Artificial Intelligence in Magnetic Resonance Imaging: A Feasible Practice?

Abstract

In this new decade, artificial intelligence (AI) is being used extensively in various applications in medicine, especially imaging (including image processing and interpretation). Future possible applications of AI in imaging could include AI-guided intervention and AI-only diagnosis. It was first used as term in 1950s and, in the last few decades, it has developed rapidly because of the introduction of machine and deep learning [[1]Lakhani P. Prater A.B. Hutson R.K. et al.Machine learning in radiology: applications beyond image interpretation.J Am Coll Radiol. 2018; 15: 350-359Abstract Full Text Full Text PDF PubMed Scopus (95) Google Scholar]. Machine learning includes self-improving algorithms, whereas deep learning systems are designed to be multilayered, thus being able to perform human-like actions, such as classifying data and detecting differences [[1]Lakhani P. Prater A.B. Hutson R.K. et al.Machine learning in radiology: applications beyond image interpretation.J Am Coll Radiol. 2018; 15: 350-359Abstract Full Text Full Text PDF PubMed Scopus (95) Google Scholar]. The applications of AI in medical imaging are more recent—in the last three years, there have been seven times more publications on the topic compared with a decade ago. This shows the exponential integration of AI in radiology, especially in magnetic resonance imaging (MRI) and computed tomography [[2]Pesapane F. Codari M. Sardanelli F. Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine.Eur Radiol Exp. 2018; 2: 35Crossref PubMed Scopus (223) Google Scholar]. Neuroradiology and MRI are the main field of innovation; there are some clinical surveys showing that AI in MRI could contribute to breast lesions [[3]Dalmiş M.U. Gubern-Mérida A. Vreemann S. et al.Artificial intelligence-based classification of breast lesions imaged with a multiparametric breast MRI protocol with ultrafast DCE-MRI, T2, and DWI.Invest Radiol. 2019; 54: 325-332Crossref PubMed Scopus (48) Google Scholar], degenerative diseases [[4]Liu X. Chen K. Wu T. Weidman D. Lure F. Li J. Use of multimodality imaging and artificial intelligence for diagnosis and prognosis of early stages of Alzheimer's disease.Transl Res. 2018; 194: 56-67Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar], musculoskeletal issues [[5]Hirschmann A. Cyriac J. Stieltjes B. Kober T. Richiardi J. Omoumi P. Artificial Intelligence in Musculoskeletal Imaging: Review of Current Literature, Challenges, and Trends.Semin Musculoskelet Radiol. 2019; 23: 304-311Crossref PubMed Scopus (23) Google Scholar,[6]Roblot V. Giret Y. Bou Antoun M. et al.Artificial intelligence to diagnose meniscus tears on MRI.Diagn Interv Imaging. 2019; 100: 243-249Crossref PubMed Scopus (39) Google Scholar], and rectal cancer [[7]Ferrari R. Mancini-Terracciano C. Voena C. et al.MR-based artificial intelligence model to assess response to therapy in locally advanced rectal cancer.Eur J Radiol. 2019; 118: 1-9Abstract Full Text Full Text PDF PubMed Scopus (30) Google Scholar]. The main dilemma is waiting to see if AI will become a radiologist's tool or a radiologist substitute. A recent trial found that a deep learning machine was more accurate than experienced radiologists in screening and diagnosing lung cancer [[8]Ardila D. Kiraly A.P. Bharadwaj S. et al.End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography.Nat Med. 2019; 25: 954-961Crossref PubMed Scopus (562) Google Scholar]. These surveys create hesitation and disbelief in the radiologist community, with thoughts that their specialty will become extinct because of AI's ‘domination.’ In our opinion, AI will not become a substitute for experts in medical imaging, but it will be responsible for a change in everyday practice. Radiologists are more than simple image interpreters because they combine recognition of pathological findings with the clinical manifestations, laboratory examinations, and medical history of patients. In addition, a radiologist, being firstly a physician, is able to communicate with the patient and other clinicians to help the diagnostic approach and the right choice of optimal medical treatment. However, AI's features are precious tools in the hands of experienced and well-trained radiologists. Deep learning systems could extract more detailed information, which are part of the data set and are not included in the actual image, thus improving diagnostic accuracy. Moreover, these clever machines can take the burden off radiologists when it comes to tedious clinical routine; this allows physicians to approach cases multivariably and not only in the fashion of repetitive image labelling. Furthermore, this change in the everyday practice enables experts to have a more productive workday; as with the contribution of AI, diagnostic time is reduced but at the same time, the standards are raised because of less fatigue. In conclusion, AI will surely change the landscape in clinical practice. Radiologists should capitalize on the impact of innovative technology to reach the potential of higher diagnostic accuracy. In our opinion, AI will not replace the specialty of radiology, but it will replace certain tasks while complementing overall decision-making.