Seven-Tesla Magnetic Resonance Imaging Has Increased Sensitivity but Decreased Specificity in Diagnosing Cartilage Defects in the Knee Compared With 1.5- and 3-Tesla Magnetic Resonance Imaging

Purpose

To assess the accuracy of 7-Tesla (T) magnetic resonance imaging (MRI) compared with standard-of-care (SOC) 1.5- or 3-T MRI in detecting and grading cartilage defects in the knee.

Methods

Participants who already underwent SOC MRI at 1.5- or 3-T and provided consent were scheduled for 7-T MRI prior to arthroscopy for all indications except revision surgery. SOC and 7-T MRI scans were independently reviewed in randomized order by 2 blinded musculoskeletal radiologists using a modified Outerbridge classification system for defect grading. Image quality (sharpness, contrast, artifact, noise) was also rated. At the time of arthroscopy, each articular surface was then graded by the surgeon, who was blinded to the 7-T MRI scans. Using arthroscopy as the gold standard, we calculated the sensitivity and specificity of SOC MRI and 7-T MRI, as well as inter-rater reliability.

Results

A total of 81 patients—41 female patients (aged 45 ± 14 years) and 40 male patients (aged 40 ± 14 years)—were enrolled. The average time between SOC and 7-T MRI was 69 days, and the average time between 7-T MRI and surgery was 6 days. Sharpness, contrast, and noise ratings were all significantly improved with 7-T MRI compared with SOC MRI ( P <.05). Seven-Tesla MRI had higher sensitivity but lower specificity in detecting cartilage defects for all 6 articular surfaces (patella, trochlea, lateral femoral condyle, medial femoral condyle, lateral tibial plateau, and medial tibial plateau) when using arthroscopy as the gold standard. Intraoperative cartilage grading was significantly closer to 7-T grading compared with SOC grading for 2 of 6 surfaces ( P <.05). Seven-Tesla MRI had reduced coefficient-of-variation values for 1 articular surface (medial femoral condyle) compared with 3-T MRI ( P =.036). There were no significant differences in other coefficient-of-variation measures or interobserver reliability with 7-T MRI compared with SOC MRI.

Conclusions

Seven-Tesla MRI showed improved imaging quality metrics and increased sensitivity but decreased specificity in diagnosing cartilage defects in the knee compared with SOC 1.5- and 3-T MRI using arthroscopy as the gold standard.

Level of Evidence

Level II, prospective comparative diagnostic accuracy study.

Osteoarthritis is a leading cause of disability, affecting more than 240 million persons worldwide. ^, As the population ages and obesity rates continue to rise, the prevalence and global burden of osteoarthritis is expected to increase. Therefore, developing disease-modifying interventions to slow and potentially reverse the progression of osteoarthritis is a public health priority. As cartilage injury plays a critical role in the pathogenesis of the disease, preserving and regenerating cartilage is a specific area of active interest. ^,

Cartilage injury can be difficult to diagnose clinically, but early diagnosis is crucial as the disease may be more responsive to treatment at an early stage. Although cartilage defects may be visualized using magnetic resonance imaging (MRI), the sensitivity of current commercial MRI systems is lower than that of direct operative evaluation. ^, As a result, arthroscopy remains the gold standard, and there remains a critical unmet need for improved noninvasive cartilage imaging techniques.

MRI works through the alignment of hydrogen nuclei by a magnetic field; accordingly, the field strength plays an integral role in the signal-to-noise ratio (SNR) and quality of the image created. ^,^, Standard-of-care (SOC) MRI is most often performed using a 1.5- or 3-T machine. Ultra–high-field 7-T MRI is a newer technology that potentially offers a 2.3-fold improvement in SNR compared with 3-T MRI, and it has been predominantly studied in the setting of neuroimaging. ^, As the technology is still in the relatively early stages, research on musculoskeletal imaging applications has been limited, largely owing to lack of dedicated surface coils. Although previous studies have found improved SNR with 7-T MRI of the knee, they have not evaluated its accuracy in detecting cartilage defects in the knee against an arthroscopic gold standard. Improved image quality and resolution with 7-T MRI could allow for better morphologic detection of cartilage defects, which in turn could allow for better monitoring of disease progression and disease-modifying treatment effects from both a clinical and research standpoint.

The purpose of this study was to assess the accuracy of 7-T MRI compared with SOC 1.5- or 3-T MRI in detecting and grading cartilage defects in the knee. We hypothesized that 7-T MRI would result in improved image quality characteristics, improved sensitivity, and improved specificity in detecting and grading cartilage injury.

Methods

In this prospective, paired, blinded study, participants scheduled for knee arthroscopy were enrolled after providing informed consent. The study was approved by our institutional review board. Participants had already undergone an MRI scan for clinical purposes on either a 1.5- or 3-T machine (SOC) and were scheduled to undergo a study-intervention MRI on a 7-T machine at their convenience prior to their planned surgical date. All indications for knee arthroscopy were included except revision surgery. Additional exclusion criteria were weight less than 30 kg, pregnancy, and contraindications to MRI (e.g., pacemaker). On the basis of a power of 0.80 at α =.05 with a minimum detectable difference of 15% in defect detection frequency, sensitivity, and specificity between 7-T and SOC MRI measures, a minimum sample size of 80 patients was required.

MRI Protocol

The 7-T examination consisted of 3-plane (coronal, sagittal, and axial) 2-dimensional fluid-sensitive sequences with or without fat suppression, a coronal T1-weighted sequence, and a 3-dimensional fluid-sensitive fat-suppressed sequence (Sampling Perfection with Application optimized Contrast using different flip angle Evolution [SPACE]; Siemens Healthineers, Erlangen, Germany). Per standard clinical practice, patients underwent SOC 1.5-T MRI (n = 21) or 3-T MRI (n = 60) at various locations of their choosing.

Image Interpretation

Both SOC and 7-T MRI scans were independently reviewed on a picture archiving and communication system workstation (Centricity; GE Healthcare, Waukesha, WI) by 2 fellowship-trained musculoskeletal radiologists. Inconsistencies were reviewed by an additional musculoskeletal radiologist. All were blinded to clinical and arthroscopic data. Scans were reviewed over a 5-day period, with a 4-week washout period between review of SOC and 7-T scans to limit recall bias. Each radiologist graded all 6 articular surfaces of the knee (patella, trochlea, lateral femoral condyle, medial femoral condyle, lateral tibial plateau, and medial tibial plateau) using a modified Outerbridge classification system (grade 0, normal cartilage; grade 1, increased T2 signal of morphologically normal cartilage; grade 2, low-grade partial thickness involving <50%; grade 3, high-grade partial thickness involving >50%; and grade 4, exposed subchondral bone). For surfaces with multiple defects, the highest-grade defect was assigned. The radiologists rated their perceived confidence in detecting and grading cartilage defects on a scale of 1 to 5, with 5 being the highest confidence level. To assess image quality, the radiologists graded sharpness based on the visibility of femoral nerve fascicles (5, all individual nerve fascicles visible; 4, some fascicles visible but not all [approximately 75%]; 3, approximately 50% of fascicles visible; 2, 25% of fascicles visible; 1, no fascicular architecture visible). Contrast between fluid and cartilage was graded on a scale of 1 to 5 using an axial cut at the mid patella. Finally, perceived artifact and noise were graded on a scale of 1 to 5, with 5 being the highest image quality.

Arthroscopy

At the time of arthroscopy, each articular surface was graded by the surgeon (P.C.M.) using the same classification system described earlier, with the exception that grade 1 indicated softening of articular cartilage. A condylar map was completed, depicting the location and grade of articular cartilage defects, as well as the presence of any other intra-articular pathology (e.g., meniscal tears and loose bodies) ( Fig 1 ). The surgeon was blinded to 7-T images but had access to SOC MRI scans, which was necessary for patient care.

Image, AltText currently not available — Fig 1

Statistical Analysis

Ratings of image quality and confidence between 7-T and SOC MRI were compared using the Mann-Whitney U test for nonparametric data. Comparisons between 7-T and SOC MRI were performed for the entire sample; then, sub-comparisons of 7-T MRI versus 1.5-T MRI and 3-T MRI were separately performed. Next, using arthroscopy as the gold standard, we calculated the sensitivity and specificity of SOC MRI and 7-T MRI for quantifying true-positive or true-negative findings (the radiologist identified the presence or absence of a defect that was confirmed intraoperatively by the surgeon) and false-positive or false-negative findings (the radiologist’s determination differed from the surgeons’ intraoperative observation). This was performed on a per–articular surface basis, with 6 articular surfaces per patient after correction for within-subject clustered data. Then, χ ² analysis was used to compare sensitivity and specificity between SOC MRI, 7-T MRI, and intraoperative assessment. Next, the Friedman test for nonparametric data was used to compare defect severity ratings between SOC MRI, 7-T MRI, and intraoperative assessment. Finally, to assess variability and repeatability in the radiologists’ assessments of defect presence, coefficients of variation (CVs) were calculated and compared across all patients using a paired t test, in which lower CVs indicate less variability between radiologists’ assessments. The type I error rate for all statistical analyses was set at α =.05.

Results

Eighty-one patients were enrolled (October 2019 to August 2021) and underwent 7-T MRI scans prior to arthroscopy. There were 41 female patients (aged 45 ± 14 years) and 40 male patients (aged 40 ± 14 years); no significant differences in results were found between sexes. The age range was 18 to 69 years. The average body mass index was 26 ± 4, with a range of 19 to 37. The most common surgical procedures performed were meniscus repair (n = 30, 37%), anterior cruciate ligament reconstruction (n = 25, 31%), and meniscectomy (n = 5, 6%). For sub-comparisons between 7-T MRI and either 1.5- or 3-T MRI, 21 patients received 1.5-T MRI scans and 60 patients received 3-T MRI scans. The average time between SOC MRI and 7-T MRI was 69 days, and the average time between 7-T MRI and surgery was 6 days.

Radiologist Ratings of Image Quality

Sharpness, contrast, and noise—each rated on a scale of 1 to 5, with 5 being superior image quality—were all significantly improved with 7-T MRI compared with SOC MRI ( P <.001, P <.001, and P =.005, respectively) ( Fig 2 ). The radiologists’ reported confidence in detecting and grading cartilage defects, rated on a scale of 1 to 5, was significantly improved with 7-T MRI compared with 1.5-T MRI ( P =.005) but not when compared with 3-T MRI ( P =.279).

Sensitivity, Specificity, and Repeatability

Compared with SOC MRI (3-T and 1.5-T combined), 7-T MRI had a higher sensitivity in detecting cartilage defects for lateral femoral condyle when using arthroscopy as the gold standard ( Table 1 ). Although no other statistically significant differences were observed between groups, on average, 7-T MRI had a higher sensitivity but lower specificity in the rest of the articular surfaces. Seven-Tesla MRI had reduced CV values for only 1 articular surface (medial femoral condyle) compared with 3-T MRI ( P =.036) ( Table 2 ). Aside from this finding, there were no significant differences in CV measures or interobserver reliability with 7-T MRI compared with SOC MRI.

Table 1

Sensitivity and Specificity

	Overall						7-T vs 1.5-T				7-T vs 3-T
	Sensitivity, %			Specificity, %			Sensitivity, %		Specificity, %		Sensitivity, %		Specificity, %
	SOC	7-T	P -Value	SOC	7-T	P -Value	SOC	7-T	SOC	7-T	SOC	7-T	SOC	7-T
PAT	70.8	72.9	.410	33.3	27.3	.296	69.2	69.2	37.5	50.0	71.4	74.3	32.0	20.0
TRO	41.9	58.1	.102	62.0	52.0	.156	30.0	30.0	63.6	54.5	47.6	71.4	61.5	51.3
MTP	35.3	52.9	.071	70.2	68.1	.412	22.2	44.4	83.3	83.3	40.0	56.0	65.7	62.9
LTP	46.7	57.8	.146	52.8	38.9	.118	50.0	40.0	63.6	45.5	45.7	62.9	48.0	36.0
MFC	50.0	62.0	.113	51.6	38.7	.154	33.3	50.0	66.7	44.4	55.3	65.8	45.5	36.4
LFC	42.9	67.9	.030	60.4	50.9	.164	20.0	20.0	75.0	43.8	47.8	78.3	54.1	54.1