RPS 1410 - Detecting damage: sports and trauma imaging

Author Block: M. Shoura¹, S. M. Elhennawy², Y. A. A. E. F. Hassan²; ¹Boston, MA/US, ²Cairo/EG
Purpose: Evaluate predictive value of MRI-based BAMIC & MLG-R classification systems for RTP time and re-injury risk with thigh muscle injuries in professional football players with the potential enhancement of AI integration.
Methods or Background: Retrospective study on 48 professional football players with recent thigh muscle injuries in biceps femoris, adductor longus and rectus femoris muscles between 2018 and 2024. Two professional MSK radiologists independently classified injuries.. Statistical analysis included Pearson correlation, ROC analysis, inter-rater reliability (κ), and others. AI model was trained to predict time to RTP and re-injury rate
Results or Findings: BAMIC showed strong correlation with RTP time (r = 0.72) and good predictive accuracy (AUC = 0.81,). MLG-R also demonstrated high correlation (r = 0.66) with AUC = 0.77. Anatomical location significantly impacted outcomes: biceps femoris distal myotendinous T-junction injuries showed longest RTP time and highest re-injury rate (37.5%). Length of injury and intermuscular fluid also showed prolonged RTP. Combined classification model improved correlation to r = 0.76 with AUC = 0.84. AI based model achieved cross-validation accuracy of 79.2% with improved correlation (r = 0.79, AUC = 0.87).
Conclusion: BAMIC and MLG-R classification systems demonstrate good-to-strong predictive value for RTP timing in professional football thigh muscle injuries. Specific anatomical injury locations significantly influence recovery outcomes. Other MRI parameters including cranio-caudal length and intermuscular fluid volume provide additional prognostic value. AI-enhanced models incorporating anatomical and quantitative factors achieve superior predictive accuracy
Limitations: Retrospective design with inherent selection bias. Limited sample size (n=48) may affect statistical power and generalizability. Study limited to Egyptian and Gulf region professional football players, potentially affecting generalizability to other sports. Machine learning model requires external validation in larger independent cohorts. Rehabilitation protocols varied between clubs, introducing potential confounding variables
Funding for this study: Saudi German Hospital and Paxera-health
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: A. Piacentini, K. Abdelmaguid, S. Gitto, S. Fusco, C. Messina, F. Ambrogi, A. Vanzulli, L. M. Sconfienza, D. Albano; Milan/IT
Purpose: Accurate estimation of anterior cruciate ligament (ACL) tear timing is essential for clinical and medico-legal purposes. We aimed to identify MRI parameters linked to lesion chronological age and develop a logistic-regression model to distinguish acute (≤12 weeks) from chronic (>12 weeks) injuries.
Methods or Background: This retrospective study included 426 patients with clinically and MRI-confirmed ACL tears (2014–2024). MRI (1.5/3 T) assessed 11 direct and indirect signs including bone oedema, effusion, tear morphology, ACL atrophy, ACL thinning, "wavy” morphology, anterior tibial translation, angulated posterior cruciate ligament and uncovered lateral meniscus. Firth-penalized logistic regression predicted injury timing; discrimination was assessed by AUC with bootstrap validation (n=200) and a simplified rule-based score was developed for clinical use.
Results or Findings: Among the 426 patients (mean age 35±14 years; 71% male), 261 had acute tears, 73 sub-acute, and 92 chronic. The regression model achieved excellent discrimination between recent and chronic tears with AUC 0.97 (bias-corrected 0.96); at the optimal Youden-index threshold (0.56) it yielded 96% sensitivity, 96% specificity, 96% accuracy, and PPV 0.99. Key discriminators of chronicity were ACL atrophy, thinning, and inhomogeneous thickening. The simplified algorithm–requiring either all major (atrophic ACL, thinned ACL, inhomogeneous thickening of the ACL) or ≥5 minor criteria (bone edema, effusion, uncovered lateral meniscus, full thickness tear with fibers edema, absent wavy ligament, absent anterior tibial translation)–retained high performance (sensitivity 94–96%, specificity 95–96%), closely mirroring the full model.
Conclusion: MRI-based multifeatured analysis accurately distinguishes acute from chronic ACL tears. The simplified rule-based model retains high diagnostic performance and it is practical for routine imaging and medico-legal timing of ACL injuries.
Limitations: Single-center retrospective design; exclusion of clinical variables (age, mechanism of injury) may limit generalizability; the model yields probability estimates rather than a definitive diagnosis.
Funding for this study: No funding received.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Comitato Etico Territoriale Lombardia 1 (RETRO-RAD, CE 61/INT/2017, 12 March 2025).

Author Block: R. Sivanandan¹, L. Tveiten², A. J. Fagerlund³, K. Malm-Nicolaisen³, B. A. Graff²; ¹Asker/NO, ²Drammen/NO, ³OSLO/NO
Purpose: To evaluate the therapeutic impact and clinical relevance of radiologist reporting on AI-positive fracture detections in trauma patients, within a workflow where orthopedists often initiate treatment prior to radiologist review.
Methods or Background: This retrospective, single-center study was conducted over two months following the implementation of an AI algorithm for fracture detection in trauma care. Of 2,394 consecutive skeletal radiographs reviewed, 1,127 (47%) AI-positive cases were included and detailed analysis on fracture cases performed. Data collected include patient demographics, fracure type, AI-report, radiologist's-report, musculoskeletal-radiologists (MSKr) report and orthopedist's-report . Discordant cases between AI alone and AI+radiologist interpretations were analyzed for therapeutic impact with orthopedists and MSKr as ground truth.
Results or Findings: In AI-positive cases, concordance between AI and radiologist reports for fractures was high (97%) and discordance was noticed in 3% (20 cases). Of the 20 discordant cases, the MSK radiologists agreed with AI in 3 cases (15%) and disagreed in 17 (85%). The orthopedists with AI in 6 cases (30%), disagreed in 12 (60%), and was uncertain in 2 cases (10%). Ground truth was defined by the orthopedist's assessment, with the MSKr used to resolve the 2 uncertain cases. False-positive AI results led to 1 over-treatment, 1 under-treatment while others received adequate necessary treatment (Fig-4).
Conclusion: In an AI-supported workflow for fracture detection, therapeutic decisions were primarily driven by orthopedists integrating AI with their clinical judgment. Our findings suggest that adding radiologist input to AI may not significantly improve accuracy in discordant cases, whereas an AI+orthopedits approach appears more clinically relevant and efficient—allowing radiologists to focus on complex, high-impact tasks.
Limitations: This is a single-site retrospective study and may be feasible when orthopedists assess the radiographs with AI as part of routine clinical care.
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: ID (24/05320) by the Data Protection Officer of the Vestre Viken Hospital Trust

Author Block: S. Fusco¹, L. Cerezal Pesquera², A. M. Perez³, F. Idoate⁴, L. M. Sconfienza¹, E. Llopis San Juan⁵; ¹Milan/IT, ²Santander/ES, ³Terrassa/ES, ⁴Pamplona/ES, ⁵Alzira/ES
Purpose: To describe MRI findings associated with ankle injuries involving the deltoid complex, with focus on fascial sleeve avulsion of the medial malleolus and periosteal stripping of the flexor retinaculum, and to assess their relationship with injury timing.
Methods or Background: Two musculoskeletal radiologists retrospectively evaluated 165 ankle MRIs with deltoid complex injury. Each study was reviewed for: superficial and deep deltoid tears, fascial sleeve avulsion, flexor retinaculum stripping, lateral ligaments and syndesmotic tears, osteochondral lesions(OCLs), soft tissue impingement, pathology of the posterior tibial tendon(PTT), and peroneal tendons. Injuries were classified as acute, chronic, or acute-on-chronic. Descriptive statistics and chi-square tests were used.
Results or Findings: Deep deltoid tears were present in 160/165 cases. Superficial deltoid proximal tear was seen in 84/165(50.9%). Fascial sleeve avulsion occurred in 74/165(44.8%), and flexor retinaculum stripping in 66/165(40.0%); both coexisted as “complete fascial sleeve” in 55/165(33.3%). These lesions were associated with superficial deltoid proximal tears (fascial sleeve in 88.1% vs 0%; retinaculum stripping in 66.7% vs 12.3%; complete fascial sleeve in 65.5% vs 0%; p<0.0001). The prevalence of complete fascial sleeve increased with chronicity, being less common in acute (25.4%) and more frequent in chronic (45.2%) and acute-on-chronic injuries (64.1%) (p<0.001). Isolated deltoid injuries were rare: >95% also had lateral ligament tears. Additional findings included syndesmotic injuries (9%), OCLs (15%), soft tissue impingement (24%), PTT abnormalities (27%), and peroneal pathology (28%).
Conclusion: Deltoid complex injuries are rarely isolated and usually coexist with lateral ligament tears. Fascial sleeve avulsion and retinacular stripping—especially in combination—are closely associated with proximal superficial deltoid tears, likely functioning as a biomechanical unit. These findings are more common in chronic injuries, reflecting their link to chronic lateral ankle instability.
Limitations: Retrospective; lack of surgical correlation; lack of long-term outcomes
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: D. Amiras, K. Chappell, C. Tsitsifylla, H. Lanz, Y. Joshi, J. Mcginley, C. Gupte, M. Ristic; London/UK
Purpose: A novel rotating low-field MRI can harness the magic angle effect to identify collagen fibre alignment and orientation in vivo. Typically, imaging collagen fibres relies on negative contrast, caused by adjacent water or haemorrhage, to depict injuries, as collagen produces inherently low signal on most MR techniques. In this study, we utilise magic angle directional imaging (MADI) to image the anterior cruciate ligament (ACL) in patients being investigated for ACL tears and correlate the findings with surgical outcomes.
Methods or Background: A novel 0.15T rotating transverse magnet allowed reorientation of the main B0 field. Seven to eight 3D T1 FLASH isotropic volumes at different B0 orientations, relative to the projected orientation of the ACL, were acquired. In post-processing, co-registration of these volumes was performed, and collagen fibre vectors were calculated for regions of interest corresponding to the ACL. These vectors were then visualised as tractograms superimposed on the previously acquired volume. Conventional MRI and magic angle MR images (MAMRI) were compared to arthroscopic findings.
Results or Findings: Five cases from four patients were compared with arthroscopic findings. Three cases had complete ACL tears, whereas one case had a 'cyclops' lesion post-ACL reconstruction, and the same patient had an old partial ACL tear on their contralateral knee. MAMRI performed at least as well as conventional MRI and, in one case, correctly predicted a complete tear when conventional MRI suggested a partial-thickness tear.
Conclusion: This is the first report demonstrating the clinical applicability of low-field magic angle MRI in the assessment of ACL tears in an injured population. Whilst only five cases had surgical correlation, the images obtained were promising and revealed detail not previously visible on conventional MRI.
Limitations: Limited sample size; Arthroscopic subjectivity; Limited data on MAMRI interpretation
Funding for this study: NIHR i4i Grant II-LA-1111-20005​
Wellcome Trust Innovator Award
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: REC reference: 24/LO/0531

Author Block: E. Höhne, S. Mahmoudi, T. Vogl, K. Eichler, C. Booz, I. Yel; Frankfurt/DE
Purpose: Prior to player transfers professional soccer clubs conduct comprehensive musculoskeletal assessments in healthy athletes to evaluate their current condition and injury risk. Magnetic resonance imaging (MRI) plays a central role in this process by detecting musculoskeletal abnormalities even in the absence of symptoms. This study presents common MRI findings in a cohort of professional soccer players and aims to improve understanding of the physical condition of elite athletes.
Methods or Background: In this retrospective analysis, 3T MRI examinations performed between August 2019 and March 2025 during pre-signing assessments were reviewed. Clinical data were extracted from medical records and supplemented with functional information. Non-age-typical structural alterations were systematically categorized, and prevalence patterns were analyzed by playing position.
Results or Findings: Fifty professional players (mean age 25.4 ± 4.7 years) were included. Secondary cleft signs (42%) and lumbar disc bulging (40%) were the most frequent abnormalities. Additional common findings comprised knee chondropathy (34%), acetabular labral degeneration (26%), femoroacetabular impingement (22%), and various soft tissue or bone-related changes. While most abnormalities were evenly distributed across positions, the prevalence of secondary clefts differed significantly (χ² = 8.07, p = 0.045), with strikers showing the highest proportion (69%).
Conclusion: MRI screening in professional soccer players most frequently revealed structural musculoskeletal abnormalities, many of which were also present in asymptomatic individuals. This underscores the ability of MRI to detect subclinical alterations that may later develop into clinically relevant conditions. Overall, the findings support routine use of MRI in pre-signing assessments to guide individualized prevention, rehabilitation, and long-term player management.
Limitations: The relatively small sample size, absence of a control group, and uneven distribution of playing positions and limb dominance may limit the generalizability of these findings.
Funding for this study: Not applicable.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethical approval was obtained from the local ethics committee of the university hospital Frankfurt (2024-2129) and informed consent was obtained.

Author Block: S. Fusco¹, F. Idoate², L. Cerezal Pesquera³, A. M. Perez⁴, L. M. Sconfienza¹, E. Llopis San Juan⁵; ¹Milan/IT, ²Pamplona/ES, ³Santander/ES, ⁴Terrassa/ES, ⁵Alzira/ES
Purpose: To describe the prevalence, distribution, and clinical associations of exercise-related signal abnormalities (ERSA) in lower limb MRI of athletes, and to assess their relationship with symptoms and concomitant muscle injuries.
Methods or Background: One musculoskeletal radiologist reviewed lower limb MRI of professional and non-professional athletes performed for suspected muscle injury or follow-up, to identify ERSA lesions (ovoid peritendinous and/or ring-shaped perifascial edema).
For each MRI, the following parameters were assessed: ERSA type (A/B/C), muscle(s) involved, longitudinal distribution, laterality, relationship to the clinically targeted compartment (site of suspected/ongoing injury).
Correlation with reported pain and with acute, subacute, or chronic lesions/scar was assessed. Descriptive statistics and Fisher’s exact/Chi-square tests were used.
Results or Findings: ERSA lesions were found in 30 athletes (mean age 26.7 years; 27 male; 21 soccer players), involving a single muscle in 19/30 (63.3%), two in 8/30 (26.7%), and ≥3 in 3/30 (10.0%). Subtypes included type B in 15(50.0%), type C in 13(43.3%), and type A in 2(6.7%). The rectus femoris was most frequently affected(n=12), followed by soleus, vastus lateralis, and adductor major (n=4). Pain in the ERSA compartment was present in 20/30(66.7%). In 12, pain was explained by concomitant acute/subacute injuries, while in 8 it occurred with isolated ERSA. Conversely, 10/30 showed ERSA in non-target compartments without pain.
No significant association was found between pain and ERSA subtype(p=0.099).
Conclusion: ERSA are a relatively frequent MRI finding in athletes investigated for suspected muscle injury. Our data suggest a mixed picture: ERSA may explain pain in the absence of associated muscle injuries but are also frequently found in asymptomatic compartments. Prospective studies are needed to clarify their clinical relevance or determine whether they represent overload-related changes.
Limitations: Retrospective design; relatively small sample size; lack of prospective symptom monitoring
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: C. Otgonbaatar¹, S. Cha², S. H. Jeon², H. Shim¹, S. M. Ko², H. Kim²; ¹Seoul/KR, ²Wonju-si/KR
Purpose: This study aimed to evaluate the quantitative and qualitative image analysis of ultra-low-dose CT with deep learning image reconstruction (DLR) compared with hybrid iterative reconstruction (IR) in patients with ankle and foot fractures.
Methods or Background: This retrospective study included 32 patients who underwent ultra-low-dose CT of the ankle and foot (effective dose ~0.86 μSv). Images were reconstructed with DLR and hybrid IR. Objective analysis included image noise, SNR, CNR, and sharpness (assessed with a no-reference perceptual sharpness metric). Subjective quality was rated by two board-certified radiologists on a 5-point scale.
Results or Findings: DLR demonstrated clear advantages over hybrid IR, with substantially lower image noise (47.33 ± 6.60 HU vs. 87.65 ± 12.48 HU) and significantly higher SNR (31.42 ± 5.81 vs. 19.15 ± 4.48) and CNR (50.59 ± 7.62 vs. 24.08 ± 3.81) (all p < 0.001). Image sharpness was also enhanced (0.76 ± 0.09 for DLR vs. 0.59 ± 0.09 for hybrid IR, p < 0.001). Subjective evaluations confirmed the superiority of DLR, with improved ratings for overall image quality, trabecular architecture, cortical bone delineation, and fracture depiction (all p = 0.001).
Conclusion: Ultra-low-dose CT reconstructed with DLR improves image quality while maintaining minimal radiation exposure in the assessment of ankle and foot fractures.
Limitations: Not applicable
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Institutional Review Board

Author Block: A. Şükün¹, C. Özenbaş², B. Sargın¹, K. Gökkuş¹, M. Ş. Sahin¹; ¹Antalya/TR, ²Izmir/TR
Purpose: To assess the diagnostic accuracy of a semi-quantitative signal scoring system for anterior cruciate ligament (ACL) tears on preoperative 1.5T knee MRI, and to compare the performance of conventional radiology reports with ChatGPT-based scoring, using arthroscopic findings as the reference standard.
Methods or Background: This retrospective single-center study received institutional review board approval (KA25/133, March 26, 2025). Between 2012 and 2024, 185 patients undergoing arthroscopic ACL surgery were screened. Exclusion criteria were the absence of preoperative MRI, poor image quality, or missing PACS data. After applying the criteria, 100 patients with surgically confirmed ACL rupture were included. Sagittal T1- and PD-weighted sequences were anonymized and exported in JPEG format (300 dpi, 1350×1200 pixels). Three PD and three T1 sagittal slices best depicting the ACL were selected. Semi-quantitative scoring was applied exclusively to PD images by freehand ROI placement, classifying ACLs as intact (<2), partial (2–3.5), or complete (>3.5) on a four-point scale. T1 images were used for anatomical correlation. Scoring was performed independently by ChatGPT, with surgery as the gold standard.
Results or Findings: A total of 100 patients (86 males [86%], 14 females [14%]; mean age, 31.99 ± 9.25 years) underwent preoperative knee MRI and arthroscopic ACL surgery. All had surgically confirmed complete ACL tears. ChatGPT achieved 64% diagnostic accuracy, compared to 45% for radiology reports. The semi-quantitative scoring system with ChatGPT improved accuracy by ~20% versus conventional reporting.
Conclusion: ChatGPT-assisted semi-quantitative scoring significantly enhanced the accuracy of preoperative MRI in diagnosing ACL tears compared to routine radiology reporting, suggesting its potential as a supportive decision-making tool in clinical practice.
Limitations: The retrospective design, single-center cohort, and restriction to surgically confirmed complete tears limit generalizability. External validation with larger, multi-center, and prospective datasets is warranted.
Funding for this study: No
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: KA25/133, March 26, 2025