RPS 1310 - Knee joint

Author Block: A-S. Björkman¹, H. Gauffin¹, S. K. Koskinen², A. Persson¹; ¹Linköping/SE, ²Stockholm/SE
Purpose or Learning Objective: To investigate the diagnostic accuracy of dual-energy CT (DECT) for the detection of ACL tears in the acutely and subacutely injured knee with arthroscopy as reference method.
Methods or Background: Patients with suspected ACL injury were imaged with DECT (Somatom Force, Siemens Healthcare, Germany) and 3.0 T MRI (Ingenia, Philips Medical Systems, Best, The Netherlands). Clinically blinded images were independently read by two radiologists. ACL was classified as normal or abnormal. Arthroscopy served as reference standard. Sensitivity and positive predictive value (PPV) were calculated. Also, sensitivity between DECT and MRI was assessed.
Results or Findings: 48 patients (26 M, 22 F, mean age 23 years, range 15-37 years) were imaged a mean 25 days following trauma. Of these, 21 patients underwent arthroscopy with a mean of 195 days after trauma. Arthroscopy revealed 19 ACL tears and 2 ACLs with no tear. The sensitivity was 76.3% (95% CI 66.8-85.9) and 86.8 (95% CI 71.9-95.6) for DECT and MRI, respectively (p=0.223). The positive predictive value (PPV) was 93.5 (95% CI 84.3-98.2) and 91.7 (95% CI 77.5-98.3) for DECT and MRI, respectively.
Conclusion: DECT has lower sensitivity to detect an ACL rupture than MRI but the difference was not statistically significant. The PPV was high in both methods.
Limitations: The age of the subjects (15-40 years) limits the generalizability of the results to older populations. The low number of true negatives did not allow to calculate specificity.
Ethics committee approval: Ethical approval from the regional ethical review board (2016/44-31 and 2017/221-32) and radiation protection committee was obtained.
Funding for this study: This NACOX-cohort study is supported by the Swedish Medical Research Council, the Swedish Research Council for Sport Science, the Medical Research Council of Southeast Sweden and ALF Grants Region Östergötland

Author Block: N. Stranger¹, C. Kaulfersch², G. P. Mattiassich², J. Mandl¹, D. h. m. Szolar¹, P. A. Hausbrandt¹, H. Schöllnast¹; ¹Graz/AT, ²Schladming/AT
Purpose or Learning Objective: To assess the frequency of ramp lesions (RL) and anterolateral ligament (ALL) injuries in MRI of patients with verified anterior cruciate ligament (ACL) tears, and to describe the coexistence of these injuries with further injuries of the knee.
Methods or Background: In this retrospective study, 163 patients with surgical repair of ACL tears were included. MRI scans were reviewed for RL, which were defined according to Greif et al., and ALL injuries. In addition, all coexisting meniscal injuries, injuries of the medial and lateral collateral band and bone marrow oedema were recorded. The correlation of RL and ALL-injuries with coexisting injuries was tested for statistical significance using Pearson's chi-square test. After Bonferroni correction for multiple testing a p<0.003 was defined as statistically significant.
Results or Findings: RL occurred in 52 patients (31.9%). RL were significantly associated with tears of the posterior horn of the medial meniscus (67.3% versus 19.8% of patients with/without RL, respectively) and with subchondral bone marrow oedema in the postero-medial tibia plateau (63.5% versus 0% of patients with/without RL, respectively). ALL-injuries were detected in 52 patients (31.9%). ALL-injuries were significantly associated with a tear of the posterior horn of the lateral meniscus (51.9% versus 16.2% of patients with/without ALL-injury, respectively), tear of the medial collateral ligament (40.4% versus 16.2% of patients with/without ALL-injury, respectively) and tear of the lateral collateral ligament (19.2% versus 0.9% of patients with/without ALL-injury, respectively).
Conclusion: The findings of our study demonstrate that in patients with ACL-tears, the frequency of RL and ALL-injuries is about 32%. Both injuries are frequently associated with further injuries of the knee, which are different between RL and ALL-injuries.
Limitations: Lesions not confirmed with arthroscopy.
Ethics committee approval: Approval of the Ethics Committee of the Medical University of Graz.
Funding for this study: No funding was received for this study.

Author Block: M. Klontzas¹, E. Vassalou¹, G. A. Kakkos¹, K. Spanakis¹, A. Zibis², A. H. Karantanas¹, K. Marias¹; ¹Heraklion/GR, ²Larissa/GR
Purpose or Learning Objective: Bone marrow lesions (BMLs) of the knee are commonly found in the context of subchondral insufficiency fractures (SIF) and advanced osteoarthritis (OA). The purpose of our work was to utilize deep learning in the form of convolutional neural networks to differentiate between the two conditions.
Methods or Background: The study dataset consisted of MRIs of knees with BMLs in the context of SIF (n=212) and OA (n=102), which were retrospectively collected and augmented to create a final dataset of 1174 images. Transfer learning was applied by utilizing an ImageNet-pretrained InceptionV3 convolutional neural network (CNN) which was fine-tuned with the use of the aforementioned MRI dataset. CNN performance was assessed on a validation cohort of 87 images of each group and was compared to that of two MSK radiologists with the use of receiver operating characteristics (ROC) curves and areas under the curve (AUC). Precision, recall and f1-scores were computed for the CNN and expert readers.
Results or Findings: InceptionV3 achieved an AUC of 93.68%, correctly classifying 82/87 OA and 81/87 SIF validation images. The first of the two MSK radiologists performed equally to the CNN, achieving an AUC of 91.95%, whereas the performance of the second expert MSK radiologist was significantly lower compared to both the CNN and the other reader (P<0.001) reaching an AUC of 82.76%.
Conclusion: A CNN model was highly accurate in differentiating between SIF and OA, achieving a higher or equal performance to MSK radiologists.
Limitations: Not applicable.
Ethics committee approval: Approved by the University Hospital of Heraklion (No 360/08/29-04-2020).
Funding for this study: Not applicable.

Author Block: C. Smekens¹, E. De Smet², E. Roelant¹, T. Vande Vyvere¹, A. Snoeckx², P. Van Dyck²; ¹Wilrijk/BE, ²Edegem/BE
Purpose or Learning Objective: To evaluate the interchangeability of multi-contrast 3D controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA) sampling perfection with application optimized contrast using different flip angle evolutions (SPACE) TSE and standard 2D TSE knee MRI in a large patient population.
Methods or Background: 250 symptomatic subjects prospectively underwent knee MRI on a 3T system with a 15-channel knee coil. The imaging protocol consisted of PD-weighted, T1-weighted and fat-suppressed T2-weighted standard 2D TSE acquisitions (12:14 minutes) and a commercially available multi-contrast 3D CAIPIRINHA SPACE TSE protocol (9:26 minutes). Approximately 13% of the patients (n=33) previously underwent anterior cruciate ligament and/or meniscus surgery. Two experienced musculoskeletal radiologists independently evaluated all datasets for technical image quality and identified pathologies of knee structures using a 4-point Likert scale representing the level of diagnostic confidence. The interchangeability of 2D and 3D protocols was tested under the same-reader scenario using a bootstrap percentile confidence interval. Finally, interreader reliability and intermethod concordance were assessed for all detected pathologies.
Results or Findings: Although 2D acquisitions scored better in terms of image quality and diagnostic confidence, standard 2D and accelerated 3D protocols were found interchangeable for diagnosing structural abnormalities, except for patellar (6.8% difference; 95%-CI: 4.0-9.6) and trochlear (3.6% difference; 95%-CI: 0.8-6.6) cartilage defects. Additionally, interreader reliability was high for both 2D and 3D protocols (range κ, 0.785-1 and κ, 0.725-0.964, respectively) and the intermethod concordance was very good for all diagnoses (range κ, 0.817-0.986).
Conclusion: This study contributes to the growing evidence that accelerated 3D protocols are a valuable time-saving alternative for standard 2D knee MRI. Yet, the radiologists' preference for 2D images needs to be further challenged to achieve a paradigm shift.
Limitations: Not applicable.
Ethics committee approval: The ethics committee approval was obtained.
Funding for this study: CS: B-Q MINDED (EU H2020 MSCA ETN #764513), UAntwerpen SEP #44883.