracy based on the deep neural network is much higher than that of the traditional algorithm

  • l

    The PSPNet network reduces manual interaction in diagnosis, reduces dependence on medical personnel, improves the efficiency of disease diagnosis, and provides auxiliary strategies for subsequent medical diagnosis systems based on cardiac coronary angiography.

Abstract

Purpose: Coronary artery disease (CAD) is known to have high prevalence, high disability and mortality. The incidence and mortality of cardiovascular disease are also gradually increasing worldwide. Therefore, our paper proposes to use a more efficient image processing method to extract accurate vascular structures from vascular images by combining computer vision and deep learning.

Method: Our proposed segmentation of coronary angiography images based on PSPNet network was compared with FCN, and analyzed and discussed the experimental results using three evaluation indicators of precision, recall and Fl-score. Aiming at the complex and changeable structure of coronary angiography images and over-fitting or parameter structure destruction, we implemented the parallel multi-scale convolutional neural network model using PSPNet, using small sample transfer learning that limits parameter learning method.

Results: The accuracy of our technique proposed in this paper is 0.957. The accuracy of PSPNet is 26.75% higher than the traditional algorithm and 4.59% higher than U-Net. The average segmentation accuracy of the PSPNet model using transfer learning on the test set increased from 0.926 to 0.936, the sensitivity increased from 0.846 to 0.865, and the specificity increased from 0.921 to 0.949. The segmentation effect in this paper is closest to the segmentation result of the human expert, and is smoother than that of U-Net segmentation.

Conclusion: The PSPNet network reduces manual interaction in diagnosis, reduces dependence on medical personnel, improves the efficiency of disease diagnosis, and provides auxiliary strategies for subsequent medical diagnosis systems based on cardiac coronary angiography.

Introduction

Coronary artery disease (CAD) has a high incidence and high fatality rate, and is one of the most important causes of death worldwide [1], [2], [3]. The pathophysiological basis of CAD is atherosclerotic lesions in coronary arteries, which will cause vascular stenosis or blockage. The primary way of early diagnosis of coronary heart disease is the detection and quantification of coronary artery stenosis.

With the constant understanding of diseases and advancement of imaging technology, these have led to the continuous emergence of new diagnostic methods. Coronary angiography (CAG) is regarded as the gold standard in clinical treatment [4]. This is a non-invasive method of examining coronary arteries. It does not require an arterial catheter, but allows the contrast fluid to flow through the heart by intravenous injection, and complete the heart scans [5]. X-ray angiography has a powerful fluoroscopic ability to examine the structure of coronary arteries, and is the most commonly used technology from in the clinical diagnosis of CAD. This diagnosis and treatment method has opened another door for people to understand CAD. It can clearly and intuitively display the detailed images of the coronary arteries, allowing doctors and patients to observe that the coronary arteries have vascular wall calcification and stenosis problems, which make the diagnosis of coronary heart disease more intuitive and scientific.

CAG image blood vessel segmentation is of great significance in the assessing of the degree of vascular disease, assisting doctors in diagnosis and treatment, and reconstruction of the three-dimensional structure of blood vessels [6], [7], [8]. A good segmentation result can save the doctor's manual segmentation time and help the doctor correctly diagnose the condition. The segmented blood vessels can be used for further research, such as blood flow velocity, degree of vascular stenosis, etc., to lay the foundation for subsequent vascular dynamics research and three-dimensional reconstruction.

Convolutional neural network (CNN), relying on weight sharing, automatic feature extraction and computer performance improvement, has achieved remarkable results [9]. The medical community has also noticed the great success of deep learning methods and hopes to apply these technologies to different tasks in medical image processing.

The use of deep learning technology to segment coronary vessels has three advantages. The first point is that it can automatically extract features [10]. In the past, meaningful or task-related parts were mainly identified by human experts. Based on their understanding of these images, these features can be carried out by non-experts in the field, using machine learning techniques. Now, deep learning has incorporated feature engineering steps into the scope of learning steps. In other words, there is no need to manually extract features. If necessary, deep learning only needs a small amount of pre-processed data, and then new features can be discovered in a self-learning manner.

The second point is that deep learning technology has good versatility [11]. Because of the difference in imaging environment and the level of radiologists, even the same patient will have different CAG images [12]. The traditional method uses the features designed by experts, who may have a good effect on a certain kind of image, but it is not universal. Deep learning technology uses a large number of high-quality labeled samples for training. This method can be applied to various types of contrast images.

The third point is that deep learning can improve efficiency and detection accuracy [13]. Coronary vascular images will inevitably introduce various noises during the imaging process, and the contrast agent will be unevenly distributed in the blood vessels due to various reasons. The vascular bifurcation will cause the traditional methods to be slow in the segmentation process. The accuracy is low and a lot of labor time is wasted. Deep learning technology relies on powerful feature extraction and classification. This enables deep learning technology to be more efficient and detection accuracy than traditional methods.

Section snippets

Target segmentation based on traditional algorithms

Currently, traditional image segmentation comprises threshold-based segmentation, region-based segmentation, edge-based segmentation, and specific theories-based segmentation. With the progress of research, segmentation algorithms are divided into pixel-based classification, threshold segmentation, edge detection, color image segmentation, depth image segmentation, and fuzzy set-based methods.

Data source

We will perform segmentation tasks and experiments on coronary angiography images. At present, the available coronary angiography image data is collected. This study established a database derived from real patient cases. We collected CAG images of 109 patients from the Fuwai Central China Cardiovascular Hospital.

Data preprocessing

During subsequent training, the network can only accept grayscale images as input, so we convert the blood vessel RGB value to a grayscale value from 0 to N-1 according to the

Discussion

At present, deep learning has achieved good diagnostic accuracy comparable to clinical experts and medical doctors in the field of medical imaging. However, in general, there are still many challenges in applying deep learning to the field of medical image processing.

First, there is still a gap between the segmentation accuracy and the current best method. Although deep learning has proven its great success in other fields, it has just entered the field of coronary artery segmentation, and many

Conclusion

Based on the coronary vascular imaging, the artery is a thin tubular structure, which has relatively low contrast and artifacts, and is difficult to accurately segment and effectively annotate the scarcity of samples. A PSPNet-based multi-scale CNN model is proposed. This algorithm first down-sample the pre-processed image to obtain images of multiple scales, and sends them to the CNN. The features of different scales are selected and fused in the fully connected layer, and finally the features

Declaration of Competing Interest

No.

References (31)

  • M Li et al.

    Retinal Blood Vessel Segmentation Based on Multi-Scale Deep Learning [C]

  • R L Mcmillan

    Coronary atherosclerotic heart disease [J]

    North Carolina medical journal

    (1958)
  • N Sigfusson et al.

    Declining coronary heart disease mortality in Iceland: contribution by incidence, recurrence and case fatality rate [J]

    Scandinavian Journal of Thoracic & Cardiovascular Surgery

    (2002)
  • W Yong et al.

    Clinical analysis of chronic cor pulmonale in elderly merges coronary atherosclerotic heart disease [J]

    Journal of Clinical Pulmonary Medicine

    (2010)
  • C Jadrian et al.

    Automatic multiscale vascular image segmentation algorithm for coronary angiography [J]

    Biomedical Signal Processing & Control

    (2018)
  • M T Dehkordi et al.

    Retraction: A Review of Coronary Vessel Segmentation Algorithms [J]

    Journal of Medical Signals and Sensors

    (2019)
  • S Jiangping et al.

    Assessment of coronary artery stenosis by coronary angiography: a head-to-head comparison with pathological coronary artery anatomy [J].Circulation: Cardiovascular Interventions

    Journal of the American Heart Association

    (2013)
  • A G Blaiech et al.

    Impact of Enhancement for Coronary Artery Segmentation Based on Deep Learning Neural Network[M]

    Pattern Recognition and Image Analysis

    (2019)
  • W Huang et al.

    Coronary Artery Segmentation by Deep Learning Neural Networks on Computed Tomographic Coronary Angiographic Images[C]

  • Huang Weimin et al.

    Coronary Artery Segmentation by Deep Learning Neural Networks on Computed Tomographic Coronary Angiographic Images [C]

  • K Jo et al.

    Segmentation of the Main Vessel of the Left Anterior Descending Artery Using Selective Feature Mapping in Coronary Angiography [J]

    Transactions of Atmospheric Sciences

    (2019)
  • F Jiang et al.

    Medical image semantic segmentation based on deep learning [J]

    Neural computing & applications

    (2018)
  • S Pan et al.

    Diagnostic Model of Coronary Microvascular Disease Combined With Full Convolution Deep Network With Balanced Cross-Entropy Cost Function [J]

    IEEE Access

    (2019)
  • P Moeskops et al.

    Deep Learning for Multi-task Medical Image Segmentation in Multiple Modalities[C]

  • M. Jia-Qing

    Study of image segmentation based on morphology and watershed algorithms for coronary angiography [J]

    Modern Electronics Technique

    (2015)
  • Cited by (0)

    View full text