New trend in artificial intelligence-based assistive technology for thoracic imaging

Abstract

Although there is no solid agreement for artificial intelligence (AI), it refers to a computer system with intelligence similar to that of humans. Deep learning appeared in 2006, and more than 10 years have passed since the third AI boom was triggered by improvements in computing power, algorithm development, and the use of big data. In recent years, the application and development of AI technology in the medical field have intensified internationally. There is no doubt that AI will be used in clinical practice to assist in diagnostic imaging in the future. In qualitative diagnosis, it is desirable to develop an explainable AI that at least represents the basis of the diagnostic process. However, it must be kept in mind that AI is a physician-assistant system, and the final decision should be made by the physician while understanding the limitations of AI. The aim of this article is to review the application of AI technology in diagnostic imaging from PubMed database while particularly focusing on diagnostic imaging in thorax such as lesion detection and qualitative diagnosis in order to help radiologists and clinicians to become more familiar with AI in thorax.

Keywords: Artificial intelligence; Convolutional neural network; Deep learning; Explainable AI; Thoracic imaging; Vision transformer.

Fig. 1

Flowchart of studies selection process

Fig. 2

Nodule detection on chest radiograph with bone suppression by a commercially available AI model. A 42-year-old woman. A low radiolucent area is suspected on the right supraclavicular bone (arrow, A). However, it is difficult to distinguish between pulmonary and bone lesions. A pulmonary nodule hidden behind the right clavicle can be detected on the chest radiograph with bone suppression using a commercially available AI model (arrow, B). A part-solid ground-glass nodule with about 3 cm in diameter can be seen in the right upper lobe (arrow, C). This nodule was confirmed as a lung adenocarcinoma after surgery. AI artificial intelligence

Fig. 3

Nodule detection on chest radiograph by a commercially available AI-based CAD system. A 70-year-old man. A nodule can be seen in the right middle lung field and a mass can be seen in the left lower lung field overlapping the cardiac shadow (arrows, A). A commercially available AI-based CAD system can highlight both the nodule and mass. A nodule with 2 cm in diameter can be seen in the right upper lobe on CT (arrow, C). A mass with 5 cm in diameter can be seen in the left lower lobe on CT (arrow, D). This mass was confirmed as a lung metastasis from renal cell carcinoma after biopsy. AI artificial intelligence, CAD computer-aided detection/diagnosis

Fig. 4

Quantification by a commercially available AI-based CAD system. A commercially available AI-based CAD system can automatically segment a part-solid ground-glass nodule, providing quantitative values that are exactly the same regardless of the observers. AI artificial intelligence, CAD computer-aided detection/diagnosis

Fig. 5

Differences of attention map in the diagnosis of lung nodule. This is a case confirmed as invasive adenocarcinoma after surgery (a and d, axial images; b and e, coronal images; c and f, sagittal images). Red areas in a, b, and c are focused by the AI in the diagnosis of invasive adenocarcinoma. Even for the same nodule, red areas in d, e, and f are focused when diagnosed as noninvasive adenocarcinoma. The numbers in the upper row indicate the diagnostic probability of invasive adenocarcinoma (maximum value of 1) calculated by the AI. Depending on the purpose of diagnosis and the cross section of the image, the focused part and the probability of results by the AI may differ. AI artificial intelligence