Deep learning in rheumatological image interpretation

Abstract

Artificial intelligence techniques, specifically deep learning, have already affected daily life in a wide range of areas. Likewise, initial applications have been explored in rheumatology. Deep learning might not easily surpass the accuracy of classic techniques when performing classification or regression on low-dimensional numerical data. With images as input, however, deep learning has become so successful that it has already outperformed the majority of conventional image-processing techniques developed during the past 50 years. As with any new imaging technology, rheumatologists and radiologists need to consider adapting their arsenal of diagnostic, prognostic and monitoring tools, and even their clinical role and collaborations. This adaptation requires a basic understanding of the technical background of deep learning, to efficiently utilize its benefits but also to recognize its drawbacks and pitfalls, as blindly relying on deep learning might be at odds with its capabilities. To facilitate such an understanding, it is necessary to provide an overview of deep-learning techniques for automatic image analysis in detecting, quantifying, predicting and monitoring rheumatic diseases, and of currently published deep-learning applications in radiological imaging for rheumatology, with critical assessment of possible limitations, errors and confounders, and conceivable consequences for rheumatologists and radiologists in clinical practice.Deep learning is a powerful technique with great potential for the analysis and interpretation of rheumatological images. To successfully use deep learning, rheumatologists should understand the tasks involved in image processing and the potential confounders and limitations that can affect the analysis of clinical data.The number of research studies on deep learning in rheumatological imaging has grown rapidly during the past 5 years, but they mainly consist of pilot studies that require external validation.Confounding factors and errors in deep-learning methods need to be ruled out before deep learning can be applied in clinical practice, for which the intended use should be strictly defined.Deep-learning techniques, together with mapping to explain their reasoning, will enable hypothesis-free image analysis and could identify new imaging biomarkers.Deep learning might assist rheumatologists and radiologists in interpreting rheumatological images, increasing their diagnostic, prognostic and monitoring accuracy, and decreasing workloads and costs.