Activity cliff prediction - identifying positions where small structural changes cause large potency shifts - has been a persistent challenge in computational medicinal chemistry. This work focuses on a parsimonious definition: which small modifications, at which positions, confer the highest probability of an outcome change. Position-level sensitivity is calculated using 25 million matched molecular pairs from 50 ChEMBL targets across six protein families, revealing that two questions have fundamentally different answers. "Which positions vary most?" is answered by scaffold size alone (NDCG@3 = 0.966), requiring no machine learning. "Which are true activity cliffs?" - where small modifications cause disproportionately large effects, as captured by SALI normalization - requires an 11-feature model with 3D pharmacophore context (NDCG@3 = 0.910 vs. 0.839 random), generalizing across all six protein families, novel scaffolds (0.913), and temporal splits (0.878). The model identifies the cliff-prone position first 53% of the time (vs. 27% random - 2x lift), reducing positions a chemist must explore from 3.1 to 2.1 - a 31% reduction in first-round experiments. Predicting which modification to make is not tractable from structure alone (Spearman 0.268, collapsing to -0.31 on novel scaffolds). The system is released as open-source code and an interactive webapp.
ObjectiveThe predictive value of a nomogram model constructed by integrating radiomics features and clinical risk factors for the functional outcomes of patients after rotator cuff repair was evaluated.MethodsA total of 367 patients who underwent rotator cuff repair from January 2021 to December 2023 were selected. Pre - operative shoulder MRI images were collected and radiomics features were extracted, and clinical baseline data were also collected. The patients were randomly divided into a training set (n = 257) and a validation set (n = 110) at a ratio of 7:3. In the training set, univariate analysis was used to identify factors associated with postoperative functional outcomes, which were evaluated by the Constant-Murley score at 12 months after surgery and classified into good or poor categories. Least absolute shrinkage and selection operator (LASSO) regression was used for radiomics feature dimensionality reduction and variable screening, and then independent predictive factors were identified by multivariate Logistic regression. A nomogram model was established accordingly. The area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis (DCA) were used to evaluate the discrimination, calibration, and clinical utility of the model, respectively.ResultsMultivariate analysis showed that age, pre - operative visual analog scale score, tear area, tear maximum length, tendon retraction distance, standard deviation of gray - scale, and entropy of the gray - level co - occurrence matrix were independent predictive factors for poor postoperative functional outcomes in patients undergoing rotator cuff repair (P < 0.05). The AUCs of the nomogram model developed based on these factors were 0.817 (95% CI: 0.750–0.883) in the training set and 0.721 (95% CI: 0.600–0.843) in the validation set, respectively. The calibration curve showed good consistency between the predicted probability and the actual risk.ConclusionThe nomogram model integrating radiomics features and clinical factors has potential utility in predicting functional outcomes after rotator cuff repair, and may thus provide a valuable reference for clinical individualized treatment and prognosis assessment.
Aliaksandr Marozau, Barbara Karpowicz, Tomasz Kowalewski
et al.
Mixed Reality (MR) technologies such as Virtual and Augmented Reality (VR, AR) are well established in medical practice, enhancing diagnostics, treatment, and education. However, there are still some limitations and challenges that may be overcome thanks to the latest generations of equipment, software, and frameworks based on eXtended Reality (XR) by enabling immersive systems that support safer, more controlled environments for training and patient care. Our review highlights recent VR and AR applications in key areas of medicine. In medical education, these technologies provide realistic clinical simulations, improving skills and knowledge retention. In surgery, immersive tools enhance procedural precision with detailed anatomical visualizations. VR-based rehabilitation has shown effectiveness in restoring motor functions and balance, particularly for neurological patients. In mental health, VR has been successful in treating conditions like PTSD and phobias. Although VR and AR solutions are well established, there are still some important limitations, including high costs and limited tactile feedback, which may be overcome with implementing new technologies that may improve the effectiveness of immersive medical applications such as XR, psychophysiological feedback or integration of artificial intelligence (AI) for real-time data analysis and personalized healthcare and training.
Large language models (LLMs) have demonstrated exceptional capabilities in general domains, yet their application in highly specialized and culturally-rich fields like Traditional Chinese Medicine (TCM) requires rigorous and nuanced evaluation. Building upon prior foundational work such as TCM-3CEval, which highlighted systemic knowledge gaps and the importance of cultural-contextual alignment, we introduce TCM-5CEval, a more granular and comprehensive benchmark. TCM-5CEval is designed to assess LLMs across five critical dimensions: (1) Core Knowledge (TCM-Exam), (2) Classical Literacy (TCM-LitQA), (3) Clinical Decision-making (TCM-MRCD), (4) Chinese Materia Medica (TCM-CMM), and (5) Clinical Non-pharmacological Therapy (TCM-ClinNPT). We conducted a thorough evaluation of fifteen prominent LLMs, revealing significant performance disparities and identifying top-performing models like deepseek\_r1 and gemini\_2\_5\_pro. Our findings show that while models exhibit proficiency in recalling foundational knowledge, they struggle with the interpretative complexities of classical texts. Critically, permutation-based consistency testing reveals widespread fragilities in model inference. All evaluated models, including the highest-scoring ones, displayed a substantial performance degradation when faced with varied question option ordering, indicating a pervasive sensitivity to positional bias and a lack of robust understanding. TCM-5CEval not only provides a more detailed diagnostic tool for LLM capabilities in TCM but aldso exposes fundamental weaknesses in their reasoning stability. To promote further research and standardized comparison, TCM-5CEval has been uploaded to the Medbench platform, joining its predecessor in the "In-depth Challenge for Comprehensive TCM Abilities" special track.
In this article, we propose a novel approach for plant hierarchical taxonomy classification by posing the problem as an open class problem. It is observed that existing methods for medicinal plant classification often fail to perform hierarchical classification and accurately identifying unknown species, limiting their effectiveness in comprehensive plant taxonomy classification. Thus we address the problem of unknown species classification by assigning it best hierarchical labels. We propose a novel method, which integrates DenseNet121, Multi-Scale Self-Attention (MSSA) and cascaded classifiers for hierarchical classification. The approach systematically categorizes medicinal plants at multiple taxonomic levels, from phylum to species, ensuring detailed and precise classification. Using multi scale space attention, the model captures both local and global contextual information from the images, improving the distinction between similar species and the identification of new ones. It uses attention scores to focus on important features across multiple scales. The proposed method provides a solution for hierarchical classification, showcasing superior performance in identifying both known and unknown species. The model was tested on two state-of-art datasets with and without background artifacts and so that it can be deployed to tackle real word application. We used unknown species for testing our model. For unknown species the model achieved an average accuracy of 83.36%, 78.30%, 60.34% and 43.32% for predicting correct phylum, class, order and family respectively. Our proposed model size is almost four times less than the existing state of the art methods making it easily deploy able in real world application.
Luohong Wu, Nicola A. Cavalcanti, Matthias Seibold
et al.
Ultrasound-based bone surface segmentation is crucial in computer-assisted orthopedic surgery. However, ultrasound images have limitations, including a low signal-to-noise ratio, and acoustic shadowing, which make interpretation difficult. Existing deep learning models for bone segmentation rely primarily on costly manual labeling by experts, limiting dataset size and model generalizability. Additionally, the complexity of ultrasound physics and acoustic shadow makes the images difficult for humans to interpret, leading to incomplete labels in anechoic regions and limiting model performance. To advance ultrasound bone segmentation and establish effective model benchmarks, larger and higher-quality datasets are needed. We propose a methodology for collecting ex-vivo ultrasound datasets with automatically generated bone labels, including anechoic regions. The proposed labels are derived by accurately superimposing tracked bone CT models onto the tracked ultrasound images. These initial labels are refined to account for ultrasound physics. A clinical evaluation is conducted by an expert physician specialized on orthopedic sonography to assess the quality of the generated bone labels. A neural network for bone segmentation is trained on the collected dataset and its predictions are compared to expert manual labels, evaluating accuracy, completeness, and F1-score. We collected the largest known dataset of 100k ultrasound images of human lower limbs with bone labels, called UltraBones100k. A Wilcoxon signed-rank test with Bonferroni correction confirmed that the bone alignment after our method significantly improved the quality of bone labeling (p < 0.001). The model trained on UltraBones100k consistently outperforms manual labeling in all metrics, particularly in low-intensity regions (320% improvement in completeness at a distance threshold of 0.5 mm).
Abstract Background Regular walking has been reported to improve metabolically-associated steatotic liver disease (MASLD) by altering the metabolic environment. However, no studies to date have focused on older individuals in both conditions. Therefore, this study aimed to investigate the effects of a 12-week walking intervention on metabolic syndrome risk factors, liver function indicators, and liver ultrasound findings in older adults with both metabolic syndrome and MASLD. Methods A total of 66 participants aged 65–85 years had average ages, heights, and weights of 75.3 ± 5.8 years, 159.3 ± 9.3 cm, and 68.6 ± 6.8 kg, respectively. The participants resided in four senior living communities, and their diets were uniform. The participants from two facilities were assigned to the control group (CON, n = 33), whereas those from the other two facilities were allocated to the treadmill walking program group (WPG, n = 33). Each group comprised 13 males and 20 females. The intervention consisted of a low- to moderate-intensity walking program, conducted for 30 min per day, 6 days per week, totaling 180 min per week. The total daily calorie expenditure was recorded based on the values calculated from the treadmill. The walking intensity was adjusted by modifying the treadmill incline according to each participant’s heart rate corresponding to their maximal oxygen consumption (VO₂max). The exercise intensity was set at 50% on Mondays and Fridays, 60% on Tuesdays and Thursdays, and 70% on Wednesdays and Saturdays. Sundays were designated as rest days. Results Although there were no significant differences in caloric intake between the groups, the WPG exhibited a 52.5% increase in physical activity levels (p < 0.001), resulting in significant reductions in body weight (-10.2%), fat mass (-17.2%), and abdominal fat (-4.8%). The WPG showed a 16.1% increase in VO₂max, along with significant reductions in systolic blood pressure (-9.6%) and blood glucose (-16.9%), as well as notable improvements in lipid profiles (p < 0.001). The WPG also demonstrated significant reductions in aspartate aminotransferase (-40%), alanine aminotransferase (-23.5%), total protein (-14.4%), albumin (-8.1%), bilirubin (-17.6%), and liver ultrasound scores (-31.8%), with all changes showing significant intergroup differences (p < 0.001). Conclusions Along with a consistent diet, a 12-week walk has been shown to induce significant changes in the body composition and cardiometabolic factors of older adults, as well as notable improvements in liver function markers and imaging findings. Trial registration This study was registered with the Clinical Research Information Service of the Korea Centers for Disease Control and Prevention under Clinical Trials KCT0010079 on 26/12/2024.
Takuya Takakuwa, Kentaro Endo, Nobutake Ozeki
et al.
Abstract Background Mesenchymal stem cell (MSC) therapy has emerged as a promising treatment option for knee osteoarthritis. Adipose MSCs are commonly used due to their easy accessibility; however, synovial MSCs have demonstrated a superior capacity for cartilage matrix synthesis. The mechanism underlying this difference in therapeutic efficacy may involve lubricin, a crucial glycoprotein that maintains joint lubrication and protects cartilage. Notably, lubricin levels decrease during the progression of osteoarthritis. The aim of this study was to compare extracellular lubricin secretion by human synovial MSCs versus adipose MSCs. We also analyzed potential correlations between synovial MSC lubricin secretion and synovial inflammation. Methods Tissues for MSC isolation were obtained from 16 human donors with osteoarthritis who underwent total knee arthroplasty. Synovium was collected from the suprapatellar pouch on the femoral side, whereas adipose tissue was harvested from the subcutaneous layer of the knee skin incision. The synovial and adipose MSCs from each donor were cultured for 48 h (six replicate wells per donor), and lubricin concentrations in culture supernatants were measured using ELISA. For 11 donors, lubricin concentrations were measured directly, whereas the concentrations were normalized to cell number for the other 5 donors. MSC identity was assessed by flow cytometry and trilineage differentiation assays. Correlations between synovial MSC lubricin secretion and clinical parameters, including age, CRP, WBC, numerical rating scale for knee pain, synovial redness, synovial hyperplasia, and Krenn’s synovitis score, were assessed. Results Lubricin secretion was significantly greater from synovial MSCs than from adipose MSCs in 8 of the 11 directly analyzed MSC supernatants (p = 0.014 by Wilcoxon matched-pairs signed-rank test). However, the results from all 5 donors whose lubricin concentrations were normalized to cell number revealed a consistently higher lubricin secretion by synovial MSCs than by adipose MSCs. Every representative preparation of synovial and adipose MSCs fulfilled standard MSC identity criteria. No correlations were found between lubricin secretion and synovial inflammation or any clinical parameters. Conclusions More lubricin was secreted from human synovial MSCs than from adipose MSCs in the majority of donors examined. These findings support the potential therapeutic advantages of using synovial MSCs for osteoarthritis treatment.
Jeremy Cabrolier-Molina, Alexandra Martín-Rodríguez, Vicente Javier Clemente-Suárez
This systematic review, conducted in accordance with PRISMA guidelines and registered in PROSPERO (CRD42024619693), aimed to evaluate the effects of physical exercise interventions on muscle function and fall risk in older adults with and without sarcopenia. <b>Methods</b>: A comprehensive search of PubMed and Web of Science databases identified 11 randomized controlled trials (RCTs) published between 2015 and 2025. A total of 792 participants (mean age 75.13 ± 4.71 years; 65.53% women, 34.47% men) were included. Interventions varied in type—strength, balance, aerobic, and multi-component programs—with a minimum duration of 8 weeks. <b>Results</b>: The reviewed studies showed that physical exercise interventions significantly improved neuromuscular function, physical performance, and postural control in older adults. Positive effects were observed in gait speed, stair-climbing ability, grip strength, muscle mass, and bone density. Specific modalities such as Tai Chi improved postural control and neuromuscular response; dynamic resistance and functional training increased muscle strength and improved posture; Nordic walking reduced postural sway; and multi-component and combined walking-resistance training enhanced mobility and force efficiency. Programs integrating strength and balance components yielded the most consistent benefits. However, reporting on FITT (Frequency, Intensity, Time, Type) principles was limited across studies. <b>Conclusions</b>: Exercise interventions are effective in improving neuromuscular outcomes and reducing fall risk in older adults, both with and without sarcopenia. The findings support the need for tailored, well-structured programs and greater methodological standardization in future research to facilitate broader clinical application and maximize health outcomes.
Objective: The extension of life span and the decrease of fertility rate in the elderly population lead to the increasing aging proportion, and the natural growth of age is accompanied by aging. Cognitive diseases will emerge in large numbers at this stage, affecting the cognitive function of the elderly. Methods: The purpose of this review is to comprehensively explore the effects of open movement and locked exercise on cognitive function in the elderly. It discuss the specific effects and related neurobiological mechanisms of different types of exercise on cognitive function (including attention, memory, executive ability, etc.) in the elderly. Results: Open sports can effectively improve attention, executive function and cognitive flexibility in the elderly, and closed sports can have positive effects on memory and concentration by enhancing brain neuroplasticity and regulating neurotransmitters. Conclusion: although two types of movement on the cognitive function of the elderly has its emphasis, but overall, compared with closed movement, open movement in the elderly specific cognitive function (such as executive function, inhibitory control, task conversion, etc.) show more prominent effect, and help to reduce the brain caused by the age of atrophy and the risk of neurodegenerative changes.
Purpose: To predict parameters associated with patellar instability from magnetic resonance imaging (MRI) measurements using a machine learning model and to quantify the relative importance of radiographic risk factors that are associated with the presence of instability. Methods: Patients with a confirmed clinical diagnosis of patellar instability and age- and sex-matched controls without patellofemoral pathology were identified retrospectively. Multiple measurements to describe patella alta, malalignment, and trochlear dysplasia were performed on knee MRI scans. Univariate and multivariable logistic regressions were used to identify MRI measurements associated with patellar instability. Machine learning models were developed and evaluated for accuracy, discrimination, and calibration in predicting patellar instability. Shapley additive explanations (SHAP) were used to evaluate global and local variable importance. Results: A total of 256 patients were included in this study (128 with patellar instability and 128 controls, 63% female sex). Multivariable logistic regression found significant associations between diagnosis of patellar instability and lower patellotrochlear index (OR, 1.39 [95% CI, 1.15-1.69]; P < .001), greater Insall-Salvati ratio (OR, 1.65 [95% CI, 1.37-2.02]; P < .001), greater tibial tubercle–trochlear groove (TT-TG) distance (OR, 1.12 [95% CI, 1.06-1.19]; P < .001), and lower trochlear depth (OR, 1.42 [95% CI, 1.09-1.87]; P = .009). The random forest model had the highest performance among machine learning models, with an area under the receiver operating characteristic curve of 0.85. In this model, the variables with the greatest importance were Insall-Salvati ratio, TT-TG distance, and trochlear depth. Conclusions: The final model was able to reliably predict MRI-based parameters associated with patellar instability. Insall-Salvati ratio, TT-TG distance, and trochlear depth were the most important risk factors both in the machine learning models and using conventional statistical analysis. Clinical Relevance: This model has the potential to improve the diagnostic accuracy of patellar instability from MRI scans. The explanations provided by the model could enable clinicians to personalize care and understand the factors driving patellar instability in individual patients.
Manuel A Ballester Herrera, Josep M Muñoz Vives, Agusti Marti
# Background
Cryotherapy, long used for its pain-relieving and anti-inflammatory effects, is widely applied in sports medicine, physiotherapy, and postoperative care. Advances in cryotherapy methods, including the use of topical vapocoolant sprays, offer enhanced pain management and support recovery from musculoskeletal injuries by alleviating pain and reducing swelling in a targeted manner.
# Hypothesis/Purpose
The primary aim was to assess both immediate and sustained pain relief in subjects with mild to moderate musculoskeletal complaints.
# Study Design
Observational cohort study.
# Methods
Fifty-nine participants with mild to moderate musculoskeletal pain participated in a three-week observational study evaluating a cryotherapy spray. Pain and range of motion (ROM) were assessed using the Visual Analogue Scale (VAS) and goniometer measurements at multiple time points, including T-1 (before treatment), T0 (beginning of treatment), T2-min, T5-min, T10-min, T15-min, T30-min, T60-min, T7-days, T14-days, while participants were actively receiving treatment exclusively with the cryotherapy spray, and T21-days (1 week after treatment cessation). Acceptability Test and additional subjective questionnaires evaluated participants analysis of cooling sensation and product tolerance. Data were analyzed using Cumulative Logit Mixed Models (CLMM) and the non-parametric Friedman test for repeated measures.
# Results
The cryotherapy spray significantly reduced pain (VAS) from baseline (p \< 0.001, Hedges' g = -1.90) and improved joint mobility (ROM) with derived scores increasing from 3 (3--4) to 4 (4--4) by Day 21 (p \< 0.001). Rapid pain relief was reported by 35% of participants within 10 seconds, with 80% experiencing relief within three minutes. Comfort ratings were consistently high, with 95% of participants expressing satisfaction at Day 0, rising to 99% by Day 14. Global efficacy satisfaction measured by a subjective Likert scale also increased from 75% at baseline to 95% by Day 14. No adverse events were reported.
# Conclusion
This study supports the effectiveness of the cryotherapy spray in reducing pain and improving joint mobility, with both immediate and sustained benefits. High patient satisfaction and a favorable safety profile suggest its potential for clinical use. Further controlled studies could confirm its efficacy in broader populations.
# Level of Evidence
Level 3
Mendelian randomization uses genetic variants as instrumental variables to make causal inferences about the effects of modifiable risk factors on diseases from observational data. One of the major challenges in Mendelian randomization is that many genetic variants are only modestly or even weakly associated with the risk factor of interest, a setting known as many weak instruments. Many existing methods, such as the popular inverse-variance weighted (IVW) method, could be biased when the instrument strength is weak. To address this issue, the debiased IVW (dIVW) estimator, which is shown to be robust to many weak instruments, was recently proposed. However, this estimator still has non-ignorable bias when the effective sample size is small. In this paper, we propose a modified debiased IVW (mdIVW) estimator by multiplying a modification factor to the original dIVW estimator. After this simple correction, we show that the bias of the mdIVW estimator converges to zero at a faster rate than that of the dIVW estimator under some regularity conditions. Moreover, the mdIVW estimator has smaller variance than the dIVW estimator.We further extend the proposed method to account for the presence of instrumental variable selection and balanced horizontal pleiotropy. We demonstrate the improvement of the mdIVW estimator over the dIVW estimator through extensive simulation studies and real data analysis.
Daniel Wolf, Tristan Payer, Catharina Silvia Lisson
et al.
Self-supervised pre-training of deep learning models with contrastive learning is a widely used technique in image analysis. Current findings indicate a strong potential for contrastive pre-training on medical images. However, further research is necessary to incorporate the particular characteristics of these images. We hypothesize that the similarity of medical images hinders the success of contrastive learning in the medical imaging domain. To this end, we investigate different strategies based on deep embedding, information theory, and hashing in order to identify and reduce redundancy in medical pre-training datasets. The effect of these different reduction strategies on contrastive learning is evaluated on two pre-training datasets and several downstream classification tasks. In all of our experiments, dataset reduction leads to a considerable performance gain in downstream tasks, e.g., an AUC score improvement from 0.78 to 0.83 for the COVID CT Classification Grand Challenge, 0.97 to 0.98 for the OrganSMNIST Classification Challenge and 0.73 to 0.83 for a brain hemorrhage classification task. Furthermore, pre-training is up to nine times faster due to the dataset reduction. In conclusion, the proposed approach highlights the importance of dataset quality and provides a transferable approach to improve contrastive pre-training for classification downstream tasks on medical images.
Background: Deep learning models have shown promise in diagnosing neurodevelopmental disorders (NDD) like ASD and ADHD. However, many models either use graph neural networks (GNN) to construct single-level brain functional networks (BFNs) or employ spatial convolution filtering for local information extraction from rs-fMRI data, often neglecting high-order features crucial for NDD classification. Methods: We introduce a Multi-view High-order Network (MHNet) to capture hierarchical and high-order features from multi-view BFNs derived from rs-fMRI data for NDD prediction. MHNet has two branches: the Euclidean Space Features Extraction (ESFE) module and the Non-Euclidean Space Features Extraction (Non-ESFE) module, followed by a Feature Fusion-based Classification (FFC) module for NDD identification. ESFE includes a Functional Connectivity Generation (FCG) module and a High-order Convolutional Neural Network (HCNN) module to extract local and high-order features from BFNs in Euclidean space. Non-ESFE comprises a Generic Internet-like Brain Hierarchical Network Generation (G-IBHN-G) module and a High-order Graph Neural Network (HGNN) module to capture topological and high-order features in non-Euclidean space. Results: Experiments on three public datasets show that MHNet outperforms state-of-the-art methods using both AAL1 and Brainnetome Atlas templates. Extensive ablation studies confirm the superiority of MHNet and the effectiveness of using multi-view fMRI information and high-order features. Our study also offers atlas options for constructing more sophisticated hierarchical networks and explains the association between key brain regions and NDD. Conclusion: MHNet leverages multi-view feature learning from both Euclidean and non-Euclidean spaces, incorporating high-order information from BFNs to enhance NDD classification performance.
Irina A. Balandina, Aleksandr S. Terekhin, Anatolii A. Balandin
et al.
Aim – to study the dynamics of pubic symphysis parameters in men in the early adulthood, early and middle old age according to computed tomography (CT) data.
Material and methods. In the study, we used the results of a CT examination of 80 men without bone or pelvic organ pathology. All participants gave their consent to routine examination to exclude possible pathology of the pelvic bones. The CT investigation included the measurement of the height, width and thickness of the pubic symphysis in 3D reconstruction mode. The subjects were divided into three groups according to anatomical age classification. The first group included 25 early adulthood men (21 to 35 years old); the second group included 29 early old age men (56 to 74 years old); the third group included 26 middle old age men (75 to 88 years old).
Results. When comparing the parameters of height, width and thickness of the pubic symphysis, their statistically significant decrease by middle old age was revealed. Its height decreased from the early adulthood to early old age by 7.1% (t = 12.82, p 0.01) and further remained unchanged in middle old age. The width of the pubic symphysis was decreasing by 22.7% (t = 8.3, p 0.01) from the early adulthood to early old age and by 26.5% (t = 8.32, p 0.01) from early to middle old age. The symphysis thickness was growing from the early adulthood to early old age by 6.4% (t = 6.10, p 0.01) and from early to middle old age – by 1.1% (t = 1.08, p 0.05).
Conclusion. The results obtained in this study can be helpful for doctors of such specialties as traumatology, sports medicine and rehabilitation, forensic science, forensic medicine and many others.
David Freire-Obregón, Javier Lorenzo-Navarro, Oliverio J. Santana
et al.
We present a transfer learning analysis on a sporting environment of the expanded 3D (X3D) neural networks. Inspired by action quality assessment methods in the literature, our method uses an action recognition network to estimate athletes' cumulative race time (CRT) during an ultra-distance competition. We evaluate the performance considering the X3D, a family of action recognition networks that expand a small 2D image classification architecture along multiple network axes, including space, time, width, and depth. We demonstrate that the resulting neural network can provide remarkable performance for short input footage, with a mean absolute error of 12 minutes and a half when estimating the CRT for runners who have been active from 8 to 20 hours. Our most significant discovery is that X3D achieves state-of-the-art performance while requiring almost seven times less memory to achieve better precision than previous work.
Sacha Lewin, Maxime Vandegar, Thomas Hoyoux
et al.
The long-standing problem of novel view synthesis has many applications, notably in sports broadcasting. Photorealistic novel view synthesis of soccer actions, in particular, is of enormous interest to the broadcast industry. Yet only a few industrial solutions have been proposed, and even fewer that achieve near-broadcast quality of the synthetic replays. Except for their setup of multiple static cameras around the playfield, the best proprietary systems disclose close to no information about their inner workings. Leveraging multiple static cameras for such a task indeed presents a challenge rarely tackled in the literature, for a lack of public datasets: the reconstruction of a large-scale, mostly static environment, with small, fast-moving elements. Recently, the emergence of neural radiance fields has induced stunning progress in many novel view synthesis applications, leveraging deep learning principles to produce photorealistic results in the most challenging settings. In this work, we investigate the feasibility of basing a solution to the task on dynamic NeRFs, i.e., neural models purposed to reconstruct general dynamic content. We compose synthetic soccer environments and conduct multiple experiments using them, identifying key components that help reconstruct soccer scenes with dynamic NeRFs. We show that, although this approach cannot fully meet the quality requirements for the target application, it suggests promising avenues toward a cost-efficient, automatic solution. We also make our work dataset and code publicly available, with the goal to encourage further efforts from the research community on the task of novel view synthesis for dynamic soccer scenes. For code, data, and video results, please see https://soccernerfs.isach.be.
Nour Neifar, Afef Mdhaffar, Achraf Ben-Hamadou
et al.
In this paper, we present a systematic literature review on deep generative models for physiological signals, particularly electrocardiogram (ECG), electroencephalogram (EEG), photoplethysmogram (PPG) and electromyogram (EMG). Compared to the existing review papers, we present the first review that summarizes the recent state-of-the-art deep generative models. By analyzing the state-of-the-art research related to deep generative models along with their main applications and challenges, this review contributes to the overall understanding of these models applied to physiological signals. Additionally, by highlighting the employed evaluation protocol and the most used physiological databases, this review facilitates the assessment and benchmarking of deep generative models.