This narrative review explores recent advancements and applications of modern low-field (≤ 1 Tesla) magnetic resonance imaging (MRI) in musculoskeletal radiology. Historically, high-field MRI systems (1.5 T and 3 T) have been the standard in clinical practice due to superior image resolution and signal-to-noise ratio. However, recent technological advancements in low-field MRI offer promising avenues for musculoskeletal imaging. General principles of low-field MRI systems are being introduced, highlighting their strengths and limitations compared to high-field counterparts. Emphasis is placed on advancements in hardware design, including novel magnet configurations, gradient systems, and radiofrequency coils, which have improved image quality and reduced susceptibility artifacts particularly in musculoskeletal imaging. Different clinical applications of modern low-field MRI in musculoskeletal radiology are being discussed. The diagnostic performance of low-field MRI in diagnosing various musculoskeletal pathologies, such as ligament and tendon injuries, osteoarthritis, and cartilage lesions, is being presented. Moreover, the discussion encompasses the cost-effectiveness and accessibility of low-field MRI systems, making them viable options for imaging centers with limited resources or specific patient populations. From a scientific standpoint, the amount of available data regarding musculoskeletal imaging at low-field strengths is limited and often several decades old. This review will give an insight to the existing literature and summarize our own experiences with a modern low-field MRI system over the last 3 years. In conclusion, the narrative review highlights the potential clinical utility, challenges, and future directions of modern low-field MRI, offering valuable insights for radiologists and healthcare professionals seeking to leverage these advancements in their practice.

To evaluate the reader’s diagnostic performance against the ground truth with and without the help of a novel content-based image retrieval system (CBIR) that retrieves images with similar CT patterns from a database of 79 different interstitial lung diseases. We evaluated three novice readers’ and three resident physicians‘ (with at least three years of experience) diagnostic performance evaluating 50 different CTs featuring 10 different patterns (e.g., honeycombing, tree-in bud, ground glass, bronchiectasis, etc.) and 24 different diseases (sarcoidosis, UIP, NSIP, Aspergillosis, COVID-19 pneumonia etc.). The participants read the cases first without assistance (and without feedback regarding correctness), and with a 2-month interval in a random order with the assistance of the novel CBIR. To invoke the CBIR, a ROI is placed into the pathologic pattern by the reader and the system retrieves diseases with similar patterns. To further narrow the differential diagnosis, the readers can consult an integrated textbook and have the possibility of selecting high-level semantic features representing clinical information (chronic, infectious, smoking status, etc.). We analyzed readers’ accuracy without and with CBIR assistance and further tested the hypothesis that the CBIR would help to improve diagnostic performance utilizing Wilcoxon signed rank test. The novice readers demonstrated an unassisted accuracy of 18/28/44%, and an assisted accuracy of 84/82/90%, respectively. The resident physicians demonstrated an unassisted accuracy of 56/56/70%, and an assisted accuracy of 94/90/96%, respectively. For each reader, as well as overall, Sign test demonstrated statistically significant (p < 0.01) difference between the unassisted and the assisted reads. For students and physicians, Chi²-test and Mann-Whitney-U test demonstrated statistically significant (p < 0.01) difference for unassisted reads and statistically insignificant (p > 0.01) difference for assisted reads. The evaluated CBIR relying on pattern analysis and featuring the option to filter the results of the CBIR by predominant characteristics of the diseases via selecting high-level semantic features helped to drastically improve novices’ and resident physicians’ accuracy in diagnosing interstitial lung diseases in CT.

Contrast agents are frequently administered in computed tomography (CT) scans used for opportunistic screening of osteoporosis. The objective of this study is to compare the impact of contrast-related bone mineral density (BMD) increase between phantom-based and internal CT calibration techniques. Phantom-based and internal CT calibration techniques were used to determine trabecular BMD in 93 existing clinical CT scans of the lumbar spine of 34 subjects, scanned before and after administration of contrast agents. BMD differences between native (Nat) and contrast-enhanced scans in arterial (Art) and portal venous (PV) phases were compared among calibration methods. Three pairs of internal reference materials were investigated: blood/air, subcutaneous adipose tissue (SAT)/air and muscle tissue/air. Contrast agents increased CT values of blood in the aorta and the inferior vena cava in the Art phase by 210 ± 99HU and 47 ± 35HU, respectively and in the PV phase by 110 ± 49HU and 73 ± 28HU, respectively. Effects on CT values of muscle and SAT were < 5HU in the Art phase and < 10HU in the PV phase. BMD obtained with phantom calibration increased significantly (p < 0.01) in Art and PV phases by 8.6 ± 14.7 mg/cm³ and 14.4 ± 16.3 mg/cm³, respectively. If SAT/air or muscle/air were used as internal reference materials, increases in internally calibrated BMD (<12 mg/cm³ in AT phase and < 17 mg/cm³ in PV phase) did not significantly differ (p > 0.1 for Art and PV phases) from increases of phantom-calibrated BMD. Although CT values of internal reference materials increased following the administration of contrast, internal calibration was unable to mitigate the contrast-related BMD increases observed in phantom-based calibration.

To evaluate patients’ radiation exposure undergoing CT-guided joint injection in preparation of MR-arthrography. We developed a novel ultra-low-dose protocol utilizing tin-filtration, performed it in 60 patients and compared the radiation exposure (DLP) and success rate to conventional protocol (26 cases) and low-dose protocol (37 cases). We evaluated 123 patients’ radiation exposure undergoing CT-guided joint injection from 16 January–21 March. A total of 55 patients received CT-guided joint injections with various other examination protocols and were excluded from further investigation. In total, 56 patients received shoulder injection and 67 received hip injection with consecutive MR arthrography. The ultra-low-dose protocol was performed in 60 patients, the low-dose protocol in 37 patients and the conventional protocol in 26 patients. We compared the dose of the interventional scans for each protocol (DLP) and then evaluated success rate with MR-arthrography images as gold standard of intraarticular or extracapsular contrast injection. There were significant differences when comparing the DLP of the ultra-low-dose protocol (DLP 1.1 ± 0.39; p < 0.01) to the low dose protocol (DLP 5.3 ± 3.24; p < 0.01) as well as against the conventional protocol (DLP 22.9 ± 8.66; p < 0.01). The ultra-low-dose protocol exposed the patients to an average effective dose of 0.016 millisievert and resulted in a successful joint injection in all 60 patients. The low dose protocol as well as the conventional protocol were also successful in all patients. The presented ultra-low-dose CT-guided joint injection protocol for the preparation of MR-arthrography demonstrated to reduce patients’ radiation dose in a way that it was less than the equivalent of the natural radiation exposure in Germany over 3 days—and thereby, negligible to the patient.

We thank Prof. Boyd and colleagues for their interesting comments highlighting the growing importance of internal bone mineral density (BMD) calibration techniques for use in opportunistic screening. We agree that the phantomless BMD calibration method developed at the University of Calgary ¹ is easier to use than other methods because it does not require an external reference data set. We had addressed this point in our publication ² in the first sentence of the section “Phantomless calibration method 4: Voxel-specific calibration.” It is true that at the time of writing, we were not aware of the public availability of the voxel-specific calibration method developed in Calgary. In retrospect, we have noted that a reference to the website appears to have been published only in an article focusing on muscle density calibration. ³ We congratulate the authors for providing a public domain tool. We have now implemented the Calgary software as provided in the Github archive and compared the results with the software we implemented for our publication. ² There were two remaining differences between the two software. First, as pointed out in the authors’ letter, the Calgary software is based on 5 internal reference materials. Thus, for the comparison specifically performed to address this letter, we used the same set of materials in our software. Second, the Calgary software used mass absorption coefficients (MAC) to determine the effective energy, although the paper by Millner et al. ⁴ cited by Michalski as a reference for the method to determine the effective energy suggested the use of linear absorption coefficients of the internal reference materials. Consequently, we had also used linear absorption coefficients for all calculations performed for our publication. ² However, for the purpose of comparison, we have now implemented MAC in our software. We applied both softwares to our DS1 dataset of 350 computed tomography (CT) scans from female residents of Minnesota, described in detail in our publication. The BMD accuracy errors obtained with the Calgary software were virtually identical to our software (modified for comparison as described above), as can be seen in Figure 1.