Clinical utility of fat-suppressed 3-dimensional controlled aliasing in parallel imaging results in higher acceleration sampling perfection with application optimized contrast using different flip angle evolutions MRI of the knee in adults

Abstract

Objective:

To compare htree-dimensional CAIPIRINHA SPACE and two-dimensional turbo spin echo (2D TSE) MRI in the diagnosis of knee pathology in symptomatic adult patients.

Methods:

From February to September in 2018, 120 patients who underwent a knee MRI using both 3D CAIPIRINHA SPACE and 2D TSE MRI were enrolled. The signal-to-noise ratios (SNRs) and contrast-to-noise ratio (CNR) of the 2D and 3D MRI were compared using a paired t-test. Two radiologists independently evaluated both 2D and 3D MRI images using scoring systems for the menisci, ligaments, and cartilage. Intermethod, inter- and intrareader agreements were determined using an intraclass correlation coefficient (ICC). The diagnostic performance of both methods was measured in 44 patients with arthroscopy.

Results:

The mean scan time of 3D CAIPIRINHA SPACE MRI (4’ 43”) was shorter than that of 2D TSE MRI (17’ 27”). The mean SNR and CNR of 3D CAIPIRINHA SPACE was higher than those of 2D TSE MRI (mean difference, 3.97 of SNR and 1.58 of CNR; p < 0.001 and p = .038, respectively). Intermethod (ICC, 0.84–1.0) and inter-reader (ICC, 0.75–0.97), and intra-reader agreements (ICC, 0.87–1.0) were good or excellent. The diagnostic accuracy of 3D CAIPIRINHA SPACE sequence was equal for ligament (95.5%) and better for meniscal and cartilage evaluation (84.1% each), compared to 2D TSE MRI (79.5% each).

Conclusion:

The fat-suppressed 3D CAIPIRINHA SPACE MRI maybe useful in clinical practice for the evaluation of the knee in place of the 2D conventional MRI protocol.

Advances in knowledge:

1. The 3D CAIPIRINHA SPACE MRI of the knee joint may be acceptable to be used in clinical practice showing comparable imaging quality compared to conventional 2D TSE MRI.

2. Compared with arthroscopic findings as the gold-standard, the diagnostic performance of 3D CAIPIRINHA SPACE MRI was equal or better for knee joint evaluation than that of 2D TSE MRI, as well as with shorter scan time.

Introduction

MRI of the knee plays a substantial role in morphologic assessment of anatomic structures and their pathologic changes and is widely performed with two-dimensional (2D) fast or turbo spin echo (TSE) sequences. However, conventional 2D MRI has some limitations, as separate multiorthogonal scanning is time consuming. Moreover, musculoskeletal imaging may require particularly additional curved and oblique planar images due to the non-rectilinear anatomic course of ligaments, tendons, and cartilage.¹

Three-dimensional (3D) isotropic volumetric acquisition techniques, such as sampling perfection with application optimized contrast using different flip angle evolutions (SPACE), produce the signal through volume excitation and, therefore, offer a small isotropic voxel size for multi planar reconstruction from single parent raw images.² Unfortunately, 3D TSE MRI has other limitations, including reduced spatial resolution, image blurring, and long acquisition time.^3–6 A recently introduced technique, controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA), allows for the optimized use of coil sensitivity and parallel imaging acceleration in two dimensions.This enables a reduction in aliasing artifacts, degrades image noise, and reduces data acquisition time.^7–9 Therefore, 3D CAIPIRINHA SPACE is thought to be a promising technique for isotropic, and high spatial resolution 3D MRI in musculoskeletal imaging.^2,8–11

Fat suppression techniques contribute diagnostic confidence in musculoskeletal MRI. It specifically improves conspicuity of bone marrow lesions, including edema, and also improves the differentiation of soft tissue structures from the surrounding fat.^12,13 Furthermore, it is useful in cartilage imaging, as it reduces chemical shift artifacts.¹⁴ There is some controversy, however, as to whether fat suppression improves the diagnostic performance of knee MRIs.^15–17

There have been a few studies investigating image quality and diagnostic confidence of knee disease in adults using 3D CAIPIRINHA SPACE MRI.^10,11 To the best of our knowledge, however, no study has examined diagnostic performance using arthroscopic findings as the gold-standard in symptomatic adult patients.Therefore, we aimed to evaluate the clinical utility of fat-suppressed 3D CAIPIRINHA SPACE MRI compared to 2D TSE MRI of the knee in patients.

Methods and materials

This retrospective study was approved by our hospital’s (Chung-Ang University Hospital, Seoul, Korea) institutional review board and the requirement for informed consent was waived.

Study population

Between February and September in 2018, a total of 144 consecutive symptomatic patients underwent a knee MRI using both 2D conventional and 3D CAIPIRINHA methods via our institution’s orthopedic clinic. Among them, three patients underwent MRI exams for both knees. The inclusion criteria were as follows: adult >18 years of age and no previous history of knee surgery. The exclusion criterion was suboptimal images that could not be evaluated or for which 3D CAIPIRINHA method or multiplanar reconstruction (MPR) was omitted. Based on these criteria, 24 patients were excluded from our study: suboptimal images in 16 patients, omitted MPR in 8 patients. Ultimately, 120 patients were enrolled. There were 52 males and 68 females with a mean age of 49.7 years (range of age, 18–88 years: standard deviation (SD), 17.58 years). Also there were 58 right knees and 62 left knees in our study. Among the 120 patients, 44 underwent arthroscopy after their MRI exams.

MRI techniques

All MRI exams were performed using commercially available 3.0 T MRI scanners (MAGENETOM Skyra, Siemens Healthcare, Germany) and 15-channel knee surface coils from the same vendorwith straight position. The fat-suppressed 3D T₂ weighted CAIPIRINHA SPACE images were scanned through the sagittal plane, and axial and coronal images were reconstructed from raw images within 10 s. The detailed parameters of 2D TSE and 3D CAIPIRINHA SPACE MRI are listed in Table 1.

Table 1.

MR pulse sequence protocol

Parameters	PD FS axial	PD coronal	T2W sagittal	PD sagittal	PD oblique coronal	3D CAIPIRINHA
						T2W sagittal
TR (ms)	5500	3180	5040	4990	2880	1200
TE (ms)	27	35	81	35	58	108
Flip angle	150	150	150	150	150	120
Bandwidth	225	530	220	220	220	225
FOV (mm)	160 × 160	150 × 150	150 × 150	150 × 150	160 × 160	160 × 160
Matrix	384 × 384	448 × 448	320 × 320	320 × 320	384 × 384	256 × 256
Slice thickness (mm)	2.5	2.5	1.5	1.5	1.5	0.6
Slice gap (mm)	0.5	0.5	0.5	0.5	0.5	0
Slice numbers	45	55	55	55	55	204
Acceleration factor	-	-	-	-	-	4
Scan time	3 min 40 s	3 min 33 s	3 min 20 s	3 min 32 s	3 min 22 s	4 min 33 s

CAIPIRINHA, controlled aliasing in parallel imaging results in higher acceleration; FOV, field of view; FS, fat-saturated; PD, proton density; TE, echo time; TR, repetition time; T₂ W, T₂ Weighted.

Image quality analysis

One radiologist (SL) measured the signal-to-noise ratio (SNR) of 3D CAIPIRINHA SPACE and 2D TSE sequences in a total of 120 fat-suppressed MRI axial images, as described by Wolff et al.^18,19 Axial images of 2D TSE MRI were used for SNR measurement to compare 3D fat-suppressed axial images, because which was only fat-suppressed scan in the TSE protocol. One patient was excluded due to omission of the 2D fat-suppressed MRI. The region of interest (ROI) was drawn manually at the cancellous bone marrow space of the proximal tibia or distal femur (S_A), avoiding pathologic areas for measurement of signal intensities. Noise was measured as standard deviation in air and not on an anatomic structure (S_N). This was done more than three times and an average value was used.

$$\mathrm{S}\mathrm{N}\mathrm{R}=\mathrm{}\frac{S\text{A}}{S\text{N}}$$

However, this conventional measurement of SNR could not be applied for parallel imaging not following Rician noise distribution.²⁰ The SNR of parallel imaging is affected by acceleration value (R-value) due to the use of phased array coil and the acquisition of fewer phase encoding time, and geometry factor (g-factor) based on data processing of reducing aliasing artifact.^20,21 Plus, the SNR of CAIPIRINHA image is improved with the number of simultaneous excited slices (NS factor) in compared with sequential excitation.²² It is necessary to take a pre-scan noise measurement, correlation in the phased-array receiver and single accelerated image acquisition for measurement of g-factor.²¹ Unfortunately, this study could not take this method because of its retrospective design. Despite it was not accurate measurement of the SNR, subjective comparison of SNR values between 2D and 3D MRI images was chosen in limited setting of retrospective study. In this study, R-value was four and shift factor was 1.

$${\displaystyle SNR(SENSE)=\frac{SNR}{g\cdot \sqrt{R}}}$$ (23)

In addition, contrast-to-noise ratio (CNR) in 3D CAIPIRINHA SPACE and 2D TSE fat-suppressed axial images was determined. CNR was defined as subtraction of cartilage signal and ligament signal divided by standard deviation of air. One radiologist (SL) manually drew ROIs in the posterior femoral condyle cartilage and posterior cruciate ligament avoiding pathologic signal at same section of axial image. These anatomic regions were selected, considered as the most reliable and easy to draw in the axial plane.

$$\mathrm{C}\mathrm{N}\mathrm{R}=\mathrm{}\frac{S\text{Cartilage}-S\text{Ligament}}{S\text{N}}$$

MR imaging finding analysis

Two radiologists (SL with 4 experience years and GYL with 14 experience years in musculoskeletal radiology) independently evaluated the menisci, ligaments, and cartilage on the knee MRI using 2D TSE and 3D CAIPIRINHA SPACE. Three-point scales for the menisci and ligaments and 4-point scales of the scoring system of the international cartilage repair society²³ for the cartilage were used, as listed in Table 2. Both medial and lateral menisci were divided into three anatomic portions for scoring: the anterior horn, body, and posterior horn. The anterior cruciate ligament, posterior cruciate ligament, medial collateral ligament, and lateral collateral ligament were evaluated with this scale. Cartilaginous lesions were estimated according to the anterior, middle, and posterior portions of the medial and lateral femoral condyles, medial and lateral tibial plateaus, medial and lateral trochleae, and medial and lateral facets of the patellae. If there were more than one chondral lesion in the same anatomic region, the most severe lesion according to scoring system was considered.

Table 2.

Scoring systems for abnormalities of the menisci, the ligaments, and the cartilage

Meniscus
Grade 0	Normal
Grade 1	No definite tear but focal edema
Grade 2	Tear with extension to one surface
Grade 3	Tear with penetrating high signal
Ligament
Grade 0	Normal
Grade 1	Edema less than 50% thickness of the ligament
Grade 2	Edema more than 50% thickness of the ligament
Grade 3	Evident discontinuity of the ligament
Cartilage
Grade 0	Normal
Grade 1	Superficial lesion, crack and indentation
Grade 2	Fraying, lesions extending less than 50% thickness of the cartilage
Grade 3	Partial loss of cartilage thickness of more than 50%
Grade 4	Complete loss of the cartilage thickness

After the first review of the MRI images, the two readers had a consensus meeting for inconsistent findings and re-scored any inconsistent cases (17 cases for the menisci, 11 for the ligaments, and 21 for the cartilage). 1 month later, a second review was done independently in the same manner.

In the 44 patients who underwent arthroscopic surgery, diagnostic accuracy was determined based on the arthroscopic findings. For statistical analysis, ratings of a Grade 3 meniscal tear, Grade 3 complete ligament tear, and a Grade 4 chondral lesion by two readers were considered to be positive.

Arthroscopic surgery

One of two orthopedic surgeons (YBP with 14 experience years and HJL with 24 experience years in knee arthroscopy) performed all arthroscopic surgeries of 44 patients and wrote about the operative findings on electric medial record (EMR) of our hospital. The mean interval between MRI exam and surgery was 21.8 days (range, 0–151 days; SD, 37.7 days).

Statistical analysis

A paired t-test was used to compare the SNR and CNR values of the 2D TSE and 3D CAIPIRINHA MRI. Intermethod, inter-reader and intrareader reliabilities of each sequence were analyzed with an intraclass correlation coefficient (ICC) for detection of meniscal or ligament tear and cartilage abnormality. The ICC was interpreted as poor (less than 0.5), moderate (0.5–0.75), good (0.75–0.9), excellent (more than 0.9).²⁴ The sensitivity, specificity, accuracy, positive- and negative predictive values of 2D TSE and 3D CAIPIRINHA images were determined about diagnosis of meniscal or ligament tear and cartilage lesions based on arthroscopic findings. Statistical analysis was done with SPSS statistical software (Version 25, IBM Inc., NY, USA). A p-value of less than 0.05 was considered to be statistically significant.

Results

In imaging quality analysis, the mean ROI size was 10.42 mm² (SD, 3.55 mm²) in 3D CAIPIRINHA and 2.18 mm² (SD, 0.86 mm²) in 2D TSE MRI, respectively. The average SNR values were 16.23 (SD, 5.16) in fat-suppressed 3D CAIPIRINHA images and 12.25 (SD, 3.06) in fat-suppressed 2D TSE MRI. The mean SNR of 3D CAIPIRINHA was significantly higher than that of 2D TSE MRI (mean difference (MD), 3.97; 95% confidence interval (CI), 2.36–5.59; p < 0.001). The average CNR values were 14.02 (SD, 5.31) in 3D CAIPIRINHA images and 12.69 (SD, 3.61) in 2D TSE. The mean CNR of 3D CAIPIRINHA was also significantly higher than that of 2D TSE (MD, 1.58; 95% CI 0.092–3.06; p = .038).

The intermethod agreements between the 3D CAIPIRINHA and 2D TSE MRI results were good or excellent in all anatomic regions by both Reader 1 (ICC, 0.84–0.99) and Reader 2 (ICC, 0.89–1.0). Intrareader agreements by Reader 1 were good or excellent on both the 2D TSE (ICC, 0.91–0.99) and 3D CAIPIRINHA images (ICC, 0.91–0.99), as were those by Reader 2 on both the 2D TSE (ICC, 0.87–1.0) and 3D CAIPIRINHA sequences (ICC, 0.92–1.0). The inter-reader agreements were also good or excellent on both the 2D TSE (ICC, 0.75–0.97) and 3D CAIPIRINHA MRI (ICC, 0.76–0.96), as shown in Table 3.

Table 3.

Intermethod, inter- and intrareader agreements between 2D conventional and 3D CAIPIRINHA MRI

Region	Intermethod agreement		Inter-reader agreement		Intrareader agreement
	R1	R2	3D	2D	R1		R2
					3D	2D	3D	2D
MM, AH	0.90	0.99	0.80	0.80	0.92	0.91	1	0.98
MM, body	0.96	0.97	0.91	0.92	0.96	0.93	1	0.99
MM, PH	0.98	0.98	0.95	0.95	0.98	0.98	1	1
LM, AH	0.97	0.98	0.93	0.94	0.94	0.96	0.99	0.99
LM, body	0.92	0.96	0.91	0.94	0.99	0.98	0.98	0.99
LM, PH	0.92	0.99	0.9	0.91	0.94	0.94	1	0.99
ACL	0.94	0.91	0.89	0.92	0.95	0.95	0.99	0.99
PCL	0.84	0.89	0.76	0.82	0.93	0.91	0.92	0.87
MCL	0.95	0.94	0.93	0.9	0.95	0.95	0.98	0.99
LCL	0.95	0.92	0.79	0.75	0.95	0.93	0.99	0.97
Anterior MFC	0.95	0.98	0.84	0.89	0.97	0.98	0.97	0.96
Central MFC	0.99	1	0.96	0.97	0.98	0.97	1	1
Posterior MFC	0.99	0.99	0.95	0.95	0.95	0.95	1	0.99
Anterior LFC	0.92	0.99	0.79	0.78	0.91	0.93	0.99	1
Central LFC	0.94	0.96	0.87	0.89	0.92	0.94	0.99	1
Posterior LFC	0.94	0.97	0.92	0.93	0.95	0.93	1	1
MTP	0.98	0.99	0.92	0.93	0.97	0.96	0.99	1
LTP	0.98	0.98	0.96	0.93	0.98	0.98	0.98	0.97
Medial P	0.98	0.99	0.94	0.93	0.97	0.98	1	0.99
Lateral P	0.99	0.99	0.95	0.96	0.98	0.99	0.98	0.98
Medial T	0.94	0.95	0.87	0.86	0.97	0.96	0.99	0.99
Lateral T	0.99	1	0.92	0.91	0.98	0.98	1	0.99

ACL, anterior cruciate ligament; AH, anterior horn; CAIPIRINHA, controlled aliasing in parallel imaging results in higher acceleration; ICC, intraclass correlation coefficient; LCL, lateral collateral ligament; LFC, lateral femoral condyle; LM, lateral meniscus; LTP, lateral tibial plateau; MCL, medial collateral ligament; MFC, medial femoral condyle; MM, medial meniscus; MTP, medial tibial plateau; P, patella; PCL, posterior cruciate ligament; PH, posterior horn; R1, reader 1; R2, reader 2; T, trochlea; TSE, turbo spin echo.

^aData presented as intra class correlation coefficient (ICC) values.

MRI findings were correlated with arthroscopic results in the 44 patients who underwent arthroscopic surgery, summarized in Table 4. Conventional 2D TSE MRI revealed 26 true meniscal tears and 9 true-negative meniscal lesions (83.9% sensitivity, 69.2% specificity, and 79.5% accuracy). On the contrary, 28 true meniscal tears were identified on fat-suppressed 3D CAIPIRINHA MRI, including 2 that were missed with 2D TSE MRI (one of them in Figure 1), providing better diagnostic performance in the evaluation of menisci (90.3% sensitivity, 69.2% specificity, and 84.1% accuracy) than 2D TSE MRI. In evaluation of the ligaments, 6 true complete ligament tears and 36 true-negative ligamentous lesions were demonstrated both on 2D TSE and 3D CAIPIRINHA MRI (100% sensitivity, 94.7% specificity, and 95.5% accuracy) (Figure 2). In the evaluation of chondral lesions, 20 true chondral defects and 15 true-negative findings were foundon 2D TSE MRI (80.0% sensitivity, 78.9% specificity, and 79.5% accuracy). Otherwise, 3D CAIPIRINHA MRI showed additional two true chondral lesions (88.0% sensitivity, 78.9% specificity, and 84.1% of accuracy) (Figure 3).

Table 4.

Diagnostic performances of knee pathology in 2D TSE and 3D CAIPIRINHA MRI

Parameters	Meniscal tear		Ligament tear		Cartilage abnormality
	2D	3D	2D	3D	2D	3D
Sensitivity	83.9% (26/31)	90.3% (28/31)	100% (6/6)	100% (6/6)	80.0% (20/25)	88.0% (22/25)
Specificity	69.2% (9/13)	69.2% (9/13)	94.7% (36/38)	94.7% (36/38)	78.9% (15/19)	78.9% (15/19)
Accuracy	79.5% (35/44)	84.1% (37/44)	95.5% (42/44)	95.5% (42/44)	79.5% (35/44)	84.1% (37/44)
PPV	86.7% (26/30)	87.5% (28/32)	75.0% (6/8)	75.0% (6/8)	83.3% (20/24)	84.6% (22/26)
NPV	64.3% (9/14)	75.0% (9/12)	100% (36/36)	100% (36/36)	75.0% (15/20)	83.3% (15/18)

NPV, negative predictive value; PPV, positive predictive value.

^aRow data is presented in parentheses.

Figure 1.

A 45-year-old male with a twisting injury in the left knee shows a subtle surface irregularity of lateral meniscus on the 2D proton-density sagittal scan (A, arrow). However, a focal undersurface defect of lateral meniscus is more clearly identified on the fat-suppressed 3D CAIPIRIHA sagittal image (B, arrow). Additionally, a focal radial defect of lateral meniscus is well depicted on the fat-suppressed 3D CAIPIRINHA axial image (C, arrow), which is confirmed on arthroscopy (D). 2D, two-dimensional; 3D, three-dimensional.

Figure 2.

The sagittal scan of the 2D TSE MRI reveals the wavy contour and increased signal intensity of posterior cruciate ligament suspected of partial thickness tear (A, arrow) in a 47-year-old-male after fall-down injury. On the contrary, more prominent torn appearance of posterior cruciate ligament is depicted on the fat-suppressed 3D CAIPIRINHA sagittal view (B, arrow), confirmed as complete rupture of posterior cruciate ligament on arthroscopy. 2D, two-dimensional; 3D, three-dimensional; TSE, turbo spin echo.

Figure 3.

The coronal scan from the 2D TSE MRI of a 49-year-old male scheduled for a right high tibial osteotomy due to coxa vara reveals a mild cartilage edema of the bilateral tibial plateaus and subtle cartilage irregularity in the medial tibial plateau (A, arrow), whereas the fat-suppressed 3D CAIPIRINHA SPACE MRI shows a definite chondral depression of the medial tibial plateau (B, arrow). 2D, two-dimensional; 3D, three-dimensional; TSE, turbo spin echo.

Discussion

The 3D TSE MRI has often been considered as an appealing future technique in musculoskeletal imaging. However, its clinical application is limited out of its relatively long acquisition time and low image quality.^3–6,11,25 Various attempts to reduce acquisition time, such as large voxel sizes, long echo trains, one-dimensional parallel imaging, and partial Fourier under sampling, resulted in image blurring and aliasing artifacts.^4,26,27 The 3D CAIPIRINHA SPACE MRI optimizes the maximum sensitivity of the coil and allows a twofold acceleration in in-plane and through-plane phase-encoding directions.^7,9,11 This technique uses true isovoxel size, limited echo train length, and full Fourier sampling, which leads to the preservation of image quality and the uniformity of reconstructions.^2,10,11,28 With the combined use of parallel compressed sensing, under sampling of k-space with iterative reconstruction can achieve a sixfold acceleration, however, it is not currently used due to longer image reconstruction times.^29,30 The faster acquisition time and immediate multiplanar reconstruction of 3D CAIPIRINHA SPACE MRI could improve clinical efficiency and accessibility. The 3D CAIPIRINHA SPACE MRI protocols in the 1.5- and 3.0 T machines received US Food and Drug Administration clearance in February 2018.

Only a few validation studies of the clinical utility and diagnostic performance of 3D CAIPIRINHA SPACE MRI of the knee have been done with a scan time of a few minutes.^10,11,28 The fat suppression technique was first added in knee MRI studies using 3D CAIPIRINHA SPACE to improve tissue contrast and enhance the clarity of pathologic lesions.^13,15,16 The 3D CAIPIRINHA MRI protocol in our study takes less than 5 min, possibly because of using less matrix size and slice number, which is a half of scan time as that of previous studies.^10,28 Previously Fritz et al¹¹ reported knee MRI scan of 5 min, however, which was a study for asymptomatic subjects. In addition, to the best of our knowledge, our study is the first to perform arthroscopic validation and estimate diagnostic performance of 3D CAIPIRINHA SPACE MRI of the knee in adults.

In our study, the intermethod agreements between 3D CAIPIRINHA SPACE and 2D TSE MRI were good or excellent. Despite different statistical methods, these results followed similar tendency with those of a previous study by Del Grande, et al,¹⁰ which showed moderate to very good intermethod concordance (κ value, 0.59–1.0). Meanwhile, their study suggested only moderate intermethod concordance in the evaluation of cartilage defects with both 1.5 and 3.0 T, that might be due to higher spatial resolution, minimization of partial volume artifacts, and ultimately, diagnostic superiority in 3D MRI for cartilage defects.¹⁰ In our study, the posterior cruciate ligament demonstrated relatively lower ICC values by both readers. It is thought that the posterior cruciate ligament sometimes exhibits an exaggerated high signal on fat-suppressed 3D CAIPIRINHA MRI, possibly owing to emphasized signals of tiny fluid drops or vessel structures.

The inter- and intrareader reliabilities of assessments of knee pathology were also good or excellent on both MR sequences. The structures or regions in which the ICC values were less than 0.8 may be concerned about accurate diagnosis on 3D CAIPIRINHA MRI such as posterior cruciate ligament in our study. However, there have been no previous reports of interpretation reliability in detailed anatomic regions of the knee to be compare with our results.

The results of the fat-suppressed 3D CAIPIRINHA SPACE MRI suggested a slightly better diagnostic performance in meniscal evaluation than the 2D TSE MRI. Previous 3D MRI studies of meniscal tears in adults reported a wide range of sensitivities, specificities, and accuracies of 58–100%, 68–97% and 74–97%, respectively.^{4,16,26,27,31,32} Furthermore, diagnostic performance of cartilage evaluation of 3D CAIPIRINHA SPACE MRI was also slightly superior to that of 2D TSE MRI in our study. In the assessment of cartilage, the direct comparison of our results with those of previous studies may be limited, as the International Cartilage Repair Society (ICRS) grading system was applied with a binary interpretation for correlation with surgical findings, which was widely used in our hospital because of the surgeons’ preference.²³ A meta-analysis by Shakoor et al reported that the pooled sensitivity and specificity of chondral lesions in 3D MRI were 74.8 and 93.3%, respectively, using the Modified Noyes system or Outer bridge classification,³³ which could be like to our results.

In this study, the SNR and CNR were significantly higher in 3D CAIPIRINHA than 2D TSE in the knee MRI. Previous study reported that SNR and CNR values in 2D TSE were higher than 3D CAIPIRINHA,¹¹ in contrast with our results. Direct comparison could be impossible because fat-suppression was applied in 3D CAIPIRINHA sequence as well as g-factor was not measured, unfortunately. Nevertheless, the SNR and CNR values of 3D CAIPIRINHA SPACE MRI would be acceptable in clinical practice compared to those of 2D conventional MRI.

Our study had some limitations. As a retrospective study, arthroscopic recordof EMR was insufficient for correlation with all lesions documented on MRI, and surgeons could not be blinded of the MRI findings prior to surgery. Second, the sample size for arthroscopic correlation was relatively small. In the future, large clinical validation studies are necessary to evaluate the clinical application of 3D CAIPIRINHA SPACE protocols. Third, our institution’s 3D CAIPIRINHA protocol was designed with fat suppression, whereas, only the proton-density axial image was designed with fat suppression for the 2D TSE MRI. There may be some limitations in the comparison between fat saturated images and non-fat saturated images, as this allowed readers to recognize the type of sequences and personal preferences may have interfered with interpretation.

In conclusion, the fat suppressed 3D CAIPIRINHA SPACE MRI may be useful in clinical practice in place of the 2D conventional MRI protocol in evaluation of the knee.