Analysis of Varicose Veins of Lower Extremities Based on Vascular Endothelial Cell Inflammation Images and Multi-Scale Deep Learning

Ruizong Zhu,Ningning Yin,Tianjiao Wu,Huiping Niu,Yapei Zhao

doi:10.1109/access.2019.2954708

Abstract

The doctor determines whether there are lesions in the human body through the diagnosis of medical images, and classifies and identifies the lesions. Therefore, the automatic classification and recognition of medical images has received extensive attention. Since the inflammatory phenomenon of vascular endothelial cells is closely related to the varicose veins of the lower extremities, in order to realize the automatic classification and recognition of varicose veins of the lower extremities, this paper proposes a varicose vein recognition algorithm based on vascular endothelial cell inflammation images and multi-scale deep learning, called MSDCNN. First, we obtained images of vascular endothelial cells in patients with varicose veins of the lower extremities and normal subjects. Second, multiple convolutional layers extract multi-scale features of vascular endothelial cell images. Then, the MFM activation function is used instead of the ReLU activation function to introduce a competitive mechanism that extracts more features that are compact and reduces network layer parameters. Finally, the network uses a 3 × 3 convolution kernel to improve the network feature extraction capability and use the 1 × 1 convolution kernel for dimensionality reduction to further streamline network parameters. The experimental results tell us that the network has the advantages of high recognition accuracy, fast running speed, few network parameters, and is suitable for small-embedded devices.

Highlights

With the continuous development of science and technology, digital medicine and digital images are increasingly developing into society
Al-Antari et al [6] used a structured support vector machine combined with deep learning to segment breast lumps, indicating that the model improves the accuracy and effectiveness of breast image segmentation
DATA ACQUISITION DEVICE The Internet of Things refers to the connection of any object to the network through the information-sensing device according to the agreed protocol

Summary

INTRODUCTION

With the continuous development of science and technology, digital medicine and digital images are increasingly developing into society. This paper uses the multiscale deep learning algorithm to extract the characteristics of vascular endothelial cell inflammation images, and classifies and recognizes varicose veins of lower extremities. Multiple convolutional layers extracted multi-scale features of vascular endothelial cell images. The technical contributions of our paper can be concluded as follows: In this paper, vascular endothelial cells were used as research objects to classify and identify varicose veins of lower extremities by extracting the inflammatory features of endothelial cells. Similar studies have found increased baseline levels of intercellular adhesion molecules icam-1, vcam-1, and elam-1 and von will brand factor in plasma of patients with varicose veins of the lower extremities, as well as those that reflect vascular endothelial cell injury. This paper takes vascular endothelial cells as the research object, extracts the characteristics of cell images, and classifies and recognizes varicose veins of lower extremities. The filter is band-passable, so it can better reflect the visual characteristics of the image, and it is widely used in edge detection

DEEP CONVOLUTIONAL NEURAL NETWORK

INCEPTION MODEL

ACTIVATION FUNCTION

DEPTH SEPARABLE CONVOLUTION

MULTI-SCALE DEEP LEARNING NETWORK STRUCTURE

CONCLUSION