Jessica Lauren Asay1, Krithika Balaji2, Anthony A Gatti1, Arjun D Desai1, Michael Mendoza2, Zimu Huo2, Akshay S Chaudhari1,3, Feliks Kogan1, Peter J Lally4,5, Neal K Bangerter2, and Garry E Gold1
1Radiology, Stanford University, Stanford, CA, United States, 2Bioengineering, Imperial College London, London, United Kingdom, 3Biomedical Data Science, Stanford University, Stanford, CA, United States, 4Brain Sciences, Imperial College London, London, United Kingdom, 5UK Dementia Research Institute Centre for Care Research and Technology, London, United Kingdom
Synopsis
Keywords: Quantitative Imaging, Data Processing, Repeatability, cartilage, knee
Cartilage T2 relaxation times (T2), used to detect early knee osteoarthritis, lack standardization in acquiring and processing data, making comparisons between studies difficult. Standardizing image post-processing could possibly control for biases. Here, we assess qDESS cartilage T2 repeatability across two different sites and 7T scanner vendors with identical automatic segmentation and T2 mapping software. Within-site repeatability was good (ICC≥ 0.75) for most cartilage regions, while cross-vendor repeatability was good for the tibial and femoral posterior cartilage. This preliminary study shows standardizing acquisition and post-processing can lead to repeatable T2 values across different vendors.
Introduction
T2 relaxation time, a surrogate measure of collagen microstructure and hydration levels of articular cartilage, has been used to study early osteoarthritis (OA) changes. Increased T2 has been linked to OA progression1. To date, most T2 measurement protocols have been developed, collected, and validated at 1.5 or 3T2. Limited work has assessed T2 at ultra-high field strengths (7T). Benefits of using an increased field strength include improved the signal-to-noise ratio (SNR), which offers the potential for faster acquisitions and/or higher resolutions. A challenge of using 7T systems is the increase in field inhomogeneities, which may affect quantitative measurements and compromise the reliability of the data. Previous work has shown that repeatability of T2 relaxation measurements collected at 7T and 3T are similar3. However, previous studies at 3T have shown that acquiring T2 relaxation times using different scanners with different vendors produces systematically different T2 values4. Furthermore, it has been shown that inter-rater, or even inter-algorithm segmentation variability produces biases in cartilage segmentation, which may affect calculated T2 values5. This adds additional variability for comparing T2 between sites or vendors. The purpose of our study was to evaluate the repeatability of cartilage T2 relaxation time measurements at 7T from two sites with different scanner vendors, while standardizing for scanning sequence, coil, segmentation, and reconstruction, cartilage segmentation, and T2 computation algorithms.Methods
Five healthy adults (3 males, age range: 20-50 years) without history of lower limb joint trauma were enrolled in this Institutional Review Board-approved study with appropriate informed consent in place. Their right knees were imaged using the following whole-body 7T scanners: a GE950 7T (GE Healthcare, Waukesha, WI, USA) and a 7T MAGNETOM Terra (Siemens Healthineers, Erlangen, Germany). Both scanners used a 28 Channel Transmit/Receive Knee Coil (Quality Electrodynamics, Cleveland, OH USA). qDESS images were acquired to calculate the knee cartilage T2 relaxation times6; sequence parameters and subject scanning conditions were standardized across vendors where possible (Table 1).
At each scanner, subjects were scanned twice. Between scans, the subjects were removed from the scanner and repositioned. Cartilage segmentation was performed using a publicly available deep-learning model provided by DOSMA, an open-source Python framework for musculoskeletal analysis7,8. All segmentations were manually checked, and no corrections were made. Voxelwise cartilage T2 values were determined using a previously validated method based on extended phase graph modeling6. Mean T2 was calculated across the full thickness of the cartilage of six tibial and femoral regions (anterior, central, and posterior regions of the medial and lateral compartments9,10) (Figure 1).
Reliability between-scans (same vendor and scanner) and between-vendors (different scanner) was assessed using generalizability theory11. Generalizability theory is a measurement theory framework that improves robustness of reliability estimates by allowing the calculation of specific reliability coefficients (between-scan or between-vendor) while simultaneously using all available data. Generalizability theory was used to compute relative reliability (intraclass correlation coefficient; ICC) and absolute reliability (standard error of measurement; SEM) for measurements between scans and between vendors. The computed SEM was then used to calculate a minimum detectable change at 90% confidence (MDC90) using the equation: MDC = SEM * sqrt(2) * z-score with a z-score of 1.645 used to calculate the MDC90. ICCs and SEMs were calculated separately for each anatomic region of interest. ICCs ≥ 0.75 were considered to have good to excellent reliability12.Results
T2 relaxation showed better repeatability between scans than between vendors taken from different vendors (Table 2). T2 relaxation times between scans had excellent repeatability (ICCs ≥ 0.75) in most regions (Table 2). Cross-vendors, the medial and lateral posterior cartilage of the tibia and femur had good to excellent repeatability (ICCs ≥ 0.75). The SEMs between repeated scans were relatively small at 1.4 to 2.6ms, yielding minimum detectable change values ranging from ~3-5ms. SEMs between vendors were larger, and thus required between 4-11ms of change to be detectableDiscussion
Between scan repeatability of cartilage T2 values aligned with those previously reported for a single vendor3. Across sites/vendors, the posterior regions of the femoral and tibial cartilage were repeatable. Some limitations of this preliminary study include that subjects were not scanned at the same time of the day due to scanner availability and may not have been isocenter due to scanner bore and coil placement constraints. Suboptimal scanner stability may have contributed to some of the larger differences both at one of the sites and between sites. Despite the study’s limitations, we found T2 to be repeatable between 7T scans and vendors. To our knowledge, this is the first repeatability study across 7T vendors analyzing T2 values in multiple compartments.Conclusion
This preliminary work compared cartilage T2 relaxation times calculated from scans acquired using the qDESS pulse sequences on 7T scanners from different vendors while standardizing the acquisition parameters, coil, automatic segmentation, and reconstruction algorithm across vendors. Results showed excellent between scan repeatability in most regions of tibial and femoral cartilage and good repeatability in the posterior cartilage of the tibia and femur between vendors. Having such repeatable quantitative measures tested on scanners from different vendors is promising for future cross-site T2 comparative studies conducted at 7T.Acknowledgements
This work was supported by the National Institutes of Health (R01EB002524, R01AR077604), GE Healthcare, the Wu Tsai Human Performance Alliance, and a CIHR Postdoctoral Fellowship.
References
1. Prasad AP, Nardo L, Schooler J, et al. T1r and T2 relaxation times predict progression of knee osteoarthritis. Osteoarthr Cartil. 2013; 21(1):69–76.
2. Chaudhari AS, Stevens KJ, Sveinsson B, et al. Combined 5‐minute double‐echo in steady‐state with separated echoes and 2‐minute proton‐density‐weighted 2D FSE sequence for comprehensive whole‐joint knee MRI assessment. Magn Reson Imaging. 2019;49:e183-e194.
3. Asay JL, Gatti AA, Desai AD, et al. Repeatability of Cartilage T2 relaxation times measures at 3T and 7T using quantitative double-echo in steady-state. Proc Intl Soc Mag Reson Med. 2022; 2307.
4. Balamoody S, Williams TG, Wolstenholme C, et al. Magnetic resonance transverse relaxation time T2 of knee cartilage in osteoarthritis at 3-T: a cross-sectional multicentre, multivendor reproducibility study. Skeletal Radiol. 2013;42(4):511-520.
5. Schneider E, Nevitt M, McCulloch C, et al. Equivalence and precision of knee cartilage morphometry between different segmentation teams, cartilage regions, and MR acquisitions. Osteoarthritis and Cartilage. 2012;20(8):869-879.
6. Sveinsson B, Chaudhari AS, Gold GE, Hargreaves BA. A simple analytic method for estimating T2 in the knee from DESS. Magn Reson Imaging. 2017;38:63–70.
7. Desai AD, Barbieri M, Mazzoli V, et al. DOSMA: A deep-learning, open-source framework for musculoskeletal MRI analysis. Proc Intl Soc Mag Reson Med. 2019; 1106(27).
8. Desai AD, Schmidt AM, Rubin EB, et al. SKM-TEA: A Dataset for Accelerated MRI Reconstruction with Dense Image Labels for Quantitative Clinical Evaluation. arxiv :2203.06823 [eess.IV]. https://arxiv.org/pdf/2203.06823.pdf.
9. Monu UD, Jordan CD, Samuelson BL, Hargreaves BA, Gold GE, McWalter EJ. Cluster analysis of quantitative MRI T2 and T1ρ relaxation times of cartilage identifies differences between healthy and ACL-injured individuals at 3T. Osteoarthritis and Cartilage 2017;25(4):513-520.
10. Black MS, Young K, Chaudhari AS, Sveinsson B, Kogan F, Monu U, McWalter E, Levenston M, Gold GE, Hargreaves BA. T2-mapping of Femoral Cartilage 3-months Following ACL Reconstruction Surgery. 2018; Paris, France.
11. Brennan RL. Generalizability theory. New York: Springer; 2001. https://doi.org/10.1007/978-1-4757-3456-0.
12. Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med. 2016;15(2):155-163.