Deep Learning Approach for Anterior Cruciate Ligament Lesion Detection: Evaluation of Diagnostic Performance Using Arthroscopy as the Reference Standard

The most commonly injured ligament in the human body is an anterior cruciate ligament (ACL). ACL injury is standard among the football, basketball and soccer players. The study aims to detect anterior cruciate ligament injury in an early stage via efficient and thorough automatic magnetic resonance imaging without involving radiologists, through a deep learning method. The proposed approach in this paper used a customized 14 layers ResNet-14 architecture of convolutional neural network (CNN) with six different directions by using class balancing and data augmentation. The performance was evaluated using accuracy, sensitivity, specificity, precision and F1 score of our customized ResNet-14 deep learning architecture with hybrid class balancing and real-time data augmentation after 5-fold cross-validation, with results of 0.920%, 0.916%, 0.946%, 0.916% and 0.923%, respectively. For our proposed ResNet-14 CNN the average area under curves (AUCs) for healthy tear, partial tear and fully ruptured tear had results of 0.980%, 0.970%, and 0.999%, respectively. The proposing diagnostic results indicated that our model could be used to detect automatically and evaluate ACL injuries in athletes using the proposed deep-learning approach.

Download Full-text

Deep Learning for Detection of Complete Anterior Cruciate Ligament Tear

Journal of Digital Imaging ◽

10.1007/s10278-019-00193-4 ◽

2019 ◽

Vol 32 (6) ◽

pp. 980-986 ◽

Cited By ~ 10

Author(s):

Peter D. Chang ◽

Tony T. Wong ◽

Michael J. Rasiej

Keyword(s):

Anterior Cruciate Ligament ◽

Deep Learning ◽

Cruciate Ligament ◽

Anterior Cruciate Ligament Tear ◽

Ligament Tear ◽

Anterior Cruciate

Download Full-text

Diagnostic performance of 3D TSE MRI versus 2D TSE MRI of the knee at 1.5 T, with prompt arthroscopic correlation, in the detection of meniscal and cruciate ligament tears

Radiologia Brasileira ◽

10.1590/0100-3984.2015.0042 ◽

2016 ◽

Vol 49 (2) ◽

pp. 69-74 ◽

Cited By ~ 15

Author(s):

Francisco Abaeté Chagas-Neto ◽

Marcello Henrique Nogueira-Barbosa ◽

Mário Müller Lorenzato ◽

Rodrigo Salim ◽

Maurício Kfuri-Junior ◽

...

Keyword(s):

Anterior Cruciate Ligament ◽

Diagnostic Performance ◽

Spin Echo ◽

Cruciate Ligament ◽

Three Dimensional ◽

Meniscal Tears ◽

Turbo Spin Echo ◽

Turbo Spin ◽

Anterior Cruciate ◽

Cruciate Ligament Tears

Abstract Objective: To compare the diagnostic performance of the three-dimensional turbo spin-echo (3D TSE) magnetic resonance imaging (MRI) technique with the performance of the standard two-dimensional turbo spin-echo (2D TSE) protocol at 1.5 T, in the detection of meniscal and ligament tears. Materials and Methods: Thirty-eight patients were imaged twice, first with a standard multiplanar 2D TSE MR technique, and then with a 3D TSE technique, both in the same 1.5 T MRI scanner. The patients underwent knee arthroscopy within the first three days after the MRI. Using arthroscopy as the reference standard, we determined the diagnostic performance and agreement. Results: For detecting anterior cruciate ligament tears, the 3D TSE and routine 2D TSE techniques showed similar values for sensitivity (93% and 93%, respectively) and specificity (80% and 85%, respectively). For detecting medial meniscal tears, the two techniques also had similar sensitivity (85% and 83%, respectively) and specificity (68% and 71%, respectively). In addition, for detecting lateral meniscal tears, the two techniques had similar sensitivity (58% and 54%, respectively) and specificity (82% and 92%, respectively). There was a substantial to almost perfect intraobserver and interobserver agreement when comparing the readings for both techniques. Conclusion: The 3D TSE technique has a diagnostic performance similar to that of the routine 2D TSE protocol for detecting meniscal and anterior cruciate ligament tears at 1.5 T, with the advantage of faster acquisition.

Download Full-text

Increased Lateral Femoral Condyle Ratio Measured by MRI Is Associated with Higher Risk of Noncontact Anterior Cruciate Ligament Injury

10.21203/rs.3.rs-1087807/v1 ◽

2021 ◽

Author(s):

Miao He ◽

Jie Li

Keyword(s):

Risk Factors ◽

Logistic Regression ◽

Anterior Cruciate Ligament ◽

Diagnostic Performance ◽

Femoral Condyle ◽

Cruciate Ligament ◽

Acl Injury ◽

Tibial Slope ◽

Lateral Femoral Condyle ◽

Anterior Cruciate

Abstract Background Studies have shown a significant association between the radiographically measured lateral femoral condyle ratio (LFCR) and anterior cruciate ligament (ACL) injury. However, it is unclear whether the the LFCR measured by magnetic resonance imaging (MRI) is associated with risk of noncontact ACL injury. Objective 1 To investigate the effect of the LFCR on the risk of noncontact ACL injury by MRI. 2 To investigate the association of the LFCR measured by MRI with multiple bone morphological risk factors and evaluate the most sensitive risk predictors of noncontact ACL injury. Methods A total of 116 patients, including 58 subjects with noncontact ACL injury and 58 age-matched and sex-matched controls with only meniscus injury, were included in this retrospective case-control study. The LFCR, lateral tibial slope (LTS), lateral tibial height (LTH), medial tibial slope (MTS), and medial tibial depth (MTD) were measured on MRI. The differences in each index between the two groups were compared, and risk factors were screened by single-factor logistic regression analysis. Indicators with P values <0.1 were included in the logistic regression equation. The critical values and areas under the curve (AUCs) of independent risk factors were determined by receiver operating characteristic (ROC) curve analysis. Finally, the diagnostic performance of each risk factor was evaluated by the Z-test. Results A total of 116 patients who met the inclusion criteria were included in the final analysis (58 cases in the noncontact ACL injury group and 58 cases in the control group). Patients with noncontact ACL injury had a higher femoral LFCR (63.5±2.7%) than patients with simple meniscus tear. Among all the risk factors for ACL injury, the AUC for the LFCR was the largest, at 0.81 (95% CI, 0.73-0.88), and when the critical value was 61.35%, the sensitivity and specificity for the diagnosis of ACL injury were 79% and 67%, respectively. Combined with the LTH (> 2.35 mm), the diagnostic performance was improved. The AUC was 0.85 (95% CI, 0.78-0.92), the sensitivity was 0.83, and the specificity was 0.76 Conclusion This study shows that an increased LFCR is related to an increased risk of noncontact ACL injury by MRI. The LFCR and LTH are the most sensitive risk factors for noncontact ACL injury and may help clinicians identify individuals prone to ACL injury, allowing prevention and intervention measures to be applied.

Download Full-text

Fully Automated Diagnosis of Anterior Cruciate Ligament Tears on Knee MR Images by Using Deep Learning

Radiology Artificial Intelligence ◽

10.1148/ryai.2019180091 ◽

2019 ◽

Vol 1 (3) ◽

pp. 180091 ◽

Cited By ~ 15

Author(s):

Fang Liu ◽

Bochen Guan ◽

Zhaoye Zhou ◽

Alexey Samsonov ◽

Humberto Rosas ◽

...

Keyword(s):

Anterior Cruciate Ligament ◽

Deep Learning ◽

Cruciate Ligament ◽

Mr Images ◽

Automated Diagnosis ◽

Anterior Cruciate ◽

Cruciate Ligament Tears

Download Full-text

Diagnostic Performance of Artificial Intelligence for Detection of Anterior Cruciate Ligament and Meniscus Tears: A Systematic Review

Arthroscopy The Journal of Arthroscopic and Related Surgery ◽

10.1016/j.arthro.2020.09.012 ◽

2020 ◽

Author(s):

Kyle N. Kunze ◽

David M. Rossi ◽

Gregory M. White ◽

Aditya V. Karhade ◽

Jie Deng ◽

...

Keyword(s):

Artificial Intelligence ◽

Systematic Review ◽

Anterior Cruciate Ligament ◽

Diagnostic Performance ◽

Cruciate Ligament ◽

Meniscus Tears ◽

Anterior Cruciate

Download Full-text

Meniscal ramp lesions: diagnostic performance of MRI with arthroscopy as reference standard

La radiologia medica ◽

10.1007/s11547-021-01375-3 ◽

2021 ◽

Author(s):

Marcello Zappia ◽

Luca Maria Sconfienza ◽

Salvatore Guarino ◽

Michele Tumminello ◽

Germano Iannella ◽

...

Keyword(s):

Diagnostic Performance ◽

Cruciate Ligament ◽

Bone Bruise ◽

Reference Standard ◽

Meniscal Lesion ◽

Exact Test ◽

Orthopedic Surgeons ◽

Meniscal Lesions ◽

Anterior Cruciate ◽

Meniscotibial Ligament

Abstract Background The posteromedial meniscal region is gaining interest among orthopedic surgeons, as lesions of this area has been reported to be significantly associated with anterior cruciate ligament tears. The current imaging literature is unclear. Purpose To evaluate the diagnostic performance of MR in the detection of meniscal ramp lesions having arthroscopy as reference standard. Materials and methods We retrospectively included 56 patients (mean age of 25 ± 7 years; 14 females) from January to November 2017 with a arthroscopically proved ACL tear and posterior meniscocapsular separation. On preoperative MRI, two radiologists with 13 and 2 years’ experience in musculoskeletal imaging assessed the presence/absence of ramp lesion, meniscotibial ligament lesion, peripheral meniscal lesion, or their combination, bone bruise. Having arthroscopy as reference standard, diagnostic performance of MRI in the evaluation of ramp area lesions was calculated. Cohen’s kappa (k) and Fisher's Exact Test statistics were used. Results Agreement between radiologists ranged from κ = 0.784 (meniscotibial ligament lesions) to κ = 0.918 red–red meniscal lesion. Sensitivities were 97.4% for ramp lesions, 95.8% for meniscotibial ligament lesion, 94.4% for peripheral meniscal lesions; specificities were 88.9%, 81.3%, and 97.4%, respectively; accuracies were 94.6%, 87.5%, and 96.4%, respectively. Agreement between MR and arthroscopy was almost perfect in identification of ramp lesions (κ = 0.871) and red–red zone meniscal lesions (κ = 0.908). The agreement between the two methods was substantial (κ = 0.751) for meniscotibial lesion. No significant association between tibial plateau bone bruise and the different type of lesions was found (κ ≥ 0.004 and p ≥ 0.08). Conclusion MR has high diagnostic performance in meniscal ramp area lesion assessment, with substantial to almost perfect inter-reader agreement.

Download Full-text