RPS 810 - Hot Topic: musculoskeletal imaging meets sustainability

Author Block: J. Mendonca, J. D. A. Batista Araújo Filho, A. G. ORMOND FILHO, T. R. Giacometti, A. Motoyama Caiado, B. A. Rocha, N. Y. Hashimoto, M. Nico, D. Bahia; São Paulo/BR
Purpose: To describe our institutional experience deploying a deep learning reconstruction (DLR) algorithm for musculoskeletal MRI in an outpatient setting, assessing its impact on MRI acquisition time, image quality, and collaborative workflow adaptation.
Methods or Background: This retrospective study evaluated MSK MRI exams performed at a large private radiology network in Brazil. Following vendor-guided scanner upgrades, a k-space-based DLR algorithm (AIR™ Recon DL) was integrated into native reconstruction pipelines across multiple systems. Workflow adjustments and protocol standardization were co-developed in collaboration with the vendor and internal radiologist leads. We compared MRI studies from 12 months before (Group 1) and after (Group 2) DLR implementation. Acquisition time was automatically calculated per exam. A subset of cases was randomly selected for blinded qualitative image review using a 5-point Likert scale. Estimated energy use and CO₂e emissions were calculated using scanner-specific consumption metrics and regional emission factors.
Results or Findings: A total of 22,165 MSK MRI scans were analyzed: 10,492 in Group 1 and 11,673 in Group 2. Median scan duration decreased by 53% post-implementation, with the most substantial reductions in shoulder (62%), wrist (59%), knee (52%), spine (38%) and hip (33%) scans. Workflow optimization reduced scanner idle time, improved patient throughput, and lowered anxiety- and motion-related interruptions. Radiologist image quality assessments remained stable, with occasional feedback on altered image texture. Annual energy savings exceeded 2.3 MWh per scanner, avoiding over 1 metric ton of CO₂e emissions.
Conclusion: This real-world implementation of AI-driven DLR demonstrated tangible workflow optimization, improved patient experience, and meaningful environmental benefits. Ongoing collaboration among vendors, radiologists, and technologists was essential for successful adoption, underscoring the importance of clinician–industry partnerships in ensuring safe and efficient AI deployment.
Limitations: Retrospective study with limited image quality sampling and no cost-effectiveness analysis.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved in 05-14-2025

Author Block: F. Herr, N. Hohmann, C. A. Dascalescu, V. M. Schäfer, H. Gildein, B. Hoppe, J. Ricke, J. Arnholdt, P. Reidler; Munich/DE
Purpose: To evaluate whether a 3D isotropic MRI localizer (FastView) can reliably and efficiently assess femoral and tibial torsion compared to a dedicated MRI protocol, offering a faster alternative to a conventional dedicated MRI protocol.
Methods or Background: In this retrospective single-center study, 60 lower limbs from 30 patients (mean age 27.1 ± 11.5 years; 19 female, 11 male) were examined using FastView and a dedicated MRI protocol. FastView imaged the entire lower limb in 17.4 seconds (voxel size 5×5×5 mm³), whereas the standard protocol required nearly 7 minutes. Two independent readers measured femoral and tibial torsion angles. Agreement between methods was evaluated using intraclass correlation coefficients (ICCs), Bland–Altman plots, and Pearson’s R².
Results or Findings: Across 60 limbs, no significant differences were observed in all measured angles between FastView and the dedicated protocol (p > 0.05). Femoral torsion showed excellent inter-modality agreement (ICC: 0.91–0.96), with mean differences ranging from –2.2 ± 3.4° (p = 0.001) to –3.7 ± 4.4° (p < 0.001). Tibial torsion also showed excellent agreement (ICC: 0.91–0.94), with mean differences from –1.2 ± 4.6° (p = 0.16) to –2.1 ± 4.6° (p = 0.02). Inter-reader agreement was consistently high for both modalities (ICC: 0.95–0.99). Bland–Altman analyses and Pearson’s R² (range: 0.84–0.94) confirmed strong correlation and reliability.
Conclusion: FastView MRI localizer showed excellent agreement with the dedicated torsion protocol for femoral and tibial measurements. By reducing scan time and avoiding radiation, FastView promotes a more sustainable approach and may replace conventional multi-sequence protocols in practice.
Limitations: This retrospective single-center study has limited generalizability, and the sample size was relatively small. Although scan acquisition with FastView is reduced to ~17 seconds, preparation time remains similar to standard protocols.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the institutional review board (approval number: 23-0020; date of approval: 09 March 2023).

Author Block: J. Roshardt¹, S. Steppacher¹, M. K. Meier¹, S. Sommer², R. Sutter², F. Schmaranzer²; ¹Bern/CH, ²Zürich/CH
Purpose: Hip deformities require precise diagnosis before joint-preserving surgery, but standard radiographs and CT scans involve ionizing radiation, raising concerns about cumulative exposure, particularly in younger patients. MRI offers detailed, radiation-free imaging of bone morphology and soft tissues. Advances in deep learning enable automatic segmentation and MRI-based 3D pelvic modeling, allowing virtual radiographic-like projections. This study determined if standard radiographic parameters (lateral center-edge angle [LCEA], retroversion index [RI], and acetabular index [AI]) can be accurately measured on MRI-based projections compared to conventional anteroposterior (AP) pelvic radiographs.
Methods or Background: Fifty patients (100 hips, mean age 28 years) with symptomatic hip deformities underwent evaluation with standard AP pelvic radiographs and direct hip MR arthrography (including T1-w VIBE DIXON). Osseous pelvis was segmented using a validated 3D nnU-Net, cortical bone contrast augmentation and radiograph-like projections were generated using a cone beam projection algorithm. Arthrography and non-contrast sides were analyzed separately. Parameters were measured on both imaging modalities and compared using t-tests and Pearson's correlation.
Results or Findings: The LCEA showed no significant difference between conventional and MRI-based projections, with absolute mean differences of 0.2±1.2° (arthrography, p=0.71) and 0.6±1.7° (non-contrast, p=0.29). Similarly, AI showed no significant difference (0.6±1.5° arthrography, p=0.20); 1.±2.1° (non-contrast, p=0.075). For RI, absolute mean differences were 0.5±4% (arthrography, p=0.80) and 2.±5.3% (non-contrast, p=0.22), indicating no significant difference. Moderate to strong correlations were found between MRI-based and conventional radiographs (LCEA: r=0.84/0.77, AI: r=0.75/0.66, RI: r=0.60/0.66 for arthrography/non-contrast; all p<0.001).
Conclusion: MRI-based radiographic-like projections, whether with or without contrast agents, provide measurements comparable to AP pelvic radiographs for key hip morphology parameters. This approach eliminates radiation exposure while maintaining diagnostic accuracy, offering significant advantages for young patients requiring repeated imaging for hip deformity assessment and surgical planning.
Limitations: No external validation
Funding for this study: Funding received by the Swiss National Science Foundation (porject no. 205091)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee of the University of Bern

Author Block: M. Avanesov¹, C. Zickert², M. Stojkova³, G. F. Peter¹, J. Yamamura¹, H. Ragab¹; ¹Hamburg/DE, ²Lüneburg/DE, ³Duisburg/DE
Purpose: This dual-centre study analysed the additional value of CT imaging besides conventional radiography (x-ray) on the diagnosis and outcome of acute scaphoid fractures at two university medical centres in Europe.
Methods or Background: Two hundred and twelve consecutive patients (103 patients from Hamburg, Germany (mean age 38 ±19 years, 82 males) and 109 patients from Skopje, North Macedonia (mean age 38 ±18 years, 86 males)) who were diagnosed with scaphoid fractures were included in the European dual-centre study. All patients underwent conventional radiography in three views and all fractures were staged according to the dichotomic Herbert classification (Herbert A: stable fracture; Herbert B: unstable fracture). In Hamburg, all patients underwent additional CT imaging, whereas in Skopje only patients with suspected instability (Herbert Type B) underwent additional CT imaging. Demographics, treatment strategies, as well as clinical and functional outcomes were retrospectively compared between the two centres.
Results or Findings: All patients with a scaphoid fracture in Hamburg underwent additional CT imaging, whereas only 23 out of 109 patients (21 %) in Skopje had a complementary CT scan (p < .001). Fracture type distribution and stability were comparable between centres. While 42 out of 103 patients (41%) underwent surgical treatment in Hamburg, only 13 out of 109 patients (12%) were operated in Skopje (p < .001). However, there were no statistically significant differences in complication rates between the two sides (p = .489).
Conclusion: This dual-centre study suggests that using additional CT imaging only for unstable acute scaphoid fractures does not lead to worse clinical or functional outcomes or an increased rate of complications.
Limitations: Retrospective study design
Limited follow-up period
MRI imaging was not investigated
Funding for this study: No external funding sources.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethical approval was provided by the local institutional ethics committee. The requirement for written informed consent was waived due to the retrospective nature of the study. The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments.

Author Block: S. Mahmoudi, K. Eichler, L. D. Grünewald, T. J. Vogl, I. Yel; Frankfurt/DE
Purpose: The purpose of this study was to investigate the impact of deep learning-accelerated MRI on acquisition time and image quality in muscle injuries of professional soccer athletes.
Methods or Background: 28 professional male soccer athletes (mean age 25.4±3.4) who underwent 3-Tesla-MRI of the lower limb due to suspicion of muscle injuries were included. Standard sequences were acquired consisting of PDw imaging in axial, paracoronal and parasagittal plane, as well as T1-weighted imaging in axial plane. Additionally, PDw imaging sequences were acquired using the deep learning algorithm (PDDL).
Subjective evaluation was independently performed by three blinded radiologists. 5-point Likert scales were utilized to subjectively assess diagnostic confidence, image quality, and lesion sharpness. Objective analysis encompassed time efficiency and quantitative imaging parameters, including signal-to-noise-ratio (SNR) and contrast-to-noise-ratio (CNR), as well as operational costs including electricity and personnel.
Results or Findings: MRI revealed structural injuries in 21 of 28 athletes (75%). Subjective evaluations of PDDL sequences revealed excellent diagnostic confidence (median 5, interquartile range (IQR): 5-5), image quality (median 5, IQR: 5-5), and lesion sharpness (median 5, IQR: 4.5-5). In 91.7 % of cases, PDDL sequences were favored over conventional PD imaging. Objective image analysis of PDDL revealed significant higher SNR (115.0±9.1) and CNR values (80.5±12.8) when compared to conventional PD sequences (81.3±5.9 and 54.3±5.4, respectively)(p<0.05). Acquisition times could be reduced by an average of 39.7% using PDDL sequences.
Conclusion: Our findings suggest that deep learning MRI of muscle injuries among professional athletes allow a reduction in acquisition time while additionally improving image quality. The integration of these sequences may enhance operational efficiency and reduce ongoing costs, promoting a more sustainable and value-based approach in radiology.
Limitations: The study population is relatively small, which is attributable to the specific subcohort examined.
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the ethics committee of the University Hospital Frankfurt

Author Block: C. Zhu¹, I. d. Mata², C. Zhu³, E. Santana²; ¹Barcelos/PT, ²Lisboa/PT, ³Lisbon/PT
Purpose: Healthcare sustainability requires imaging strategies that maximize diagnostic yield while minimizing resource use. Musculoskeletal (MSK) findings are frequently visible on routine scans obtained for other indications, such as abdominal CT, pelvic MRI, or vascular imaging. Leveraging these “incidental windows” provides clinically relevant information without additional examinations, contributing to more efficient and environmentally responsible practice.
The aim of this presentation is to demonstrate how MSK findings can be opportunistically assessed on routine non-MSK scans, to discuss their role in the early detection of conditions such as osteoporosis or sarcopenia, and to illustrate how this approach contributes to sustainable imaging by reducing redundant examinations while enhancing patient care.
Methods or Background: A literature review and pictorial synthesis were performed, focusing on CT and MRI examinations where bone and soft-tissue structures are included but not primarily targeted. Examples include vertebral bone mineral density from abdominal CT, femoral muscle bulk in pelvic MRI, and vascular CT revealing both arterial calcifications and adjacent bone health.
Results or Findings: Analysis of routine scans demonstrates that clinically relevant MSK information can often be extracted without additional imaging. Vertebral Hounsfield units derived from abdominal CT correlate with osteoporosis risk, muscle cross-sectional area on pelvic MRI reflects sarcopenia, and joint or tendon morphology may be partially assessed on pelvic or vascular studies. Incorporating this opportunistic evaluation into structured reporting allows early recognition of disease, reduces the demand for dedicated MSK examinations, and contributes to a more sustainable imaging workflow.
Conclusion: “Buddy imaging” transforms routine scans into multipurpose tools, aligning clinical value with sustainability goals. Recognizing MSK features beyond the primary indication supports earlier intervention, avoids duplicate examinations, and exemplifies a greener radiology practice.
Limitations: No limitations were identified.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: T. C. Arnold¹, L. Wang¹, A. Shankaranarayanan¹, L. N. Tanenbaum²; ¹Menlo Park, CA/US, ²RIVERSIDE, CT/US
Purpose: Musculoskeletal MR protocols include multiple sequences with different acquisition planes (axial, sagittal, coronal) and contrasts (T1w, T2w, T2w fat-sat, PD, PD fat-sat). The exact combination of sequences varies widely across institutions. Previously, we demonstrated the ability to synthesize sequences in knee protocols, thus enabling faster or more comprehensive exams. However, that algorithm's utility is limited to knee exams. In this study, we present a unified model capable of synthesizing sequences across a wide range of musculoskeletal anatomies.
Methods or Background: We retrospectively analyzed 554 musculoskeletal MR exams to identify image synthesis sets containing a target image, a sequence-type match, and an acquisition-plane match. We held out 55 subjects for testing, which spanned 6 target sequences, 8 musculoskeletal regions, and 39 target/sequence-type/acquisition-plane combinations. The remaining subjects were used to train a patch-based neural network. Model performance was quantitative assessed using peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM).
Results or Findings: Across all synthesis combinations, average PSNR was 23.6±3.3 dB and SSIM 0.66±0.15. Averaging across target sequence, quantitative values ranged from PSNR=[22.5-25.7 dB] and SSIM=[0.59-0.80] with Ax PD FS having the highest (25.7±5.0 dB, 0.80±0.12) followed by Ax T2 FS (24.8±0.2 dB, 0.77±0.01). Across anatomies, values ranged from PSNR=[20.1-27.8 dB] and SSIM=[0.59-0.81] with elbow and ankle showing the best performance and wrist having the worst performance.
Conclusion: We provide a preliminary demonstration that a single model can synthesize a range of sequences from various musculoskeletal anatomies. This offers a flexible and generalizable approach that can match site-specific musculoskeletal protocols and adapt to protocol changes.
Limitations: Future studies should include a qualitative reader study by radiologists.
Funding for this study: Funding provided by Subtle Medical.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: