Medical Engineering & Physics最新文献

{"title":"MCCNet: A Multi-granularity Collaborative Cascaded-perception Network for Colorectal Cancer Histopathological Diagnosis.","authors":"Shuzhi Su, Kexue Zhang, Yanmin Zhu, Xiaoni Zhong, Yong Dai","doi":"10.1088/1873-4030/aeb605","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb605","url":null,"abstract":"<p><p>Colorectal cancer (CRC) is among the most lethal malignancies worldwide, and histopathological examination remains the clinical gold standard for diagnosis. In the histopathological examination, collaborative perception of fine-grained and coarse-grained lesions in histopathological samples is crucial for CRC diagnosis. However, existing methods often struggle to simultaneously model discriminative cues across granularities because of fine-grained glandular disorganization and coarse-grained morphological overlap among tissue categories. To tackle these issues, we propose a novel Multi-granularity Collaborative Cascaded-perception Network (MCCNet), which learns cross-granularity discriminative representations to overcome the limitations of single-granularity modeling. For fine-grained glandular disorganization, we design an Adaptive Glandular Offset Convolution (AGO-Conv). The convolution adaptively learns sampling offsets by injecting prominent edge gradients to actively focus on disorganized glands, and dynamically adjusts the sampling radius by a continuous domain constraint mechanism, thereby preventing out-of-bound glandular sampling. For coarse-grained similar lesions, we construct a Grouped Residual Linear-Attention (GRLA) module. GRLA linear approximate exponential kernel by segmented decomposable kernel mapping, and the discriminative inter-class representations across multiple subspaces are further preserved by subspace grouping and deep residual paths, thereby alleviating inter-class similarity among different tissue categories. To enable dynamic cross-granularity feature coupling, a Multi-Granularity Cascaded Feature Fusion (MGCFF) module is proposed that recalibrates the discriminative responses along channel and spatial dimensions via cascaded hierarchical representations and feature recalibration units, and further eliminates low-response background noise by multi-scale gating mechanism, thereby mitigating cross-granularity semantic conflicts. Extensive experimental results on the Chaoyang, Kather-5K, EBHI, and Private dataset show that MCCNet consistently outperforms existing methods.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225197","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Effects of LBBAP Pacing-Site Position under a Modelled Proximal LBBB: A 24-Model Electrophysiological Simulation Study.","authors":"Biao Tang, Zhiqiang Pei, Tianyi Zhang, Qichao Zhai, Yunzhang Cheng","doi":"10.1088/1873-4030/aeb604","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb604","url":null,"abstract":"<p><p>Left bundle branch area pacing (LBBAP) may recruit the distal left-sided conduction &#xD;system, but the electrical effect of pacing-site position along the left bundle branch (LBB) &#xD;remains incompletely defined. We evaluated this effect in 24 four-chamber virtual heart &#xD;anatomies using a Reaction-Eikonal finite-element model with rule-based myocardial &#xD;fibres and algorithmically generated His-Purkinje networks. A modelled proximal His&#xD;to-LBB disconnection preserved the distal LBB-Purkinje network while interrupting its &#xD;proximal connection. Six conditions were simulated in each anatomy: a sinoatrial-node &#xD;(SAN)-initiated left bundle branch block (LBBB) control, right ventricular apical pacing, &#xD;right ventricular outflow-tract pacing and a modelled non-selective LBB capture &#xD;configuration at proximal, nominal and distal positions. The primary outcome was &#xD;BiV_TAT95, the 95th percentile of activation times across left-ventricular, right&#xD;ventricular and interventricular-septal nodes. The control-versus-paced analysis used a &#xD;common simulation-clock origin and was interpreted as contextual benchmarking, &#xD;whereas the positional analysis used matched direct ventricular stimulation. In the &#xD;matched positional analysis, median BiV_TAT95 was 104.5, 93.5 and 88.5 ms for &#xD;proximal, nominal and distal LBBAP, respectively (Friedman chi-square = 48.000, P < &#xD;0.001, Kendall's W = 1.000); every anatomy followed distal < nominal < proximal. In &#xD;three selected anatomies, a 20% decrease or increase in the prescribed myocardial &#xD;conduction speeds altered absolute activation times but preserved this ordering in all &#xD;nine anatomy-by-speed combinations. Under the tested homogeneous substrate and &#xD;modelled non-selective capture configuration, the distal tested position produced the &#xD;lowest simulated BiV_TAT95; this does not imply superiority for every activation metric. &#xD;This conditional prediction requires validation against patient-specific electrical data.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225172","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Practical Pipeline for Patient-Specific Musculoskeletal Modeling: A Comprehensive MRI-Based Framework.","authors":"Yunsub Jung, Pavel E Galibarov, Morten Bilde Simonsen, Ilias Theodorakos, Emil Toft Petersen, Maiken Stilling, Michael Skipper Skipper Andersen","doi":"10.1088/1873-4030/aeb603","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb603","url":null,"abstract":"<p><p>Patient-specific musculoskeletal (MS) modeling can provide estimates of internal joint loading that canot be measured directly. However, limitations in automated segmentation due to data-scarce environments, morphing difficulty caused by anatomical variability, and functional non-alignment are major bottlenecks that hinder the practical application of modeling and reduce productivity in musculoskeletal research. To comprehensively address these practical issues, this study proposes and validates an integrated, automated modeling framework.. First, we automatically segmented the lower extremity bones of the 4 knee osteoarthritis patients using a multi-atlas-based segmentation algorithm. By integrating the segmented bone with motion capture and force plate data, a modeling pipeline including multi-stage morphing and functional alignment processes was implemented. Using the constructed customized model, the medial and lateral knee contact forces (FMC and FLC) were estimated during a single-leg step-up exercise, and these were compared and evaluated against the results based on manual segmentation. The contact forces estimated through the automated segmentation framework demonstrated high consistency with the manual results. The mean absolute error ranged from 0.04 ± 0.03 to 0.06 ± 0.03 BW for FMC, and from 0.06 ± 0.04 to 0.09 ± 0.07 BW for FLC, and the intraclass correlation coefficient demonstrated a high correlation of over 0.99 across all patients. This study demonstrated that the proposed automated pipeline provides a practical approach to address misalignment caused by anatomical variability, while generating joint load estimates equivalent to the manual approach. By addressing key methodological challenges, this study will bring patient-specific models closer to future clinical application, where they could help assess individual loading across different activities and inform treatment and rehabilitation strategies.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Quality-Guided Extension of Convolution Kernel Compensation for High-Density Surface EMG Decomposition.","authors":"Li Jiang, Donghao Lv, Fuhao Yang, Heng Zhang","doi":"10.1088/1873-4030/aeb54e","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb54e","url":null,"abstract":"<p><p>High-density surface electromyography (HD-sEMG) decomposition provides a non-invasive way to identify motor unit (MU) discharge patterns. However, the number of MUs extracted by blind source separation-based methods is often limited by repeated convergence to dominant sources, particularly when signal quality decreases or contraction level increases. In this study, we propose QE-CKC, a quality-guided extension framework built upon gradient convolution kernel compensation (gCKC), to recover more reliable MU discharge trains while maintaining decomposition accuracy. The proposed method expands candidate search, evaluates candidates using multiple quality measures, and removes redundant outputs through global ranking and spike-train-based deduplication.&#xD;&#xD;QE-CKC was evaluated using 150 simulated HD-sEMG trials covering five signal-to-noise ratio levels and three contraction levels, together with 66 real forearm HD-sEMG recordings from 11 participants. Across all simulation conditions, QE-CKC identified $14.06 pm 4.79$ matched MUs per trial, compared with $11.72 pm 3.18$ for gCKC and $6.51 pm 6.76$ for KMCKC. F1 score and rate of agreement remained comparable to gCKC. In the real recordings, QE-CKC identified $16.17 pm 10.59$ valid MUs per recording, compared with $13.29 pm 8.23$ for gCKC and $8.53 pm 9.39$ for KMCKC. After accounting for repeated measurements, QE-CKC identified 2.88 more valid MUs per recording than gCKC (95% CI: 1.22--4.54, Holm-adjusted $p<0.001$). Among MUs independently reproduced from non-overlapping channel subsets, QE-CKC-only MUs showed discharge-train agreement comparable to that of MUs shared with gCKC. These results indicate that QE-CKC increases MU yield while maintaining comparable decomposition accuracy and reproducibility.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149208269","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Influence of screw diameter on the pull-out resistance of additively manufactured auxetic bone screws in cancellous and cortical bone.","authors":"Ahmet Meram, Mehmet Emin Çetin, Haluk Yaka, Kemal Ödev, Mustafa Özer","doi":"10.1088/1873-4030/aeaee9","DOIUrl":"10.1088/1873-4030/aeaee9","url":null,"abstract":"<p><p>This study investigates the effect of screw diameter on the pull-out strength of a bone screw with a loosening-resistant auxetic structure manufactured using additive manufacturing. The variation in mechanical performance depending on screw diameter was examined experimentally and through finite element analysis in cancellous and cortical bone structures. First, screw designs with different diameters incorporating an auxetic cell structure were created in SolidWorks. The designed screws were analyzed using ANSYS Workbench finite element software, confirming that the structures exhibit a negative Poisson's ratio (NPR). The screws were manufactured from Ti6Al4V alloy using the direct metal laser sintering additive manufacturing process. The fabricated auxetic screws were inserted into synthetic cancellous and cortical bone (L4-L5 vertebral bone). The pull-out tests were performed under conditions conforming to the ASTM F543 standard. Force-displacement curves were obtained from the experiments, and tensile tests were also conducted to determine the mechanical strength of the screws. To evaluate the performance of the auxetic screws, identical pull-out tests were carried out on conventional bone screws. After the tests, SEM images were obtained to examine the damage mechanisms in detail. The results showed that auxetic bone screws exhibited approximately 4%-15% higher pull-out strength compared to conventional screws. This improvement is attributed to increased surface roughness and radial expansion under tensile loading, which enhance mechanical interlocking at the bone-screw interface. Finite element analyses further indicated that the magnitude of the NPR increased with increasing screw diameter, suggesting more effective mechanical behavior at larger diameters. In addition, the porous structure of the auxetic screw is expected to facilitate bone ingrowth, thereby supporting bone healing and reducing the risk of loosening. Finally, the developed auxetic bone screw has the potential to increase resistance to loosening. In addition, it may enhance biological integration with bone, which could contribute to improved surgical outcomes and greater patient comfort.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149117484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fuzzy Logic-Based Modeling and Multi-Objective Optimization of Blood Pump Rotational Speed Considering Hydraulic and Hemolytic Performance.","authors":"Abdoulaye Billo Diallo, Hasan Çınar, Ömer İncebay, Rafet Yapıcı","doi":"10.1088/1873-4030/aeb4bb","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb4bb","url":null,"abstract":"<p><p>This study aims to determine the optimal operating rotational speed of a centrifugal blood pump. In order to investigate this, a centrifugal blood pump was designed, and its hydraulic efficiency and hemolysis index were calculated using computational fluid dynamics (CFD) simulations over a range of 2500-3500 rpm for rotational speed and 2.5-6 L/min for flow rate. The pump design was carried out by considering key geometric parameters, including the number of blades, inlet and outlet widths, inlet and outlet angles, impeller diameter, volute inlet width, and volute outlet diameter. Subsequently, a fuzzy logic-based pump model was developed using rotational speed and mass flow rate as input parameters, and pressure difference, hemolysis index, and hydraulic efficiency as output parameters. After validating this model with CFD simulations, the optimal pump rotational speed was determined by solving an objective function incorporating hydraulic efficiency and hemolysis index using the particle swarm optimization (PSO) algorithm. The results indicate that the optimal rotational speed of the centrifugal pump is 3150 rpm, at which a hydraulic efficiency of 36% and a hemolysis index (HI) of 1.65x10-5 are achieved. In summary, this study presents a systematic methodology for determining the optimal operating speed that enables centrifugal blood pumps to operate efficiently while maintaining a low hemolysis index.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149197407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Patient and Database Shift in Lightweight ECG Arrhythmia Classification: A Controlled Evaluation of Noise, Quantization, and Embedded Inference.","authors":"Hyun Mok Park, Woojin Cho, Seok-Oh Bang, Hyun-Seok Choi, Ki-Tae Kwon, Jin-Sung Roh, Sang-Wuk Shin, Jong-Min Lim","doi":"10.1088/1873-4030/aeb4bc","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb4bc","url":null,"abstract":"<p><p>Efficient microcontroller execution does not ensure reliable electrocardiogram classification across patients and databases. We conducted a controlled benchmark of three convolutional architectures under a corrected subject-disjoint MIT-BIH split, external INCART testing, Noise Stress Test Database artifacts, 16-bit floating-point (FP16) and full-integer 8-bit (INT8) conversion, and classifier-only invocation on an ESP32 without external RAM. Five fixed seeds were used. On the MIT-BIH test cohort (DS2), INT8 four-class macro-F1 was 0.474-0.498 and supraventricular (S) recall was 0.150-0.180. Mean S-class area under the receiver operating characteristic curve (AUROC) was below 0.5 in all six architecture-regime groups, although all corresponding record-cluster intervals included 0.5. Within-record ranking could remain high despite poor pooled ranking. On INCART, S-class ranking was stronger, with all six patient-cluster AUROC intervals above 0.5, but the representative mixed-noise depthwise-separable network (DS-CNN) achieved 0.797 recall with 0.074 precision. Including record 201, which shares a subject with test record 202, increased DS2 S-recall point estimates by 2.49-14.68 percentage points. In the primary architecture panel, FP16 and INT8 met the engineering tolerances in all 120 comparisons. DS-CNN required 9.49 ms per classifier invocation on the ESP32 using preprocessed, annotation-centered inputs. Acquisition, QRS detection, preprocessing, current, and energy were not measured. Patient-dependent S-score behavior and external false-positive predictions limited this panel more than numerical precision or microcontroller resources.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149197456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A two-timescale reduced-order model coupling acute arterial compliance with stress- and shear-regulated growth and remodeling.","authors":"Ahmet Çalık","doi":"10.1088/1873-4030/aeb093","DOIUrl":"10.1088/1873-4030/aeb093","url":null,"abstract":"<p><p>Arterial growth and remodeling (G&R) models should distinguish immediate load-induced deformation from slower structural adaptation. A computationally light two-timescale model was therefore developed in which the reference-pressure structural geometry, described by inner radiusand wall thickness, is separated from the instantaneously loaded geometryand. At every slow G&R state, a nonlinear scalar equilibrium is solved for pressure-dependent dilation while annular wall area is conserved to represent acute wall incompressibility. The resulting loaded geometry determines circumferential stress and wall shear stress (WSS), whose signed deviations from homeostatic targets are transmitted through separate stress and shear signal channels. Stress signaling regulates wall growth through a lumped growth-factor variable, whereas WSS signaling regulates structural radius. The effective wall-growth and radius-rate parameters were constrained using separate rat aortic-thickness and mesenteric-artery diameter endpoints and were not interpreted as universal validation. For an acute pressure increase from 120 to 180 mmHg, the compliant model predicted an 8.11% increase in loaded radius and a 6.78% decrease in loaded thickness; the corresponding stress and WSS changes wereand. A matched rigid-wall model produced no acute geometric change,stress, and no WSS change. At day 28, loaded thickness increased by 17.82% and 19.80% in staged and sustained high-pressure protocols, respectively, while loaded radius increased by 14.11% under increased flow. These states remained transient: equilibrium-entry times were 95-242 d under the primary criterion. All 49 acute-parameter combinations yielded physical solutions, although response magnitudes varied substantially. The framework thus adds acute compliance to homeostatic feedback without abandoning reduced-order interpretability, while making the limits of endpoint anchoring and parameter identifiability explicit.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149117512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A passive thin-film mechanical device for visual respiratory monitoring during oxygen mask use: development and bench validation.","authors":"Dylan Rowe, Mariam Rowe, Phillip Melksham","doi":"10.1088/1873-4030/aeb4bd","DOIUrl":"https://doi.org/10.1088/1873-4030/aeb4bd","url":null,"abstract":"<p><p>Objective&#xD;Continuous respiratory monitoring during transfer from the operating theatre to the post-anaesthesia care unit (PACU) is frequently unavailable despite this being a period of recognised patient vulnerability. Existing electronic respiratory monitors introduce cost, complexity and set-up that may limit use in fast-paced clinical environments. We developed Flow-Bag, a fully passive mechanical breath-indicating device consisting of a thin heat-sealed high-density polyethylene (HDPE) film bag that attaches externally to an oxygen mask.&#xD;Methods&#xD;Preliminary film-thickness screening was undertaken using a programmable breathing simulator (PEMDx Pty Ltd, Australia). A 6 µm (per ply) prototype was assessed across respiratory rates of 5-60 breaths min⁻¹ and tidal volumes of 50-600 mL. Three blinded observers positioned 5 m from the device counted visible inflation-deflation cycles over 60-s intervals, with counts compared against simulator-delivered breaths. Performance was additionally assessed across four commercially available oxygen mask models.&#xD;Results&#xD;In screening, the lowest detected tidal volume fell from 400 mL with 20 µm film to 150 mL with 6 µm film, the thinnest film that could be reproducibly manufactured using the processes available to us. The final prototype produced observable breath indications in 42 of 56 tested conditions (all tidal volumes ≥150 mL) but not at 50 or 100 mL. Within these detectable conditions, observer counts agreed 100% with simulator-delivered breaths. Performance was maintained across four oxygen mask models. However, sensitivity was reduced by incomplete sealing at either the mask-to-face or device-to-mask interface.&#xD;Conclusion&#xD;Flow-Bag is a fully passive mechanical breath indicator. Under controlled bench conditions using a mechanical breathing simulator, it provided reliable visual indication of simulated breathing across respiratory rates of 5-60 breaths min⁻¹ and tidal volumes of 150-600 mL. Subject to clinical validation and safety testing, this device may offer a potentially low-cost adjunct alongside standard monitoring.&#xD.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149197403","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Application of machine learning techniques in the measurement of ground reaction force (GRF) waveforms using inertial measurement unit (IMU) sensors-a systematic literature review.","authors":"Malik Naveed Akhter, Haider Ali, Imran Khan Niazi, Umair Javaid Manj, Denise Taylor","doi":"10.1088/1873-4030/aeae3c","DOIUrl":"10.1088/1873-4030/aeae3c","url":null,"abstract":"<p><p>Three-dimensional ground reaction forces (GRFs) are important parameters in human motion analysis, yet their measurement is confined to laboratory settings. GRF estimation from wearable inertial measurement unit (IMU) data using machine learning and deep learning has emerged as a promising avenue for kinetic analysis outside the laboratory, but no consensus on standardized methods or best practices has emerged. This review synthesizes existing methods and provides a framework of current practices and recommendations for future work. The review followed PRISMA guidelines (PROSPERO CRD42024578576) and searched PubMed, Scopus, IEEE Xplore, and ScienceDirect from 2011 to 2026. Studies were included if they used IMUs and machine learning or deep learning to estimate GRFs and reported quantitative performance metrics. Data were extracted on participant characteristics, IMU specifications and placement, preprocessing, models, validation protocols, and error metrics. Twenty-six studies met the inclusion criteria, with sample sizes ranging from 5 to 47 (median = 15) and IMU counts ranging from 1-16. Most studies used triaxial accelerometers and gyroscopes sampled at 100-1000 Hz. Vertical GRF was generally estimated more accurately than the anteroposterior and mediolateral components within the same study. A single cross-study accuracy ranking was not defensible because studies used different normalization denominators, evaluation windows, tasks, sensor configurations, and validation strategies. Heterogeneity in validation protocols and demographic samples limited direct cross-study comparisons. This review synthesizes how implementation choices, including model architecture, feature-engineering strategy, sensor configuration, and validation protocol, shape GRF prediction across axes. The variability observed underscores the need for standardized methodologies, larger and more diverse cohorts, consistent performance metrics, and benchmarking on public datasets to enable fair cross-study comparisons and accelerate clinical translation.</p>","PeriodicalId":49836,"journal":{"name":"Medical Engineering & Physics","volume":" ","pages":""},"PeriodicalIF":2.4,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149117457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}