Artificial intelligence, machine learning, computer-aided diagnosis, and radiomics: advances in imaging towards to precision medicine

Abstract

The discipline of radiology and diagnostic imaging has evolved greatly in recent years. We have observed an exponential increase in the number of exams performed, subspecialization of medical fields, and increases in accuracy of the various imaging methods, making it a challenge for the radiologist to "know everything about all exams and regions". In addition, imaging exams are no longer only qualitative and diagnostic, providing now quantitative information on disease severity, as well as identifying biomarkers of prognosis and treatment response. In view of this, computer-aided diagnosis systems have been developed with the objective of complementing diagnostic imaging and helping the therapeutic decision-making process. With the advent of artificial intelligence, "big data", and machine learning, we are moving toward the rapid expansion of the use of these tools in daily life of physicians, making each patient unique, as well as leading radiology toward the concept of multidisciplinary approach and precision medicine. In this article, we will present the main aspects of the computational tools currently available for analysis of images and the principles of such analysis, together with the main terms and concepts involved, as well as examining the impact that the development of artificial intelligence has had on radiology and diagnostic imaging.

Keywords: Artificial intelligence; Computer aided diagnosis; Machine learning; Radiomics.

Figure 1

Representation of the function (matrix) of a gray-scale digital image (axial slice of a chest CT).

Figure 2

Semiautomated segmentation of the lung on a CT scan of the chest with 256 gray levels and a user-defined threshold of 115 Hounsfield units: the original CT image (image on the left) is thresholded (to detect the edges) and transformed into a binary image (to separate the lungs).

Figure 3

Semiautomated segmentation, with region growing, of a neoplastic pulmonary nodule on chest CT after placement of a user-defined seed pixel (point at the center of the nodule in the first image on the left).

Figure 4

Example of the histogram of an axial CT scan of the chest with 256 gray levels. The histogram shows the distribution of pixels or voxels according to the gray levels (or Hounsfield units, if necessary).

Figure 5

Segmentation of pulmonary nodules on chest CT in two different patients, both nodules having irregular contours and being in proximity with adjacent vessels or chest wall, factors that make it difficult to segment the edges of lesions properly. In such cases, texture analysis can facilitate proper segmentation.

Figure 6

Architecture of a multilayer ANN. The input layer receives the feature information. The output layer represents classes or outcomes (e.g., normal versus pathological).

Figure 7

Chest CT image with a pulmonary nodule as input into a CNN for analysis using deep learning technique.

Figure 8

Example of ROC curves: curve 1 represents a test with perfect performance (AUC = 1.0); curve 2 represents a test with intermediate performance (AUC = 0.75); and curve 3 represents a random test (AUC = 0.50).

Figure 9

Example of a CAD tool for detection and analysis of pulmonary nodules. The program not only indicates the presence of a right apical pulmonary nodule but also provides quantitative and three-dimensional information regarding that nodule.

Figure 10

Quantification of the heterogeneity of a pulmonary adenocarcinoma on a CT scan of the chest by radiomic/radiogenomic evaluation. Color scale refers to a feature extracted from the image, reflecting tissue and genetic subregions of the tumor.