Low-field magnetic resonance imaging (MRI) is experiencing a renaissance, with increased interest in the radiology community.1–4 New-generation 0.55 T low-field MRI systems comprise the hardware and software developments of contemporary high-field MRI machines, including flexible phased-array coils, partial Fourier image acquisition, parallel imaging, and compressed sensing techniques. These methods foremost allow for imaging acceleration. More recently, image reconstruction algorithms using neural networks were made commercially available, allowing to enhance image resolution and further accelerate image acquisition.5,6 These algorithms are available on the new-generation low-field MRI systems, improving their diagnostic capabilities.

Lower magnet strength may be beneficial in scenarios prone to large variations in susceptibility, such as metal implant imaging, but foremost provides economic benefits.7,8 Low-field systems are cheaper to manufacture, install, transport, maintain, and operate.4 Compared with a conventional, contemporary 1.5 T MRI system, the low-field machine has a 40%–45% lower purchase price. Lower weight and smaller dimensions reduce installation and transportation costs by approximately 70%. Moreover, given a similar service plan, annual servicing costs can be reduced by 45% when compared with a high-field MRI system.9 Low-field MRI machines are ultimately aimed to increase the global availability of MRI. As such, new-generation low-field MRI needs to be evaluated for common MRI referrals.

Internal derangement of the knee is one of the most common indications for MRI of the musculoskeletal system, and consequently, the knee is a commonly imaged joint. Magnetic resonance imaging is considered the imaging standard, due to the superior soft tissue contrast when compared with other imaging modalities.10 Although both 1.5 T and 3 T MRI can provide diagnostic images, 3 T MRI is generally preferred at most institutions, providing higher image resolution. This is in keeping with a contemporary meta-analysis, suggesting benefits of 3 T over 1.5 T knee MRI.11 Moreover, increasing literature is published regarding the potential of ultra-high field 7 T MRI of the knee, possibly offering superior assessment of joint cartilage.12 Perhaps contradicting this trend of greater field strength knee MRI for ever more detailed analyses, we hypothesized that new-generation low-field MRI allows for diagnostic knee examinations in clinical practice and is able to answer common referrer queries. These include the identification of trauma sequelae of the ligaments, tendons, menisci, bone, and joint cartilage. Thus, the purpose of our study was to compare the detection rate of and reader confidence in new-generation 0.55 T knee MRI findings with 3 T knee MRI in patients with knee pain after acute trauma.

MATERIALS AND METHODS Patients

This prospective single-center study was approved by the local ethics committee, and written informed consent was obtained from each patient. Twenty-five consecutive adult patients (11 women, 14 men; median age, 38 [range, 21–63] years) who consented to repeat MRI of the symptomatic knee were included. Magnetic resonance imaging was performed between August 2022 and February 2023 for queries regarding internal derangement in all knees.

Study inclusion criteria were as follows: adult patients (age ≥18 years); valid referral for MRI of the knee with suspected internal derangement including suspected meniscal, tendon, ligament, chondral, and/or bone pathology after acute trauma; available signed consent for study participation. Exclusion criteria were general MRI contraindications including any installed stimulating device such as cardiac pacemakers and prior arthroplasty of the symptomatic knee. All patients who consented to study participation were included in the given time interval.

MRI Examinations

In one setting, 0.55 T (Siemens MAGNETOM Free.Max; Siemens Healthineers, Erlangen, Germany) and 3 T MRI (Siemens MAGNETOM Prisma) were performed of the symptomatic knee. Standard diagnostic and routinely clinically used imaging protocols were used. Protocol parameters are summarized in Table 1. 0.55 T MRI was obtained using a 6-channel flex-phased array coil (size medium) and 3 T MRI with an 8-channel receive-transmit knee coil. The 0.55 T system is commercially available. Upon purchase with the scanner provided, deep learning image reconstruction algorithm was used (Deep Resolve Gain and Deep Resolve Sharp; Siemens Healthineers). This allowed for acceleration of imaging times and increased image resolution. This algorithm was not available on the 3 T system. All patients tolerated the examinations well, and none required premature termination of image acquisition.

0.55 T MRI 3 T MRI Parameter PD FS PD FS PD FS T1 PD FS PD FS PD FS T1 Plane Axial Cor Sag Cor Axial Cor Sag Cor Slice thickness 3 mm 3 mm 3 mm 3 mm 3 mm 3 mm 3 mm 3 mm FoV, mm 169 × 169 169 × 169 169 × 169 169 × 169 160 × 160 150 × 150 150 × 160 150 × 150 TR, ms 3230 2360 2560 384 3710 3570 3880 743 TE, ms 30 45 30 15 40 40 42 9.6 Bandwidth, Hz/Px 100 100 100 110 255 150 150 245 GRAPPA factor 2 2 2 2 2 2 2 2 Deep learning image reconstruction Yes Yes Yes Yes No No No No Acquisition time, min 3:00 3:04 3:19 1:59 2:02 3:19 3:37 1:35

0.55 T and 3 T MRI parameters.

FoV, field of view; cor, coronal; sag, sagittal; GRAPPA, generalized autocalibrating partial parallel acquisition.

MRI Analysis

Two board-certified radiologists, 1 subspecialized, fellowship-trained MSK radiologist (reader 1) and 1 MSK radiology fellow (reader 2), reviewed all images on PACS workstations (Sectra, Sweden) in 2 sessions. In the first reading session, each reader individually evaluated 25 knee MRI of the 25 patients in random order, including 1 study of each patient, acquired either on the 0.55 T or 3 T MRI scanner. At least 96 hours after the first reading session, the other 25 examinations were read in random order by the radiologists. At the time of analysis, the radiologists were blinded to the clinical history, previous reports, and sequence parameters.

Magnetic resonance imaging examinations were reviewed for each study, and image quality was graded using the following criteria: edge sharpness, blurring, artifacts, fluid to cartilage contrast, fluid to soft tissue contrast, small ligament delineation, with the meniscofemoral ligaments, popliteomeniscal fascicles, and ligaments of the posterolateral corner used for reference, and noise. In addition, overall image quality was graded. A 5-point Likert scale was used for all parameters, in which “5” indicated optimal image quality, “4” good image quality, “3” average image quality, “2” fair image quality, and “1” poor image quality, limiting diagnostic evaluation.

Readers performed a clinical reading for each joint compartment as per our standardized reporting template. The presence or absence of lesions of each compartment's bone, cartilage, meniscus, ligaments, and tendons was noted. Lesions were interpreted as defined by the glossary of terms for musculoskeletal radiology.13

Cartilage lesions were quantified using the ICRS scoring system.14 Meniscal lesions were further classified by location (anterior horn, body, posterior horn) and graded from 1 to 4 (1, signal alteration; 2, horizontal or vertical longitudinal tear without displacement; 3, complex or radial tear without displacement; 4, displaced meniscus tear).

Ligament and tendon lesions were graded from less to more severe as (1) strain, (2) partial tear, or (3) full thickness tear.

For each clinical parameter evaluated, the readers assigned a score for their level of confidence: 1, questionable; 2, probable; and 3, definite presence or absence of a lesion.

Statistical Analysis

Statistical analyses were performed using commercially available software (IBM SPSS Statistics Version 25; IBM Corp, Armonk, NY). The Wilcoxon test was used to assess the differences in image quality grading and reader confidence levels between 0.55 T and 3 T MRI, with P < 0.05 considered to represent a statistically significant difference. The intraclass correlation coefficient (ICC) was calculated to evaluate the intrarater agreement for MRI findings between 0.55 T and 3 T MRI.

RESULTS Image Quality

All examinations provided diagnostic images and were included for evaluation. None of the readers scored any of the examinations' overall quality as “poor.” However, all image quality parameters and overall image quality were rated higher for the 3 T examinations when compared with the low-field 0.55 T studies (each P ≤ 0.017) by both readers. Mean image quality ratings are shown in Table 2.

TABLE 2 - Image Quality Grading Reader 1 Reader 2 Image Quality 0.55 T 3 T P 0.55 T 3 T P Overall 3.52 ± 0.77 4.84 ± 0.37 <0.001 3.32 ± 0.56 4.56 ± 0.51 <0.001 Edge sharpness 3.92 ± 0.49 4.92 ± 0.28 <0.001 3.20 ± 0.41 4.36 ± 0.57 <0.001 Blurring 3.96 ± 0.79 4.8 ± 0.41 <0.001 3.12 ± 0.33 4.16 ± 0.55 <0.001 Artifacts 4.28 ± 0.79 4.76 ± 0.52 0.017 3.04 ± 0.35 4.16 ± 0.62 <0.001 Fluid-cartilage contrast 3.60 ± 0.71 4.92 ± 0.28 <0.001 2.52 ± 0.51 3.96 ± 0.61 <0.001 Fluid soft tissue contrast 4.00 ± 0.87 4.92 ± 0.28 <0.001 3.28 ± 0.46 4.12 ± 0.44 <0.001 Small ligament delineation 3.56 ± 0.71 4.92 ± 0.28 <0.001 2.56 ± 0.58 4.00 ± 0.57 <0.001 Noise 3.12 ± 0.73 4.8 ± 0.41 <0.001 2.84 ± 0.47 3.96 ± 0.61 <0.001

A 5-point Likert scale was used for all parameters, in which “5” indicated optimal image quality, “4” good image quality, “3” average image quality, “2” fair image quality, and “1” poor image quality, limiting diagnostic evaluation.

Reader 1 rated overall image quality with a mean Likert score of 4.84 at 3 T and 3.52/5 at 0.55 T (P < 0.001). In 3 patients, reader 1 rated overall image quality as equal between 3 T and 0.55 T MRI (Fig. 1). In the remaining 22 patients, overall image quality was rated superior for the 3 T examinations. Superior artifact ratings were assigned in four 0.55 T knee MRI and equal scores in 12 knee MRI when compared with the 3 T studies.

FIGURE 1:

Comparison of image quality and high-grade retropatellar cartilage defect in a 34-year-old patient; full-thickness cartilage defect with adjacent edema (arrowhead) can be appreciated on sagittal (sag) and axial (ax) images. 3 T MRI and 0.55 T MRI were rated as “excellent” overall image quality. Cor indicates coronal.

Reader 2 rated overall image quality of 3 T examinations with a mean score of 4.56 and 3.32/5 points for 0.55 T examinations, respectively (P < 0.001). In 1 patient, 3 T and 0.55 T MRI received equal overall image quality grading, with the remaining patients receiving higher scores for the 3 T MRI. Artifact grading was rated equal between 3 T and 0.55 T MRI in 4 patients, with the remaining 21 patients having superior 3 T MRI scores.

Detection and Grading of Knee MRI Findings

3 T MRI findings of reader 1 were used for the subsequent description of detection and grading of knee MRI findings. In the study cohort, bone marrow edema was found in the medial femorotibial compartment in 6 patients, in the lateral femorotibial compartment in 3 patients and in the trochlear in 2 patients. In addition, 1 medial and 2 lateral compartment fractures were found. Overall, 100 cartilage facets were evaluated on 3 T MRI (4 per patient). Medial femorotibial compartment cartilage lesions were found in 3 patients, lateral lesions in 7 patients, trochlear lesions in 2 patients, and retropatellar lesions in 4 patients. Medial meniscus lesions were found in 13 patients and lateral meniscus lesions in 4 patients. Eight patients had medial collateral ligament lesions and 2 patients lateral collateral ligament lesions. One patient showed a complete quadriceps tendon tear. Anterior cruciate ligament lesions were found in 4 patients, and no posterior cruciate ligament lesions were identified. Baker cysts were present in 9 patients and joint effusion in 19 patients. Articular bodies were found in 2 patients. No lesions of the popliteus or patella tendon were identified. Bone marrow lesions, fractures, cartilage lesions, meniscal lesions, collateral ligament, and tendon lesions are summarized across the different compartments for improved comparison between 3 T and 0.55 T MRI examinations. Table 3 illustrates the comparison of diagnostic findings between 3 T and 0.55 T MRI.

TABLE 3 - Comparison of Clinical Findings Between 3 T and 0.55 T MRI for 2 Readers Reader 1 Reader 2 Finding 3 T 0.55 T ICC 3 T 0.55 T ICC Focal bone marrow edema, n 11 11 1 9 9 1 Fracture, n 3 3 1 1 1 1 Cartilage lesions, n 14 16 0.77 [0.67–0.84] 10 6 0.73 [0.63–0.81] Mean ICRS score of cartilage lesions ± SD 2.4 ± 1.2 2.6 ± 0.9 0.77 [0.49–0.91] 2.1 ± 1.4 1.5 ± 1.8 0.88 [0.21–0.98] Meniscal lesions, n 17 15 0.91 [0.85–0.95] 12 11 0.95 [0.91–0.97] Mean grade of meniscal lesions ± SD 2.6 ± 0.9 2.7 ± 1 0.89 [0.73–0.96] 2.5 ± 1.3 2.3 ± 1.5 0.90 [0.69–0.97] Collateral ligament lesions, n 10 10 1 11 11 1 Mean grade of collateral ligament lesions ± SD 1.8 ± 0.5 1.8 ± 0.5 1 1.7 ± 0.8 1.7 ± 0.8 1 Cruciate ligament lesions, n 4 4 1 2 2 1 Mean grade of cruciate ligament lesions ± SD 2.5 ± 0.8 2.5 ± 0.8 1 3 3 1 Tendon lesions, n 1 1 1 1 1 1 Mean grade of tendon ligament lesions ± SD 3 3 1 3 3 1 Baker cysts, n 9 9 1 3 3 1 Joint effusion, n 19 19 1 9 9 1 Articular bodies, n 2 1 0.66 [0.66–0.83] 1 1 1

Cartilage lesions were quantified using the ICRS scoring system. Meniscus lesions were graded 1, signal alteration; 2, horizontal or vertical longitudinal tear without displacement; 3, complex or radial tear without displacement; 4, displaced meniscus tear. Ligament and tendon lesions were graded as 1, strain; 2, partial tear; or 3, full thickness tear.

SD, standard deviation; ICC, intraclass correlation coefficient (95% confidence interval).

Agreement between 3 T and 0.55 T MRI for the identification of clinical findings was perfect with regard to bone marrow edema and fractures, presence of collateral ligament lesions and grading, presence of cruciate ligament lesions and grading, presence of Baker cysts and joint effusion, as well as for the presence of the full-thickness quadriceps tendon tear, for reader 1 as well as reader 2, respectively (Fig. 2).

FIGURE 2:

Full thickness ACL tear (arrow) and partial thickness cartilage defect (arrowhead) in a 53-year-old patient shown on sagittal (sag) and axial (ax) 3 T and 0.55 T PD-weighted fat-saturated MRI.

Both reader's ratings regarding presence and grading of cartilage lesions as well as presence and grading of meniscal lesions showed good agreement between high- and low-field MRI (each ICC > 0.76), with average agreement for the detection of cartilage lesions by reader 2 (ICC = 0.73) (Fig. 3).

FIGURE 3:

Subtle, vertical longitudinal medial meniscus tear in a 49-year-old patient, which was only identified on 3 T PD-weighted fat-saturated MRI.

Magnetic resonance imaging findings with differences in detection and grading between 3 T and 0.55 T MRI were evaluated in greater detail. Twelve of 14 cartilage lesions identified by reader 1 on 3 T MRI were rated equally between 3 T and 0.55 T MRI. Perfect agreement was found for the four grade 4 cartilage defects and three grade 3 cartilage defects.

Five grade 2 cartilage lesions were identified at 3 T MRI. Among these 5 lesions, 3 were equally identified as grade 2 lesions at 0.55 T. One lesion was assigned grade 1 according to ICRS, and 1 lesion was not identified on 0.55 T MRI. Two grade 1 lesions were identified at 3 T MRI. One was detected and graded equally on 0.55 T MRI. The other lesion was not identified on the corresponding 0.55 T MRI. Two additional grade 2 and 2 additional grade 1 lesions were described at 0.55 T in compartments that were graded as normal on 3 T MRI (combined ICC for grade 2 and 1 cartilage lesions = 0.77). Fifteen of 17 meniscus lesions identified at 3 T were identified and graded equally on 0.55 T MRI by reader 1. All four grade 4 and three grade 3 lesions showed perfect interscanner agreement.

One grade 2 and one grade 1 lesions were not identified at 0.55 T (combined ICC = 0.49) (Fig. 3).

Reader 1 identified articular bodies in 2 patients on 3 T MRI, but only in one of these patients at 0.55 T. In retrospect, on very close inspection, the articular body in question can be identified at 0.55 T (Fig. 4).

FIGURE 4:

Articular body in a 63-year-old patient; note how the articular body can be much better appreciated on the 3 T when compared with 0.55 T PD-weighted fat-saturated MRI (arrow). It is most conspicuous on axial (ax) 0.55 T MRI and with available 3 T MRI can also be identified on sagittal (sag) and coronal (cor) images. On close inspection, further lesions of similar signal and morphology can be noted (arrowheads), which can be comfortably diagnosed on the corresponding radiograph as calcifications, most likely due to calcium pyrophosphate deposition disease.

Reader Confidence

Overall, reader confidence scores were higher for 3 T compared with 0.55 T MRI. Across all 500 evaluated individual findings of the 25 patients mean reader confidence was 2.98 ± 0.167 for 3 T and 2.84 ± 0.446 for 0.55 T MRI (P < 0.001). The comparison of diagnostic confidence levels per finding is shown in Table 4.

TABLE 4 - Mean Scores of Reader Confidence ± Standard Deviation for Clinical Findings at 3 T and 0.55 T MRI Reader 1 Reader 2 Clinical Findings 3 T 0.55 T P 3 T 0.55 T P Bone marrow edema 3 2.98 ± 0.14 0.157 3 3 1 Cartilage 2.99 ± 0.10 2.5 ± 0.64 <0.001 2.41 ± 0.52 2.22 ± 0.50 <0.001 Menisci 2.9 ± 0.46 2.66 ± 0.72 0.041 2.88 ± 0.39 2.48 ± 0.50 <0.001 Collateral ligaments 2.98 ± 0.14 2.84 ± 0.42 0.020 3 2.96 ± 0.20 0.157 Cruciate ligaments 2.98 ± 0.14 2.96 ± 0.20 0.317 3 3 1 Tendons 3 2.99 ± 0.12 0.317 3 2.99 ± 0.12 0.317 Baker cysts 3 3 1 3 3 1 Effusion 3 3 1 3 2.96 ± 0.5 0.317 Articular bodies 3 3 1 2.96 ± 0.2 2.96 ± 0.2 1

For each clinical parameter evaluated, the readers assigned a score for their level of confidence: 1, questionable; 2, probable; and 3, definite presence of absence of a lesion.

No differences in confidence levels for evaluating the presence of bone marrow edema and fractures, cruciate ligament lesions, tendon lesions, Baker cysts, joint effusions, and articular bodies were seen for both readers at 3 T versus 0.55 T MRI (each P > 0.157).

Both readers showed had higher confidence scores when reporting cartilage and meniscal lesions at 3 T compared with 0.55 T MRI (each P < 0.041). In Figure 5, reader 1 also had a higher mean confidence level score for evaluating collateral ligament lesions at 3 T when compared with 0.55 T MRI. Furthermore, reader 2 had a nonsignificantly higher absolute confidence score for the evaluation of collateral ligaments at 3 T versus 0.55 T MRI (P = 0.157).

FIGURE 5:

Low-grade, partial thickness cartilage lesion (arrowhead) in a 44-year-old patient. The lateral, fibular cartilage is difficult to evaluate on the 0.55 T PD-weighted fat-saturated sequences with low image quality (grade 2/5). Note the anteriorly displaced lateral meniscal tear (arrow).

DISCUSSION

In this prospective comparative study, we found that new-generation 0.55 T MRI, using a deep learning image reconstruction algorithm, allows for reliable detection of bone marrow edema, fractures, high-grade cartilage and meniscus lesions, ligament and tendon lesions, Baker cysts, and joint effusions with high reader confidence. In comparison with 3 T MRI, diagnostic accuracy for detection and grading of subtle, low-grade cartilage, and meniscal lesions is limited at 0.55 T. Correspondingly, reader confidence was reduced for the evaluation of cartilage and menisci at 0.55 T compared with 3 T MRI. To the authors' best knowledge, this is the first clinical study comparing the diagnostic performance of new-generation 0.55 T knee MRI with 3 T MRI in a patient population.

A recent evaluation of 20 volunteers without knee pain or acute trauma compared the image quality of knee MRI between 1.5 T and the same new-generation 0.55 T machine that was used in our study.15 In agreement with our findings, the authors concluded that 0.55 T MRI yielded diagnostic images. They also described equal diagnostic performance for the assessment of cartilage and meniscal pathologies at 0.55 T compared with 1.5 T MRI. It needs to be emphasized that only healthy volunteers were included in this publication, and only chronic findings were described, whereas we evaluated the performance of 0.55 T in symptomatic patients after acute trauma. In addition, we compared our 0.55 T MRI findings with 3 T MRI, which may be a superior criterion standard when compared with 1.5 T MRI, especially for low-grade cartilage and meniscus lesions.11,16,17 In fact, superior sensitivity for the detection of 39 arthroscopically proven cartilage defects was shown for 3 T (75.6%) versus 1.5 T MRI (70.6%).16

Previous studies concerning low-field knee MRI used earlier-generation 0.2 T and 0.3 T MRI machines. One research group found good diagnostic performance of dedicated 0.3 T knee MRI for detection of meniscal and cruciate ligament lesions, but limited accuracy for chondral lesions when compared with arthroscopy findings.18 The authors described limited accuracy for low grade (I and II) cartilage defects, which is in agreement with our study. Other researchers using previous generation machines similarly found good diagnostic performance of low-field MRI systems when compared with arthroscopy, also highlighting the limitations for cartilage and meniscus evaluation.19–21 Overall, 12/14 (86%) cartilage lesions and 15/17 (88%) meniscus lesions identified on 3 T MRI were also identified and equally graded on 0.55 T MRI. One may hypothesize that a similar difference could be found when comparing 3 T with 1.5 T knee MRI, especially when considering low-grade lesions.

The new-generation 0.55 T scanner used in this study benefits from integrated deep learning–based image reconstruction (Deep Resolve, Siemens Healthineers). Consequently, image acquisition times were comparable between 3 T and 0.55 T MRI sequences. We used a routinely used clinical knee imaging protocol, comprising 2-fold accelerated (GRAPPA) 2D turbo spin echo sequences. Although further acceleration of image acquisition, especially at 3 T, is feasible by using simultaneous multislice acquisition techniques, rapid imaging was not the goal of this analysis.22 Instead, we aimed to perform a comparative study between imaging protocols as they would be used in clinical routine. Although further imaging acceleration can be achieved by using compressed sensing and synthetic MRI techniques, they require advanced licensing and are not part of our routine protocols.23

The 0.55 T scanner used in this study benefits from an 80-cm bore and flex-coil technology, which can improve patient comfort when compared with conventional high-field MRI systems.8 Low-field MRI may also offer potential to improve patient safety. Decreased specific absorption rates, less risk for metallic projectiles, and a generally reduced risk for device interactions when compared with conventional high-field MRI are benefits of 0.55 T MRI.4 Accordingly, 0.55 T MRI can play a role for large metal implant imaging, due to the lower sensitivity to susceptibly alterations when compared with high-field MRI.7 For imaging diagnostics of joints without metal implants, the superior image quality of 3 T MRI is unquestioned. Consequently, and frankly unsurprising, 3 T MRI provided better image quality, and each individual image quality–related parameter was rated superior by both readers for 3 T when compared with 0.55 T MRI in our study. This is in agreement with the current body of literature showing superior image quality at 3 T when compared with 1.5 T MRI.16 The superior image quality lead to higher reader confidence at 3 T MRI for the detection and grading of lesions depending on high contrast, high-resolution images, namely, cartilage and meniscal lesions, as demonstrated in our study. With regard to the assessment of anatomical structures less dependent on subtle signal alterations, such as the continuity of ligaments and tendons, 0.55 T allows for accurate diagnosis with high reader confidence. Most importantly, no high-grade cartilage or meniscus lesions, as defined by 3 T MRI, were missed or falsely classified at 0.55 T.

One pitfall of low-field MRI needs to be noted, which may become increasingly meaningful with further hardware distribution and increasing frequency of low-field joint imaging. Most radiologists are used to a spectrum of susceptibility-related artifacts that are used for diagnostic purposes, for example, for the detection of calcium or blood products. The reduced impact of alterations in tissue susceptibility on low-field MRI signal reduces these artifacts, and consequently, some findings may become loss conspicuous on anatomic low-field MRI. This was shown for calcifications in this study and may also play a role for diagnoses partially relying on blood-product associated susceptibility artifacts such as pigmented villonodular synovitis and hemophilic antropathy; however, further research is needed on this matter. Overall, we found that new-generation low-field 0.55 T MRI is a valid alternative for conventional knee MRI in patients with acute trauma and pain to rule out significant injury.

This prospective comparative study has several limitations. First, only 25 patients were included, and consequently, the number of pathologic lesions was relatively low. For example, only 1 tendon lesion was found in this cohort. Although larger patient cohorts are desirable, studies requiring repeat examinations are time and resource intensive and therefore commonly limited to small patient numbers.24 Second, 3 T MRI is a flawed criterion standard as no arthroscopy findings were available for the presented patients. However, we specifically aimed to compare 0.55 T with 3 T MRI for an imaging comparison. Third, the MRI vendor has recently released an updated deep learning–based image reconstruction algorithm (Deep Resolve Boost; Siemens Healthineers) that has been made available for 1.5 T and 3 T MRI machines. When available for the MAGNETOM Free.Max, we will be performing a second analysis using this advanced reconstruction algorithm. Fourth, deep learning–enhanced image reconstruction is not available on the MAGNETOM Prisma. 3 T systems including this reconstruction technique will allow for faster image acquisition.

In conclusion, new-generation 0.55 T knee MRI with deep learning–aided image reconstruction allows for reliable detection and grading of joint lesions in symptomatic patients, but it showed limited accuracy and reader confidence for low-grade cartilage and meniscal lesions in comparison with 3 T MRI.

REFERENCES