Artificial intelligence (AI), primarily deep learning, is transforming medical imaging. Deep learning processes intricate data using layered neural networks. These techniques are also particularly applicable in computed tomography (CT), which utilizes X-rays to generate images, and magnetic resonance imaging (MRI), which utilizes magnetic fields and radio waves to generate detailed images. This review reviews advances made recently in AI detection, segmentation, classification, reconstruction, and the fusion of different data modalities in the CT and MRI fields. It also examines their impact in actual clinical practice. Studies show that deep learning models perform well in detecting lung nodules, brain tumors, and imaging of cardiovascular and abdominal organs. Besides this, AI methods based on reconstruction can shorten CT radiation doses and speed up MRI. And the best image quality is preserved. Despite these results, obstacles remain that stop AI from being used for medical imaging. Most difficult problems encountered include limited testing on external data of different types at different institutions, the performance differences across different kinds of datasets, also deep learning models make decisions quite obscurely. Other persisting problems are how to make models more readable, how they can be integrated within clinical workflows, and regulations. The need to build more artificial intelligence that integrates imaging with clinical, genetic, and pathological data is an essential issue for future studies. Such methods could help predict better, personalize medicine. AI alone has potentially transformed radiology: from finding new ways to make diagnoses more efficient, standardizing interpretation of images and supporting clinical decision-making. Nevertheless, AI ought to be viewed as a tool to aid radiologists — not a replacement. Such studies should be focused on testing AI in actual clinical environments, increasing the transparency of algorithms, and establishing standardized measurement methods for performance. These stages are required to securely and effectively embed AI in routine CT and MRI operations.
Paper
Full text
AI APPLICATIONS IN CT AND MRI IMAGING: CURRENT ADVANCES AND LIMITATIONS. A NARRATIVE REVIEW
Semantic Scholar · 2026
Abstract
Artificial intelligence (AI), primarily deep learning, is transforming medical imaging. Deep learning processes intricate data using layered neural networks. These techniques are also particularly applicable in computed tomography (CT), which utilizes X-rays to generate images, and magnetic resonance imaging (MRI), which utilizes magnetic fields and radio waves to generate detailed images. This review reviews advances made recently in AI detection, segmentation, classification, reconstruction, and the fusion of different data modalities in the CT and MRI fields. It also examines their impact in actual clinical practice. Studies show that deep learning models perform well in detecting lung nodules, brain tumors, and imaging of cardiovascular and abdominal organs. Besides this, AI methods based on reconstruction can shorten CT radiation doses and speed up MRI. And the best image quality is preserved. Despite these results, obstacles remain that stop AI from being used for medical imaging. Most difficult problems encountered include limited testing on external data of different types at different institutions, the performance differences across different kinds of datasets, also deep learning models make decisions quite obscurely. Other persisting problems are how to make models more readable, how they can be integrated within clinical workflows, and regulations. The need to build more artificial intelligence that integrates imaging with clinical, genetic, and pathological data is an essential issue for future studies. Such methods could help predict better, personalize medicine. AI alone has potentially transformed radiology: from finding new ways to make diagnoses more efficient, standardizing interpretation of images and supporting clinical decision-making. Nevertheless, AI ought to be viewed as a tool to aid radiologists — not a replacement. Such studies should be focused on testing AI in actual clinical environments, increasing the transparency of algorithms, and establishing standardized measurement methods for performance. These stages are required to securely and effectively embed AI in routine CT and MRI operations.