**1. Introduction**

Cancer is a major public health problem with high morbidity and mortality worldwide [1]. Early detection and diagnosis are crucial for improving the 5-year survival rate [2]. Screening examination plays an essential role in the diagnosis of diseases [3], which produces a large number of images and requires physicians to interpret. However, human interpretation has many limitations, including shortterm memory, inaccuracy, distraction, and fatigue. To solve these limitations, computer-aided diagnosis (CAD) has been applied in the biomedical imaging field.

The application of CAD in medicine can be traced back to the 1950s. In 1959, Lusted first introduced the mathematical model of CAD and put forward the model in medicine [4]. They successfully applied this mathematical model to the diagnosis of lung cancer, which became the pioneer of CAD. In 1963, Lodwick et al. applied a computer to analyze digitized chest radiographs [5]. In 1964, Becker et al. developed an automatic method to measure the cardiothoracic ratio in chest radiographs [6]. In 1966, Ledley first proposed the theoretical model of CAD [7]. In 1973, Toriwaki et al. [8] and Roellinger et al. [9] applied the CAD system in chest radiographs for the diagnosis of abnormalities and heart abnormality, respectively. In 1976, Winsberg et al. applied the CAD in mammography for diagnosis of abnormalities [10].

The CAD was further developed in the early 1980s, and the expert system applied in the field of medicine was the most noticeable one. The CAD processing includes medical information collection, quantitative and statistical analysis of medical information, and diagnosis. Popular models included Bayes theorem, maximum likelihood model, and sequential model. In the middle 1980s, researchers focused on the development and evaluation of CAD systems. Artificial neural network (ANN) has developed rapidly since the 1990s. ANN is a mathematical processing method that imitates the working principle of human brain neurons. ANN can play an assistant role in diagnosis due to it has the ability of self-learning, memory, and forecasting the development of events. Compared to the traditional methods (such as probability and statistics method, mathematical model), ANN offered better performance in classification and diagnosis. It can be said that ANN is one of the most advanced artificial intelligence technologies.

CAD studies reached a low ebb during the years after the 1960s until the early 1980s. This is because people hope to realize automatic diagnosis with the help of the computer, which is expected too much. The CAD studies are limited due to the lack of corresponding theoretical algorithms and theoretical analysis. This

dual dilemma existed until the late 1980s and early 1990s. Thanks to the rapid development of computer, mathematics, and statistics technologies, CAD has been improved qualitatively. In recent years, the CAD system has been rapidly applied in medical imaging and has made gratifying achievements in some developed countries.
