Analyze COVID-19 CT images based on evolutionary algorithm with dynamic searching space

Yunhong Gong,Yanan Sun,Zhongtai Yan,Peng Chen,Ke Yang,Dezhong Peng

doi:10.1007/s40747-021-00513-8

Abstract

The COVID-19 pandemic has caused a global alarm. With the advances in artificial intelligence, the COVID-19 testing capabilities have been greatly expanded, and hospital resources are significantly alleviated. Over the past years, computer vision researches have focused on convolutional neural networks (CNNs), which can significantly improve image analysis ability. However, CNN architectures are usually manually designed with rich expertise that is scarce in practice. Evolutionary algorithms (EAs) can automatically search for the proper CNN architectures and voluntarily optimize the related hyperparameters. The networks searched by EAs can be used to effectively process COVID-19 computed tomography images without expert knowledge and manual setup. In this paper, we propose a novel EA-based algorithm with a dynamic searching space to design the optimal CNN architectures for diagnosing COVID-19 before the pathogenic test. The experiments are performed on the COVID-CT data set against a series of state-of-the-art CNN models. The experiments demonstrate that the architecture searched by the proposed EA-based algorithm achieves the best performance yet without any preprocessing operations. Furthermore, we found through experimentation that the intensive use of batch normalization may deteriorate the performance. This contrasts with the common sense approach of manually designing CNN architectures and will help the related experts in handcrafting CNN models to achieve the best performance without any preprocessing operations

Highlights

At the beginning of 2020, the outbreak of an atypical and person-to-person transmissible pneumonia caused by severe acute respiratory syndrome coronavirus 2 (SARS-COV-2, known as COVID-19) spread worldwide
Even though Evolutionary algorithms (EAs) play a significant role in the Neural architecture search (NAS) without additional expert knowledge, the fixed-length chromosome genetic encoding strategy may not be suitable for optimizing deep learning hyperparameters, because the performance of convolutional neural networks (CNNs) highly depends on the model depth [31]
The objective of the paper was to develop a new evolutionary approach to automatically evolve the CNN architecture for the COVID-19 classification task. This goal was successfully achieved by the proposed variable-length chromosome strategy with the growth operation and the shrink operation, which are suitable for classification tasks with a dynamic searching space

Summary

Introduction

At the beginning of 2020, the outbreak of an atypical and person-to-person transmissible pneumonia caused by severe acute respiratory syndrome coronavirus 2 (SARS-COV-2, known as COVID-19) spread worldwide. Even though EAs play a significant role in the NAS without additional expert knowledge, the fixed-length chromosome genetic encoding strategy may not be suitable for optimizing deep learning hyperparameters, because the performance of CNNs highly depends on the model depth [31]. – A novel EA with variable-length chromosomes is proposed to search hyperparameter settings of CNNs for COVID-19 diagnosis tasks. With the proposed balance operation, including the growth strategy and the shrink strategy, the length of chromosomes may be changed, and the hyperparameter searching space may be altered adaptively, which can significantly reduce dependence on data quality and neural network redundancy without careful pretreatment and expert knowledge. Operation, and excess BN operations may poorly affect image analysis performance

Background and related work

Findings

Conclusion and future work