No abstract

No abstract

No abstract

Machine learning as a field of artificial intelligence is increasingly applied in medicine to assist patients and physicians. Growing datasets provide a sound basis with which to apply machine learning methods that learn from previous experiences. This review explains the basics of machine learning and its subfields of supervised learning, unsupervised learning, reinforcement learning and deep learning. We provide an overview of current machine learning applications in rheumatology, mainly supervised learning methods for e-diagnosis, disease detection and medical image analysis. In the future, machine learning will be likely to assist rheumatologists in predicting the course of the disease and identifying important disease factors. Even more interestingly, machine learning will probably be able to make treatment propositions and estimate their expected benefit (e.g. by reinforcement learning). Thus, in future, shared decision-making will not only include the patient’s opinion and the rheumatologist’s empirical and evidence-based experience, but it will also be influenced by machine-learned evidence.

Purpose To test the potential of Dixon T2-weighted fat-only sequences to replace T1-weighted sequences for the detection of bone metastases, with the hypothesis that diagnostic performance with an alternative magnetic resonance (MR) imaging protocol (sagittal spin-echo Dixon T2-weighted fat-only and water-only imaging) would not be inferior to that with the standard protocol (sagittal spin-echo T1-weighted and spin-echo Dixon T2-weighted water-only imaging). Materials and Methods A total of 121 consecutive whole-spine MR imaging examinations (63 men; mean age ± standard deviation, 61.4 years ± 11.8) performed for suspected vertebral bone metastases were included in this retrospective, institutional review board-approved study. Quantitative image analysis was performed for 30 randomly selected spine levels. Qualitative analysis was performed separately by two musculoskeletal radiologists, who registered the number of metastases for each spine level. Areas under the curve with the protocols were compared on the basis of nonparametric receiver operating characteristic curve estimations by using a noninferiority test on paired data, with a best valuable comparator as a reference. Interobserver and interprotocol agreement was assessed by using κ statistics. Results Contrast-to-noise ratio was significantly higher on the alternative protocol images than on the standard protocol images (181.1 [95% confidence interval: 140.4, 221.7] vs 84.7 [95% confidence interval: 66.3, 103.1] respectively; P < .001). Diagnostic performance was not significantly inferior with the alternative protocol than with the standard protocol for both readers in a per-patient analysis (sensitivity, 97.9%-98.9% vs 93.6%-97.9%; specificity, 85.2%-92.6% vs 92.6%-96.3%; area under the curve, 0.92-0.96 vs 0.95, respectively; all P ≤ .02) and a per-spine level analysis (all P < .01). Interobserver and interprotocol agreement was good to very good (κ = 0.70-0.81). Conclusion Dixon T2-weighted fat-only and water-only imaging provide, in one sequence, diagnostic performance similar to that of the standard combination of morphologic sequences for the detection of probable spinal bone metastases, thereby providing an opportunity to reduce imaging time by eliminating the need to perform T1 sequences. RSNA, 2017 An earlier incorrect version of this article appeared online. This article was corrected on November 6, 2017.