Article Preview
Top1. Introduction
As of 12 August 2022 at 6:36 pm CEST, 5859500085 confirmed cases of COVID-19 have been reported to WHO globally, including 6425422 deaths. The highly infectious and pathogenic nature of this virus poses a serious health risk. Patients infected with COVID-19 are likely to suffer respiratory failure to death within a short period. The general population is vulnerable to infection when in unprotected contact with people who have COVID-19. Early detection as early as possible can improve the chances of survival of patients with COVID-19. Early isolation of patients for treatment can reduce the possibility of spreading the virus.
The New Guinea pandemic was not only a health crisis but also had a huge economic impact, hitting several industries hard. In the catering industry [1, 2], all types of gatherings were drastically reduced, crowded places were shut down, large numbers of restaurants and hotels ceased operations, stocks of ready-made vegetables were sold off at low prices. In the tourism sector [3, 4], major tourist attractions were closed and major cultural events were canceled across the region. Transportation [5, 6] was restricted, with several cities closing roads, logistics halted, airports and high-speed trains running at significantly reduced frequencies, and few residents terminating their travel plans. These effects on the tertiary sector have also indirectly led to a reduction in the supply of raw materials for manufacturing [7], difficulties in transportation, higher costs, and lower output. The transport, tourism, accommodation and catering sectors in all countries have experienced varying degrees of decline.
Chest imaging technology has played an important role in the response to this global crisis. Manual diagnosis based on medical imaging relies heavily on the expertise of medical personnel. It can be time-consuming and difficult to detect hidden lesions due to individual theories or experience. Computer analysis is less likely to be missed and less time-consuming, helping doctors to save effort but get accurate information about a patient's condition quickly. The need to collate and analyze the large data associated with COVID-19 can also be met by computers.
As a result, the development or optimization of deep learning-based models [8] for medical images has become an important research direction. In this paper, we first briefly describe the characteristics of common medical images and common analysis tasks. Secondly, the common deep learning models are discussed; then, the current status of domestic and international research on deep learning for medical image classification, detection [9], segmentation, and other application areas are discussed; Finally, the challenges of deep learning methods for medical image analysis are discussed and summarised, together with the main strategies and open research directions.