Patellar tendon (PT) stiffness is highly load-sensitive and can decrease by up to 29% after 14 days of unloading, impacting force transfer and thereby rehabilitation processes and athletic performance. While automated ultrasound (US) analysis has been applied to the Achilles tendon, PT elongation—an indirect marker of stiffness—still relies on manual and time-consuming frame-by-frame annotation. We present deepPatella, a novel, automated, transformer-based, open-source software with Kalman (KF) filtering to efficiently (and objectively) track PT elongation from US frame sequences. PT elongation was assessed using US (ArtUS, Telemed) with a 6 cm linear probe (LF9-5N60-A3 @7 MHz) fixed over the PT apex and tibial tuberosity during ramped isometric knee extensions. US images were recorded at 50 fps. Elongation was defined as relative displacement of proximal and distal insertions, manually labelled by two raters. The dataset included 17.475 frames from 30 participants (13 males, age=46.5 years (35-56), weight=73.7 kg (53.1-95.1), height=171.3 cm (158.0-188.0)), split into 12.221/5.254 frames (or 27/14 videos) for training/testing. A TransUNet model with sigmoid-based heatmaps, BCE-Dice loss, and ADAM optimizer predicted insertion points; predictions were subsequently smoothed using a Kalman filter. Accuracy was evaluated via Euclidean distance to manual labels and by comparing predicted and Kalman filtered PT elongation against the manually labelled one. In the 14 test videos, PT elongation was 4.8±1.5 mm based on manual annotation, 4.6±2.8 mm from the model prediction, and 4.6 ±2.9 mm with KF. Mean Euclidean errors compared to manual labels were 0.9±0.5 mm for the distal and 1.0±0.4 mm for the proximal insertions, for both model and Kalman. Accordingly, total absolute elongation Euclidean errors were 1.0±0.9 mm for the model and 1.0±1.8 mm for Kalman. Overall, the model predicted PT insertions with high accuracy and KF smoothed but did not improve the predictions. Additionally, using deepPatella, the processing time of per test video was reduced by 90% from 30 minutes (manual labelling) to 3 minutes. Thus, deepPatella enables time-efficient and objective PT elongation analysis in US videos and provides the basis for automated stiffness estimation. The openly available deepPatella software further enables users to compute tendon elongation and stiffness on their own data.

BACKGROUND: Ultrasonography is widely used to assess skeletal muscle and tendon properties, such as architecture, cross-sectional area, and tissue stiffness. Despite its growing application in different scenarios, the lack of accessible and standardized public datasets limits large-scale studies and the development of image analysis algorithms. To address this, we developed the Universal Musculoskeletal Ultrasonography Database (UMUD), a platform designed to facilitate access to these datasets and foster standardization in musculoskeletal ultrasonography imaging research. UMUD is an online repository that aggregates and indexes metadata from publicly available musculoskeletal ultrasonography datasets hosted on platforms like the Open Science Framework and Zenodo. By offering detailed metadata descriptors—such as muscle group, ultrasound device, participant demographics, and publication details—UMUD streamlines dataset discovery and exploration through search and visualization tools. RESULTS: So far, UMUD includes 14 Datasets, including 76.124 images (and/or 2.674 videos) derived from 1.901 participants. The platform also includes benchmark datasets for training and validating image analysis algorithms. These comprise multi-expert analyses of muscle architecture and panoramic cross-sectional area images, a dataset containing overlays of muscle geometry for teaching muscle architecture analysis, and labeled datasets for training deep learning models. Additionally, UMUD lists state-of-the-art automated analysis algorithms to support users in their application. CONCLUSION: UMUD addresses relevant challenges in musculoskeletal ultrasonography by providing a centralized, standardized repository of datasets and tools. Thus, it promotes transparency and innovation in the field, supporting reproducible research and advancements in automated image analysis. Future developments include adding more datasets, refining user functionalities, and introducing community-driven challenges to enhance its impact.

Due to the increasing amount of research data, open and reproducible data practices will define the future of medical research. Despite strong conceptual frameworks such as FAIR (Findable, Accessible, Interoperable, Reusable), implementation across laboratories remains inconsistent and fragmented. The MoveD (Haas et al., 2024) initiative outlines open, FAIR, and reproducible data practices in human movement research. However, putting such frameworks into operation remains challenging, especially in imaging-based disciplines, where data are heterogeneous and often under proprietary ecosystems. The Open and Reproducible Musculoskeletal Imaging Research (ORMIR) community and the Universal Musculoskeletal Ultrasonography Database (UMUD) address this issue by implementing reproducible data sharing in musculoskeletal imaging (Bonaretti et al., 2025; Ritsche et al., 2025). Here, we demonstrate implementation of guideline-based, reproducible data practices showcasing the ORMIR Medical Imaging Data Structure (MIDS) and UMUD. One of ORMIR’s aims is to create technical standards for organizing and sharing musculoskeletal imaging data across modalities and disciplines such as MRI, CT, and ultrasound. ORMIR-MIDS defines a transparent folder hierarchy, file-naming conventions, and a metadata scheme that ensure imaging datasets are both human-readable and machine-interpretable. A publicly available Python implementation enables validation, anonymization, and bidirectional conversion between ORMIR-MIDS and common vendor formats, lowering the technical barriers to adoption. Importantly, ORMIR-MIDS is not only a data structure but also a mechanism for community convergence—encouraging laboratories to align their workflows with shared standards that support cross-study comparability, reproducibility, and long-term reusability. UMUD offers an example of how such standards can be applied in ultrasonography research. It provides an openly available webapp that indexes metadata from publicly available musculoskeletal ultrasound datasets hosted on open platforms such as Zenodo. Using a standardized metadata scheme derived from ORMIR and MoveD principles, UMUD enhances the discoverability and comparability of datasets, linking them through descriptors based on expert consensus. By providing contribution and data structuring instructions, UMUD allows the research community to easily contribute their data. UMUD demonstrates how open research data can actively support both methodological innovation and capacity building in the research community. Together, ORMIR and UMUD embody the transition from conceptual to actionable open science. They illustrate how principles articulated in MoveD and FAIR can be realized through interoperable standards, shared infrastructure, and collaboration between technical and applied domains. References Bonaretti, S., Barzegari, M., Bevers, M., Boyd, S., Burghardt, A. J., Cameron, D., Chiumento, F., Crimi, G., Degenhart, G., Durongbhan, P., Hernandez, M. A. E., Fraterrigo, G., Ghasem-Zadeh, A., Grassi, L., Hirvasniemi, J., Hosseinitabatabaei, M., Iori, G., Kok, J., Kuczynski, M., … Zukic, D. (2025). Open and reproducible research in musculoskeletal imaging: Why it matters and how to implement it with the guidelines of the ORMIR community. https://doi.org/10.5281/zenodo.17258830 Haas, M. C., Sommer, B. B., Van Rekum, S., Moerman, F., & Graf, E. S. (2024, November). MoveD - open research data guidelines for movement laboratories. Zenodo. https://doi.org/10.5281/zenodo.14179954 Ritsche, P., Sarto, F., Santini, F., Leitner, C., Franchi, M., Faude, O., Finni, T., Seynnes, O., & Cronin, N. (2025). UMUD: A Web Application for Easy Access to Musculoskeletal Ultrasonography Datasets. Open Science Framework. https://doi.org/10.31219/osf.io/syr4z

Abstract Background This cross-sectional study examined whether exposure to long-term resistance and endurance training can counteract muscular weakness on a functional, neurological and structural level in adolescents with cerebral palsy (CP) compared to typically-developed peers (TD) depending on training status. Methods Five trained (4 males; mean age: 19.8) and four untrained adolescents with CP (3 males; 20.2) were compared to nine age- and sex-matched TD trained (7 males; 19.8) and nine untrained peers (7 males; 20.3). Isometric and isokinetic measurements assessed strength in knee flexion and extension, voluntary activation (VA) was assessed using the twitch interpolation technique and ultrasound imaging of the quadriceps was performed to assess anatomical cross-sectional area (ACSA) and architecture. Results Linear regression models revealed that CP trained had lower absolute isometric strength (dominant: -18% [-48; 11]; non-dominant: -35% [-58; -11]) than TD untrained while CP untrained showed between 29% and 33% lower strength than TD untrained. VA in CP trained (dominant: -13% [-23; -3]; non-dominant: -10% [-30; 11]) and CP untrained (dominant: -14% [-23; -4]; non-dominant: -8% [-29; 13]) showed similar deficits compared to TD untrained. CP trained showed higher ACSA than TD untrained in the dominant leg of the vastus lateralis muscle (+ 16% [-7; 38]), while the non-dominant side showed lower values (-18% [-45; 9]). Conclusion Exposure to long-term resistance and endurance training is associated with a reduced gap in muscle strength and muscle volume in the dominant leg of adolescents with CP while neural drive does not seem to be affected through training exposure. It is discussed that training load might have been too low in the non-dominant leg of CP trained to induce relevant neuromuscular adaptations. Trial registration ClinicalTrials.gov Identifier NCT05859360, date of registration May 16, 2023.

Low-load blood-flow restriction (BFR) training is a potential alternative to high-load (HL) resistance training, especially when mechanical stress must be minimized. However, its effects on neuromuscular activation remain unclear. This randomized controlled trial compared changes in voluntary activation (VA) and neuromuscular performance following 8 weeks of BFR versus HL knee extensor training and examined the effects of a subsequent 2-week HL phase in both groups. Thirty-seven healthy adults (37-59 years, 22 female) underwent progressive BFR or HL training for 8 weeks (phase 1), followed by 2 weeks of HL training on knee extensor muscles (phase 2). Outcomes included VA, maximal isometric and dynamic leg extension and leg press strength, rate of force development (RFD), and jump performance. Linear mixed models were used to analyze group*time interactions; Cohen's d effect sizes are reported. After both training phases, the BFR group showed smaller improvements than HL in VA (d = -0.31 to -0.37), maximal isometric strength (d = -0.07 to -0.27), dynamic strength (d = -0.18 to -0.75), and RFD (d = -0.48 to -0.54). Jump performance showed trivial between-group differences (d = -0.01 to -0.05). Although a subsequent 2-week HL phase improved outcomes in the BFR group, it did not fully restore neural adaptations to the level of continuous HL training. These findings underscore the essential role of mechanical loading in optimizing neuromuscular function. While BFR may serve as a useful preparatory method in contexts where high loads are initially contraindicated, follow-up HL training is required to maximize neuromuscular adaptation. Trial Registration: This study was preregistered on the Open Science Framework (DOI: 10.17605/OSF.IO/DA6SV).

Open science is transforming biomedical sciences by fostering transparency, reproducibility, and data reuse across laboratories and disciplines. Yet, despite broad consensus on its importance, the community still lacks the infrastructures, standards, and practical guidance needed to make openness routine. This symposium presents three initiatives that address this gap, from identifying community needs to implementing sustainable solutions. The first presentation (Ariana Ortigas-Vásquez) summarises results from the Internation Society of Biomechanics 2025 data-sharing survey, revealing strong motivation for openness but persistent barriers related to metadata standards, legal uncertainty, and methodological inconsistency. These findings define the community’s urgent priorities. Building on this evidence, the second talk (Michelle Haas) introduces the MoveD guidelines, an empirically grounded framework for FAIR-compliant data sharing in Swiss movement laboratories. MoveD was developed through nationwide surveys and expert workshops to translate open-science ideals into feasible laboratory practice. The final presentation (Paul Ritsche) demonstrates how such frameworks can be realised in imaging research through the ORMIR and UMUD initiatives, which establish interoperable standards for musculoskeletal imaging and open ultrasonography databases. Together, these talks show a coherent path from community insights to actionable infrastructures by illustrating how shared standards can strengthen reproducible, integrative research in sport and health.

The Open and Reproducible Musculoskeletal Imaging Research community is a scientific community dedicated to promoting openness and reproducibility in musculoskeletal imaging, image processing, and computational modeling. In this perspective paper, we outline the motivations for conducting transparent research and provide practical guidelines for implementing it. We start by defining open and reproducible research and describing the benefits and challenges of working transparently. Next, we redefine the outputs of a computational research study as-ideally-a combination of data, code, and a publication, recommend a folder and file structure that reflects these three study outcomes, and describe how to maintain and update such a structure during the study and at study publication. Finally, we emphasize that working in an open and reproducible manner is a learning process, and the best way to acquire the necessary competencies is simply to start.

While several openly available tools for the automatic segmentation of the anatomical cross-sectional area (ACSA) of muscle exist, there is no open-source and peer-reviewed tool for the patellar tendon. In this study, we tested an automatic approach for the segmentation of the patellar tendon ACSA in ultrasound images. Images were acquired at 25

Introduction Muscle volume is a crucial indicator of muscle strength and neuromuscular health, traditionally assessed using Magnetic Resonance Imaging (MRI). While MRI provides high accuracy, it is costly and time-consuming. Three-dimensional ultrasonography (3DUS) offers a more accessible alternative but requires rigorous validation due to setup dependent accuracy. This study evaluates the validity and reliability of a custom 3DUS setup for measuring lower limb muscle volumes, compares it to MRI, and documents the methodology for broader research adoption. Methods We recruited 10 healthy participants (5 males, 5 females; 18–40 years). The protocol included two 3DUS sessions and one MRI session spaced one week apart. The 3DUS scans were acquired by one experienced operator, while two operators analyzed the captured volumes from the second scanning timepoint. Target muscles—biceps femoris, gastrocnemius medialis and lateralis, tibialis anterior, and vastus lateralis—were scanned using a linear probe (12L3, Siemens, Erlangen, Germany) integrated with a motion capture system (Optitrack Flex 3, NaturalPoint, Corvallis, OR, USA). Data synchronization and reconstruction were performed using 3D Slicer. MRI scans were acquired using a 3T whole-body scanner (MAGNETOM Prisma, Siemens, Erlangen, Germany), and muscle volumes were segmented using Dafne and 3D-Slicer software. Validation was enhanced with phantom models (100–600 mL) scanned via both modalities. Reliability metrics included intra-class correlation (ICC), coefficients of variation (CV%), standard error of measurement (SEM), minimal detectable change (MDC) and standardized mean bias. Results Test-retest reliability of 3DUS was high for all muscles, with ICCs ranging from 0.97 to 0.99 (95% confidence intervals between 0.86–0.99) and CV% values between 2.0 and4.6% (95% CI between 1.3–6.7%). MDC values were below 5 mL for all muscles, indicating sensitivity to detect small volume changes. The biceps femoris exhibited the highest test-retest reliability, while the gastrocnemius medialis showed lower but acceptable agreement. Inter-rater reliability was similarly high, with ICCs exceeding 0.97 (95% CIs between 0.51–0.99) for all assessed muscles and minimal variability between operators (CV%: 2.1–3.2%, 95% CIs between 1.1–4.7%). Tibialis anterior yielded the highest inter-rater reliability results, vastus lateralis the lowest. Compared to MRI, 3DUS underestimated muscle volumes, with mean biases ranging from -15.4% to -44.6%. Agreement was highest for the tibialis anterior and lowest for the gastrocnemii. Standardized mean biases were classified as moderate to very large, ranging from -0.92 to -2.64. In contrast, phantom scans confirmed high accuracy for both 3DUS and MRI, suggesting measurement errors in vivo were likely due to probe pressure and sweep inconsistencies. Discussion/Conclusion Our 3DUS system demonstrated excellent reliability for assessing muscle volumes between sessions and raters, but showed limited comparability to MRI, systematically underestimating volumes. The magnitude of underestimation appeared to be influenced by muscle architecture and location. Calibration refinements, including reduced probe pressure and optimized sweep protocols, are underway to improve accuracy. Nonetheless, 3DUS is a promising, cost-effective tool for longitudinal muscle assessments, particularly for detecting intervention-induced changes. Our openly accessible documentation of the methodology supports reproducibility and further innovation in the field.

Introduction Low-load Blood Flow Restriction training (BFRT) is a viable alternative to high-load resistance training (HLT), especially for clinical populations, due to its lower mechanical tension (Centner et al., 2019). Despite using lower loads, BFRT achieves comparable or slightly lower gains in muscle hypertrophy and strength than HLT. However, neural adaptations after low-load BFRT remain poorly understood and are hypothesized to be less pronounced (Centner & Lauber, 2020). Thus, this study aimed to compare the effects of BFRT and HLT, as well as the sequential application of HLT following BFRT, to address potential gaps in neuromuscular adaptation (Duchateau et al., 2021). Methods In this 10-week randomized controlled trial, 37 healthy male and female adults (35-60 years) completed progressive resistance training for knee extensors either 8 weeks with low-load BFRT followed by 2 weeks of HLT, or exclusively HLT. Measurements at baseline including familiarization (PRE1 and PRE2), 8 weeks (MID) and 10 weeks (POST) included maximal voluntary contraction of knee extensors (MVC), leg press strength (LP), voluntary activation (VA, via electrical muscle stimulation) and vastus lateralis muscle volume (V_VL, via ultrasound at 30-70% femur length). Linear mixed models were used to analyze group differences over time, with Cohen’s d effect sizes. Here, we report the preliminary data of 20 participants. Results Reliability between PRE1 and PRE2 was excellent for MVC and V_VL (ICC>0.98) and good for LP (ICC=0.88). From PRE2 to POST, MVC showed a moderate overall increase (+14.7%, d=0.49), while from MID to POST the BFRT group benefited slightly more (d=-0.14). LP increased similarly in both groups at MID (+5.9), but at POST, the HLT group demonstrated a greater increase compared to the BFRT group (+14.8%, d=0.32). No meaningful differences in VA changes were observed between groups or across timepoint, however, the interaction at POST favored more change in BFRT compared to HLT (d=-0.3). Both groups showed comparable increases in V_VL, with negligible differences between groups (+3.4%, d=0.05). Discussion/Conclusion The results of the analyzed sub-sample imply that BFRT and HLT lead to similar changes in maximal force, neural adaptation, and muscle size, however, a large variability was found. The greater increase in VA and MVC in the BFRT group at POST suggests that sequential HLT training after BFRT may enhance performance more than HLT alone (Duchateau et al., 2021). The analysis of the remaining participants may help clarify these effects and enlighten, how mechanical tension and metabolic stress contribute to neural adaptations. References Centner, C., & Lauber, B. (2020). A systematic review and meta-analysis on neural adaptations following blood flow restriction training: What we know and what we don’t know. Frontiers in Physiology, 11, Article 887. https://doi.org/10.3389/fphys.2020.00887 Centner, C., Wiegel, P., Gollhofer, A., & König, D. (2019). Effects of blood flow restriction training on muscular strength and hypertrophy in older individuals: A systematic review and meta-analysis. Sports Medicine, 49(1), 95–108. https://doi.org/10.1007/s40279-018-0994-1 Duchateau, J., Stragier, S., Baudry, S., & Carpentier, A. (2021). Strength training: In search of optimal strategies to maximize neuromuscular performance. Exercise and Sport Sciences Reviews, 49(1), 2. https://doi.org/10.1249/JES.0000000000000234

Objective: Deep learning approaches such as DeepACSA enable automated segmentation of muscle ultrasound cross-sectional area (CSA). Although they provide fast and accurate results, most are developed using data from healthy populations. The changes in muscle size and quality following anterior cruciate ligament (ACL) injury challenges the validity of these automated approaches in the ACL population. Quadriceps muscle CSA is an important outcome following ACL injury; therefore, our aim was to validate DeepACSA, a convolutional neural network (CNN) approach for ACL injury. Methods: Quadriceps panoramic CSA ultrasound images (vastus lateralis [VL] n = 430, rectus femoris [RF] n = 349, and vastus medialis [VM] n = 723) from 124 participants with an ACL injury (age 22.8 +/- 7.9 y, 61 females) were used to train CNN models. For VL and RF, combined models included extra images from healthy participants (n = 153, age 38.2, range 13-78) that the DeepACSA was developed from. All models were tested on unseen external validation images (n = 100) from ACL-injured participants. Model predicted CSA results were compared to manual segmentation results. Results: All models showed good comparability (ICC > 0.81, < 14.1% standard error of measurement, mean differences of <1.56 cm(2)) to manual segmentation. Removal of the erroneous predictions resulted in excellent comparability (ICC > 0.94, < 7.40% standard error of measurement, mean differences of <0.57 cm(2)). Erroneous predictions were 17% for combined VL, 11% for combined RF, and 20% for ACL-only VM models. Conclusion: The new CNN models provided can be used in ACL-injured populations to measure CSA of VL, RF, and VM muscles automatically. The models yield high comparability to manual segmentation results and reduce the burden of manual segmentation.

Muscle volume is a key indicator of strength and neuromuscular health, commonly assessed via Magnetic Resonance Imaging (MRI). While accurate, MRI is expensive and time-intensive. Three-dimensional ultrasonography (3DUS) offers a more accessible alternative but requires validation due to its setup-dependent accuracy. This study investigated the validity and reliability of a custom 3DUS setup for measuring lower limb muscle volumes. Fifteen participants (8 female; 18–40 years) underwent two 3DUS and one MRI sessions. The tibialis anterior, vastus lateralis, gastrocnemii, and biceps femoris muscles were scanned using ultrasonography integrated with a motion capture system. Phantom models were also scanned. After ten participants, the scanning protocol was adapted. 3DUS and MRI volumes were analyzed using 3D Slicer by two raters or one rater, respectively. Reliability was assessed using intra-class correlation (ICC), coefficient of variation (CV

High-Density surface Electromyography (HD-sEMG) is the most established technique for the non-invasive analysis of single motor unit (MU) activity in humans. It provides the possibility to study the central properties (e.g., discharge rate) of large populations of MUs by analysis of their firing pattern. Additionally, by spike-triggered averaging, peripheral properties such as MUs conduction velocity can be estimated over adjacent regions of the muscles and single MUs can be tracked across different recording sessions. In this tutorial, we guide the reader through the investigation of MUs properties from decomposed HD-sEMG recordings by providing both the theoretical knowledge and practical tools necessary to perform the analyses. The practical application of this tutorial is based on openhdemg, a free and open-source community-based framework for the automated analysis of MUs properties built on Python 3 and composed of different modules for HD-sEMG data handling, visualisation, editing, and analysis. openhdemg is interfaceable with most of the available recording software, equipment or decomposition techniques, and all the built-in functions are easily adaptable to different experimental needs. The framework also includes a graphical user interface which enables users with limited coding skills to perform a robust and reliable analysis of MUs properties without coding.

Introduction Lower limb muscle strength is an important predictor of sports performance, injury risk and frailty in ageing. The strength of a muscle is determined by its geometry and neuronal factors. Muscle geometry can be subdivided into architecture and morphology. Muscle morphology describes shape characteristics such as anatomical cross-sectional area (ACSA), thickness or volume (Maden-Wilkinson et al., 2021). Muscle architecture is determined by muscle fascicle length and the insertion angle of the muscle fascicles in the aponeuroses and describes the orientation of the muscle fibers relative to their force generation axis (Lieber & Friden, 2000). Muscle geometry is associated to physical performance and strength in humans (Maden-Wilkinson et al., 2021; Werkhausen et al., 2022) and is therefore a main research interest. A cost-effective and participant friendly method to validly and reliably assess muscle geometry is ultrasonography. However, a major limitation of ultrasonography is the subjectivity of image acquisition and the time-consuming image analysis (Ritsche et al., 2021; Ritsche, Wirth, et al., 2022; Ritsche et al., 2023). Moreover, image characteristics are massively influenced by the ultrasonography device used (Ritsche, Schmid, et al., 2022) as well as the muscle region scanned (Monte & Franchi, 2023). This poses constraints on the generalizability of existing automated image analysis approaches. The goal of this series of studies is therefore to optimize the ultrasonography acquisition and data analysis procedures by developing open-source software packages. Secondly, we aim to apply these methods in a sports performance context and describe the relevance of muscle geometry. Methods To streamline the time-consuming and subjective process of image analysis, we developed open-source and user-friendly software packages for muscle geometry analysis in lower limb muscles. We developed a semi-automated algorithm “ACSAuto” for assisted analysis of muscle ACSA using common image filtering processes (Ritsche et al., 2021). Given the limited generalizability and required user input of this approach, we developed two fully automated software applications, “DeepACSA” and “DL_Track_US”, using convolutional neural networks for more time efficient and robust analysis of lower limb muscle geometry (Ritsche et al., 2023; Ritsche, Wirth, et al., 2022; Ritsche et al., in press). We compared the predictions in an unseen test set to the current state-of-the-art, manual analysis, in order to evaluate the performance of our algorithms. To broaden the application of ultrasonography for evaluating muscle geometry in a sports context, we investigated the validity of a low-cost mobile ultrasonography device compared to a high-end counterpart in assessing various muscle architectural parameters in healthy adults (Ritsche, Schmid, et al., 2022).The mobile ultrasonography setup consisted of a smartphone and a portable probe, enabling practitioners high flexibility in the assessment of muscle architecture. We further investigated the link between muscle geometry and performance among soccer players. In one study, we focused on the m. biceps femoris long head in under-13 to under-15 youth players, assessing architecture and morphology at the mid-muscle point and correlating these with their sprint times and maximum velocity (Ritsche et al., 2020). In a further study, we analyzed the mm. vastus lateralis and rectus femoris in both youth and adult players of both sexes, evaluating muscle geometry at various muscle lengths alongside their knee extension strength during isometric and isokinetic conditions (Ritsche et al., in preparation and under review). Results Both ACSAuto and DeepACSA showed high comparability in assessing lower limb muscle ACSA with standard error of measurement lower than one cm2 (SEM ranging from 1.2 to 9.5%; Ritsche et al., 2021; Ritsche, Wirth, et al., 2022). Moreover, DeepACSA provided fast and objective analysis comparable to manual segmentation with no supervision of the analysis process needed. The time needed for analysis was reduced by a factor of 10. DL_Track_US demonstrated high comparability to manual muscle architecture analysis of images and videos, i.e. dynamic situations, (Ritsche et al., 2023; Ritsche et al., in press) and a reduction in the duration of analysis by a factor of 100. The mobile ultrasonography system showed a high degree of reliability and comparability only for m. gastrocnemius medialis architecture assessment, with a standard error of measurement lower than 10% for all architectural parameters (Ritsche, Schmid, et al., 2022). Thus, its reliability and comparability depended on the muscle assessed. We observed relevant correlations between muscle ACSA in young and adult male soccer players as well as in female soccer players and performance (Ritsche et al., 2020; Ritsche et al., unpublished). Moreover, we observed changes in muscle geometry with age and differences between males and females. Specifically, m. biceps femoris ACSA was strongly correlated with 30m sprint times and maximal velocity (r = -0.61 and r = 0.61, respectively), highlighting its importance in athletic performance (Ritsche et al., 2020). M. vastus lateralis ACSA at 50% of muscle length was most frequently related to knee extension strength (r = 0.40 - 0.53), which was observed in both sexes and across several age groups of male soccer players (Ritsche et al., in preparation and under review). Relevant correlations occurred more frequently in older age groups and higher knee extension velocities. Interestingly, we did not observe relevant correlations between muscle architecture and performance in the mm. biceps femoris and vastus lateralis. Discussion/Conclusion The results of this series studies so far led to three main insights. Firstly, the development of the “ACSAuto”, “DeepACSA” and “DL_Track_US” tools, utilizing semi-automated and fully automated analysis techniques applying deep learning algorithms, marked another step forward in overcoming the subjectivity and time consuming image evaluation. In a user-friendly way, these tools enable reproducible and objective analyses of muscle geometry in ultrasonography images. Secondly, with technological advancements, assessing muscle geometry with ultrasonography is possible using a smartphone and a probe, and often gives comparable results to high-end devices (Ritsche, Schmid, et al., 2022). This allows for a broader and more versatile application of muscle geometry assessment. However, our results highlight the need for a selective approach based on the muscle group being assessed and technical improvements of existing devices. Lastly, our findings across several investigations reveal a relevant positive correlation between muscle ACSA and performance metrics such as sprint times and knee extension strength (Ritsche et al., 2020; Ritsche et al., unpublished), corroborating previous research (Maden-Wilkinson et al., 2021; Monte & Franchi, 2023). The relationship was more pronounced in older age groups, suggesting that muscle geometry's influence on performance may amplify with athletic maturity. Apart from that, we observed the relationship in the m. vastus lateralis to be region- and contraction velocity-dependent. In agreement with Werkhausen et al. (2022), no relation of muscle architecture with strength when assessed in a static resting position was observed. This highlights the need for a potential shift towards assessing changes in muscle geometry during contraction rather than in static situations when evaluating the relation between muscle geometry and performance. Finally, remaining challenges include the comparability of muscle geometry assessment in the literature, the analysis methods used and the low generalizability of available automated analysis approaches (ours included). There is a clear need for methodological consensus on the assessment of muscle geometry when using ultrasonography, and more versatile analysis approaches are needed to enable an easy, generalizable and reproducible analysis of images and videos. Therefore, future works should target to establish assessment and analysis guidelines of muscle geometry in ultrasonography images to increase the comparability and reproducibility of results. Moreover, assessing changes in muscle geometry during contraction rather than during rest should be focused. References Lieber, R. L., & Friden, J. (2000). Functional and clinical significance of skeletal muscle architecture. Muscle Nerve, 23(11), 1647–1666. https://doi.org/10.1002/1097-4598(200011)23:11%3C1647::aid-mus1%3E3.0.co;2-m Maden-Wilkinson, T. M., Balshaw, T. G., Massey, G. J., & Folland, J. P. (2021). Muscle architecture and morphology as determinants of explosive strength. European Journal of Applied Physiology, 121(4), 1099–1110. https://doi.org/10.1007/s00421-020-04585-1 Monte, A., & Franchi, M. V. (2023). Regional muscle features and their association with knee extensors force production at a single joint angle. European Journal of Applied Physiology, 123, 2239-2248. https://doi.org/10.1007/s00421-023-05237-w Ritsche, P., Bernhard, T., Roth, R., Lichtenstein, E., Keller, M., Zingg, S., Franchi, M. V., & Faude, O. (2020). M. biceps femoris long head architecture and sprint ability in youth soccer players. International Journal of Sports Physiology and Performance, 16(11), 1616-1624. https://doi.org/10.1123/ijspp.2020-0726 Ritsche, P., Schmid, R., Franchi, M. V., & Faude, O. (2022). Agreement and reliability of lower limb muscle architecture measurements using a portable ultrasound device. Frontiers in Physiology, 13, Article 981862. https://doi.org/10.3389/fphys.2022.981862 Ritsche, P., Seynnes, O., & Cronin, N. (2023). DL_Track_US: A python package to analyse muscleultrasonography images. Journal of Open Source Software, 8(85), Article 5206. https://doi.org/10.21105/joss.05206 Ritsche, P., Wirth, P., Cronin, N. J., Sarto, F., Narici, M. V., Faude, O., & Franchi, M. V. (2022). DeepACSA: Automatic segmentation of cross-sectional area in ultrasound images of lower limb muscles using deep learning. Medicine & Science in Sports & Exercise, 54(12), 2188-2195. https://doi.org/10.1249/MSS.0000000000003010 Ritsche, P., Wirth, P., Franchi, M. V., & Faude, O. (2021). ACSAuto-semi-automatic assessment of human vastus lateralis and rectus femoris cross-sectional area in ultrasound images. Scientific Reports, 11, Article 13042. https://doi.org/10.1038/s41598-021-92387-6 Werkhausen, A., Gløersen, Ø., Nordez, A., Paulsen, G., Bojsen-Møller, J., & Seynnes, O. R. (2022). Rate of force development relationships to muscle architecture and contractile behavior in the human vastus lateralis. Scientific Reports, 12, Article 21816. https://doi.org/10.1038/s41598-022-26379-5

Introduction Many climbers believe that they are stronger in crimp finger position than in open hand position. However, compared to open hand, crimped fingers are associated with higher pulley forces increasing the risk of injuries. Climbing expertise may influence the estimation of strength, i.e., the better the climber, the better the self-assessment. This study therefore aimed to find out whether climbing expertise influences self-assessment of finger flexor strength in half-crimp and in open hand position. Methods Data was collected at the Hands-On Science Booth of the Climbing World Championships in Bern. Participants had to fill out a questionnaire including a self-assessment of their climbing expertise and of their maximum finger strength for both hands as well as both finger positions. Afterwards, maximum finger strength was measured on an instrumented campus board: Participants placed the to-be-measured hand on a self-selected rung (depth of 23 mm) and then tried to transfer as much force as possible from their feet to their fingers. Results The analysis was based on 38 intermediate and 36 advanced climbers. Due to the limited number of participants in the lower grade (n = 0) and elite (n = 2) level, those skill levels were not considered. Advanced climbers generated significantly greater forces than intermediate climbers across all four measured conditions (t-tests, all p < 0.01). For both groups, neither in the dominant nor in the non-dominant hand a significant difference in maximum force was observed, e.g., dominant hand, intermediates: or advanced climbers: . Intermediate climbers did neither over- nor underestimate their strength in half-crimp position compared to open hand (paired t-test, p = 0.91 for dominant, p = 0.077 for non-dominant hand). In contrast to the dominant hand, advanced climbers significantly overestimated their strength in half-crimp position for the non-dominant hand (on average 9%, Cohen’s d 0.64, p < 0.01). Discussion/Conclusion Our results confirm the positive correlation between finger strength and climbing level. We also confirm that on a 23 mm rung, greater forces can be generated with in open hand compared to half-crimp (Winkler et al., 2023). With larger hold depths, force generated in open hand significantly increases (Amca et al., 2012), while for smaller holds, force exerted in half-crimp position exceeds that of open hand (Winkler et al., 2023). Hence, at least for larger holds, we recommend adopting an open hand position as preventive measure against finger injuries. Advanced climbers may tend to inaccurately self-assess their strength due to their greater engagement with peers, potentially leading to the circulation of misinformation. Note that participants were instructed to provide a general self-assessment of their strength rather than for a 23 mm deep rung, i.e., they may have had a smaller hold in mind. References Amca, A. M., Vigouroux, L., Aritan, S., & Berton, E. (2012). Effect of hold depth and grip technique on maximal finger forces in rock climbing. Journal of Sports Sciences, 30(7), 669-677. https://doi.org/10.1080/02640414.2012.658845 Winkler, M., Künzell, S., & Auguste, C. (2023). Competitive performance predictors in speed climbing, bouldering, and lead climbing. Journal of Sport Sciences, 41(8), 736-746. https://doi.org/10.1080/02640414.2023.2239598

Purpose:Adolescent soccer players experience distinct physiological changes due to chronological and biological maturation, impacting their soccer performance. Here, we explored age-related variations and associations between quadriceps geometry and strength in male national-level adolescent soccer players. Patients and Methods:We used ultrasonography to examine the regional architecture and morphology of the rectus femoris (RF) and vastus lateralis (VL) muscles, and we assessed knee extension strength by isometric and isokinetic dynamometry. Players were categorized into four age groups: under (U) 15 (n=18, age=13.7±0.5 years), U16 (n=15, age=14.7±0.5), U17 (n=19, age=15.7±0.5), U18 (n=18, age=16.7±0.5) and U21 (n=25, age=18.5±0.5). Results:The absolute and relative strengths were higher in the U16 compared to U15 by 12-15% and 6-8%, 11-12% and 6-7% in the U17 compared to U16, 5-7% and -1-2% in the U18 compared to U17 and 0-15% and -1-11% in the U21 compared to U18 age groups, respectively. VL architecture did not change relevantly between the age groups. The muscle anatomical cross-sectional area (ACSA) of the VL and RF differed non-uniformly and muscle region-specific by 10-36%, with highest values in the U21 age group. Moderate correlations between the VL architecture and knee extension strength in both legs were observed only in the U16 age group. The quadriceps ACSA showed age-specific correlations with knee extension strength. Conclusion:Our findings highlight non-uniform differences in quadriceps muscle morphology and absolute and relative strength among male national-level adolescent soccer players in different age groups. The correlations observed between muscle morphology or architecture and strength were muscle, muscle region, leg and age dependent.

Ultrasonography can be used to assess muscle architectural parameters during static and dynamic conditions.Nevertheless, the analysis of the acquired ultrasonography images presents a major difficulty.Muscle architectural parameters such as muscle thickness, fascicle length and pennation angle are mainly segmented manually.Manual analysis is time expensive, subjective and requires thorough expertise.Within recent years, several algorithms were developed to solve these issues.Yet, these are only partly automated, are not openly available, or lack in user friendliness.The DL_Track_US python package is designed to allow fully automated and rapid analysis of muscle architectural parameters in lower limb ultrasonography images.

B-mode ultrasound is commonly used to image musculoskeletal tissues, but one major bottleneck is data analysis. Manual analysis is commonly deployed for assessment of muscle thickness, pennation angle and fascicle length in muscle ultrasonography images. However, manual analysis is somewhat subjective, laborious and requires thorough experience. We provide an openly available algorithm (DL_Track) to automatically analyze muscle architectural parameters in ultrasonography images or videos of human lower limb muscles. We trained two different neural networks (classic U-net [Ronneberger et al., 2021] and U-net with VGG16 [Simonyan & Zisserman, 2015] pretrained encoder) one to detect muscle fascicles and another to detect muscle aponeuroses using a set of labelled musculoskeletal ultrasound images. We included images from four different devices of the vastus lateralis, gastrocnemius medialis, tibilias anterior and soleus. In total, we included 310 images for the fascicle model and 570 images for the aponeuroses model, which we augmented to about 1,700 images per set. Each dataset was randomly split into a training and test set for model training, using a common 80/20 train/test split. We determined the best performing model based on intersection-over-union and loss metrics calculated during model training. We compared neural network predictions on an unseen test set consisting of 35 images to those obtained via manual analysis and two existing semi/automated analysis approaches (SMA and Ultratrack). Across the set of 35 unseen images, the mean differences between DL_Track and manual analysis were for fascicle length -2.4 mm (95% compatibility interval (CI) = -3.7 to -1.2), for pennation angle 0.6° (-0.2 to 1.4), and for muscle thickness -0.6 mm (-1.2 to 0.002). The corresponding values comparing DL_Track with SMA were for fascicle length 5.2 mm (1.3 to 9.0), for pennation angle -1.4° (-2.6 to -0.4) and for muscle thickness -0.9 mm (-1.5 to -0.3) respectively. ICC values between DL_Track and Ultratrack were 0.19 (0.00 to 0.35) for medial gastrocnemius passive contraction, 0.79 (0.77 to 0.81) for medial gastrocnemius maximal voluntary contraction, 0.88 (0.87 to 0.89) for calf raise, 0.67 (0.07 to 0.86) for medial gastrocnemius during walking, 0.80 (0.79 to 0.82) for tibialis passive plantar and dorsiflexion, and 0.85 (0.83 to 0.86) for tibialis anterior maximum voluntary contraction. Our method is fully automated and can estimate fascicle length, pennation angle and muscle thickness from single images or videos in multiple superficial muscles. For single images, the method gave results that are in agreement with those produced by SMA or manual analysis. Similarly, for videos, there was overlap between the results produced with Ultratrack and our method. In contrast to Ultratrack, DL_Track analyzes each frame independently of the previous frames, which might explain the observerd variability. References Ronneberger, O., Fischer, P., & Brox, T. (2021). U-Net: Convolutional networks for biomedical image segmentation. arXiv. https://doi.org/10.48550/arXiv.1505.04597 Simonyan, K., & Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. arXiv. https://doi.org/10.48550/arXiv.1409.1556

B-mode ultrasound is commonly used to image musculoskeletal tissues, but one major bottleneck is data interpretation, and analyses of muscle thickness, pennation angle and fascicle length are often still performed manually. In this study we trained deep neural networks (based on U-net) to detect muscle fascicles and aponeuroses using a set of labelled musculoskeletal ultrasound images. We then compared neural network predictions on new, unseen images to those obtained via manual analysis and two existing semi/automated analysis approaches (SMA and Ultratrack). With a GPU, inference time for a single image with the new approach was around 0.7s, compared to 4.6s with a CPU. Our method detects the locations of the superficial and deep aponeuroses, as well as multiple fascicle fragments per image. For single images, the method gave similar results to those produced by a non-trainable automated method (SMA; mean difference in fascicle length: 1.1 mm) or human manual analysis (mean difference: 2.1 mm). Between-method differences in pennation angle were within 1$^\circ$, and mean differences in muscle thickness were less than 0.2 mm. Similarly, for videos, there was strong overlap between the results produced with Ultratrack and our method, with a mean ICC of 0.73, despite the fact that the analysed trials included hundreds of frames. Our method is fully automated and open source, and can estimate fascicle length, pennation angle and muscle thickness from single images or videos, as well as from multiple superficial muscles. We also provide all necessary code and training data for custom model development.