Vitality capacity (VC) reflects a physiological state and is a determinant domain of intrinsic capacity but has so far remained mainly theoretical. This study validates the vitality capacity domains ‘energy and metabolism’ and ‘neuromuscular function’ and examines its link to locomotor capacity and quality of life (QoL). Exploratory factor analysis (EFA) was performed on the combined dataset from the Fatigue Resistance AMErsfoort study (FRAME, n = 1000) and the Fatigue Plot study (FATPLOT,n = 620). Confirmatory factor analyses (CFA) were subsequently performed on data from the AMersfoort COhort study on functional decline, Healthy aging and Frailty (AMCOHF,n = 367) and the BrUssels sTudy on The Early pRedictors of FraiLtY (BUTTERFLY,n = 491), to validate VC in both middle-aged and older adults. Linear hierarchical regression analysis was used to investigate the relationship between VC, locomotor capacity, and QoL. EFA indicated a one-factor model and CFA validated this with good model fit in the dataset (BUTTERFLY) (Robust CFI; 0.960, SRMR: 0.040) and (AMCOHF) (Robust CFI; 0.942, SRMR: 0.055). This model validated maximal grip strength (GSmax), 30-s chair stand test (30CST), Multidimensional Fatigue Inventory (MFI-20) and Capacity to Perceived Vitality ratio physical (CPV-physical) to measure VC. Several assessments show a significant relationship with locomotor capacity and QoL. This study indicated that VC is a coherent domain and has a relationship with locomotor capacity and QoL.

Introduction Stroke is a leading cause of long-term disability, yet recovery trajectories are highly heterogeneous and difficult to predict at the individual level. Standardised clinical assessments primarily capture task accomplishment under controlled conditions but may overlook changes in movement quality, including the distinction between compensatory strategies and behavioural restitution. Additionally, they provide limited insight into the postural control processes that underlie functional task performance. Recent consensus initiatives emphasised the need for sensitive, functionally relevant measures of balance and mobility after stroke. Integrating routine clinical assessments with a sensor-based approach during functional tasks offers new opportunities to characterise recovery more precisely and to map longitudinal trajectories of poststroke change. Within the European Union-funded TARGET project, the FOSTER study aims to generate multimodal data to support data-driven, individualised prognostic models, so-called virtual twins, for personalised rehabilitation. Methods and analysis FOSTER is a prospective, single-centre observational cohort study at Rehabilitation Hospital Inkendaal (Belgium). We aim to include 100 adults within 8 weeks poststroke, assessed at three to five time points, including an entry assessment after admission and follow-up assessments within predefined windows at approximately 5, 8, 12 and 24 weeks poststroke. The protocol combines synchronised sensor-based measurements during functional tasks (seated reaching, standing reaching and gait initiation) obtained during an exergame, with standardised clinical tests and patient-reported outcome measures (PROMs). Sensors include surface electromyography, inertial measurement units, force plates and markerless motion capture. Sensor-based measurements target movement quality, and anticipatory and reactive postural control during self-initiated functional tasks. Derived outcomes will capture kinetic, kinematic and neuromuscular features of movement strategies, including temporal and asymmetry characteristics. In addition, in-task performance metrics, real-world activity monitoring through accelerometry, standardised clinical assessments and PROMs will be collected to provide a comprehensive, multimodal characterisation of recovery. Linear mixed-effects models will characterise longitudinal trajectories and relationships between clinical and sensor-derived measures. Ethics and dissemination The FOSTER study was approved by the Ethics Committee of Rehabilitation Hospital Inkendaal (B.U.N. 2024-TEE-002). Participants will be included after providing informed consent. Data are pseudonymised and stored on a General Data Protection Regulation-compliant platform. Findings will be disseminated through peer-reviewed publications, conferences and within the TARGET consortium. Trial registration number NCT06806761 .

IntroductionNeuromuscular fatigability impairs motor performance in both healthy and neurological populations. Corticomuscular coherence (CMC), derived from EEG and EMG recordings, reflects the brain-muscle interaction during movement. However, the impact of neuromuscular fatigability on CMC in healthy and neurological populations remains unclear.MethodsA systematic search of PubMed, Web of Science, and Embase was conducted up to 02/02/2026. Eligible studies investigated CMC changes related to fatiguing tasks in healthy or neurological participants. Two reviewers independently screened, extracted data, and assessed the risk of bias.ResultsFifteen non-randomized experimental studies were included, comprising predominantly neurologically healthy adults (n= 174) and a limited number of individuals with neurological conditions (n= 14). Fatiguing tasks varied widely in muscle group, contraction type, mode, and intensity. Across studies, neuromuscular fatigability was associated with heterogeneous changes in CMC, most commonly involving reductions in beta band coherence as fatigue progressed. However, preserved or increased beta band CMC was also reported in both upper- and lower-limb tasks, particularly during sustained or low- to moderate-intensity contractions. Alpha and gamma band CMC were less reported across the included studies. No consistent or limb-specific pattern of CMC modulation emerged, with observed responses depending on task demands, contraction intensity, muscle group, and stage of fatigue. Evidence from neurological populations was sparse but suggested generally lower CMC magnitude and greater disruption during fatiguing tasks compared with healthy controls.DiscussionThese findings indicate that fatigue-related changes in CMC do not reflect a uniform loss of corticomuscular coupling but rather task- and context-dependent adaptations in brain–muscle communication. Reductions in CMC may reflect diminished efficacy of corticospinal synchronization, whereas preserved or increased coherence may represent stabilization to maintain motor output with fatigue. By synthesizing how neuromuscular fatigability reshapes CMC across different experimental contexts and highlighting key methodological limitations, this review provides a framework to inform the design of future rehabilitation or neuromodulation trials targeting fatigability in both healthy and neurological populations.

Topological data analysis (TDA) is a widely used technique that extracts robust, deformation-invariant summaries of time series via persistent homology. While persistence descriptors are typically evaluated through downstream predictive performance, the inverse question remains underexplored: how much of the original signal can be recovered from persistence summaries alone? In this work, we study this preimage problem in the context of wearable inertial measurement unit (IMU) gait data. We introduce PD2IMU, a reconstruction framework that maps persistence-based representations from IMU windows to the corresponding raw time-series segments. Persistence diagrams are obtained via time-delay embedding (TDE) of the IMU signal and encoded using two forms: an H0 death-time vector, and an H1 birth-persistence point set. The PD2IMU framework employs these two inputs in three model architectures: (i) a residual MLP for H_0 , (ii) a pointwise MLP for H_1 , and (iii) a late-fusion design that combines the two prior models for H_0 + H_1 . We evaluate reconstruction quality across two window lengths using complementary metrics capturing pointwise error, shape similarity, amplitude matching, and local dynamics, and summarize performance via a global score. Finally, we assess whether reconstructed windows preserve condition-related structure by classifying walking speeds from features extracted on reconstructed signals. Results indicate that persistence-based encodings constrain coarse signal properties but do not support accurate sample-level recovery under cross-subject generalization.

Background: Vertical ground reaction force (vGRF) is a critical biomarker of joint loading and functional recovery, yet its measurement typically requires force plates in gait laboratories, limiting use outside specialized facilities. Deep Learning (DL) models can estimate vGRF from more accessible kinematic data, but the diversity of approaches and evaluation protocols makes it difficult to identify what truly drives performance. Methods: In this study, we evaluated three popular DL models (LSTM, BiLSTM, and Transformer) for stance-phase vGRF prediction using two kinematic input representations (3D marker trajectories and joint angles), and we additionally tested marker-set reduction down to three markers and a single ankle marker. To assess whether models learn beyond the typical healthy stance profile, we benchmarked all methods against a training-set mean-waveform baseline. Results: Across input modalities and architectures, prediction accuracy was high (RMSE <10% BW and R2>0.9). However, only the Transformer significantly outperformed the baseline, whereas LSTM and BiLSTM did not improve beyond this reference, despite similarly good absolute metrics. Marker trajectories and joint angles yielded comparable accuracy, indicating that performance is robust to the chosen kinematic representation. Finally, marker-set reduction for the Transformer showed no significant degradation when using only three markers, whereas an ankle-only input no longer outperformed the baseline. Conclusions: Transformer-based vGRF estimation can remain feasible with sparse kinematic inputs, supporting simplified out-of-lab gait analysis. Reporting an explicit mean-waveform baseline is recommended to avoid over-interpreting high absolute metrics in healthy gait, where waveform variability is limited.

Spatiotemporal gait parameters (SGP) derived from inertial measurement units (IMUs) are well-established in gait analysis and fall risk assessment. These interpretable features (such as step time, symmetry, double support, etc.) require accurate gait event (GE) detection and are commonly used in clinical and research settings. In contrast, topological data analysis (TDA) is an emerging approach that captures the global geometric and temporal structure of time series without requiring step segmentation. TDA maps signals into higher-dimensional phase spaces via time-delay embedding and quantifies their topological structure using persistent homology. In this study, we directly compare SGP and TDA features for classifying fall risk in a cohort of 78 older adults (41 non-fallers, 37 fallers) recruited at the University Hospital (UZ) Brussels. Each participant completed six IMU-recorded walking trials. SGP features were computed from GE using wavelet-based detection, while TDA features were extracted from vertical-axis acceleration. We also evaluate the impact of the time-delay embedding parameter ( τ ) on TDA classification performance. Our results show that both TDA features and SGP achieved comparable classification performance, with an AUC of 0.81. While the overall performance remained stable across different τ values, subject-level analysis revealed that fallers are more affected by τ variations. These results show that TDA can equal the discriminative power of handcrafted SGP features. To our knowledge, this is the very first direct comparison of SGP and TDA features in gait analysis - across any population - not just in elderly fall-risk assessment. This positions TDA as a promising alternative for future research in wearable sensors and human activity recognition tasks.

In clinical breast imaging, microcalcifications (MCs) are routinely interpreted as small radiological signs, typically analyzed in clusters. They guide radiologists in assessing breast lesions and determining the likelihood of malignancy. Yet, increasing biological evidence indicates that active processes occur not only within the calcified core but also in the microcalcification-surrounding tissue microenvironment (MCST). This raises the possibility that current assessments may overlook critical diagnostic information embedded in the MCST. However, conventional mammography and digital breast tomosynthesis (DBT) lack the spatial resolution required to determine where diagnostic information truly resides. We used a high-resolution ( ≈ 8 μ m) 3D micro-CT scanner to scan 94 paraffin-embedded breast biopsy blocks, from which 3504 individual MCs were segmented to obtain a binary 3D mask ( M_0 ) for each MC. Unlike previous studies analyzing clustered MCs at mammographic resolution, our analysis operates at the level of individual MCs. Radiomic features were extracted from each MC and used to train machine-learning classifiers to predict the histopathological label (benign or malignant) of the lesion in which each MC occurred. To determine (i) whether discriminative information arises predominantly from the calcified core or from the MCST and (ii) to separate genuine tissue signal from preprocessing or segmentation effects, we conducted three complementary analyses. First, we held M_0 constant and varied only the size of the rectangular preprocessing window around M_0 , i.e. the 3D region of grayscale data considered for feature computation. Second, we assessed robustness to segmentation by making small, incremental changes to M_0 (erosions to restrict the calcified core; dilations to extend into immediately adjacent tissue). Third, to test MCST-only signal, we classified MCs using features computed exclusively from concentric shells defined at incremental offsets from the M_0 : outer shells (thin layers of MCST just outside M_0 ) and inner shells (thin layers within the calcified core, inside M_0 ). Increasing the size of the preprocessing window around a constant M_0 improved classification performance (AUC ≈ 0.688 → 0.811), revealing strong contextual effects in radiomics feature extraction. Moderate mask dilations likewise improved performance (AUC ≈ 0.689 → 0.813), indicating that the diagnostic signal extends beyond the calcified core into the MCST. Remarkably, even when using only concentric shell features, classification performance remained high (AUC ≈ 0.81). This study provides the first imaging-based evidence - at micrometer scale and in 3D - pinpointing where the discriminative information of breast MCs arises: within the calcified core and/or the MCST? From a radiomics and technical standpoint, the preprocessing window matters: it should be controlled and reported, as it directly affects extracted features and can influence diagnostic performance. We find that MCST is the dominant source of discriminative information for classifying individual MCs as benign or malignant: classification using features computed exclusively from concentric shells performs on par with maximally dilated mask (encompassing both the core and MCST) but clearly exceeds calcified-core-only. These findings shift the interpretative focus of MC analysis from the calcified deposits themselves to their immediate microenvironment, suggesting that future biomarkers and quantitative features should target MCST.

Azure Kinect is a popular low-cost markerless Motion Capture (MoCap) system, showing promising results in clinical applications. However, during concurrent validation studies with a marker-based gold standard, reflective markers produce passive infrared (IR) noise, which significantly interferes with its tracking accuracy. In this study, we collected motion data from 15 healthy participants performing upper and lower limb exercises, concurrently recorded by Azure Kinect and the Vicon system. We found that Kinect's skeletal tracking primarily relies on IR images rather than depth images. Therefore, we developed a simple yet effective algorithm to mitigate noise in IR images. Our method significantly improved Kinect's skeletal tracking reliability, reducing missed poses from 10% to negligible levels and decreasing bone length variability across frames. Additionally, joint angle measurements improved, with lower Mean Absolute Error (MAE) in Range of Motion (ROM) and higher Intraclass Correlation Coefficient (ICC) of ROM. The code developed for this study is available at https://github.com/spongebobbe/pyKinectAzureImageManipulation.

Background:Hospitalized older adults often spend prolonged periods of time bedridden, leading to decreased muscle strength and function. To tackle this, rehabilitation aims to keep patients active and train affected muscles. Exergames have proven to be effective in the rehabilitation of different patient populations and offer a motivating solution to combat inactivity associated with hospitalization. Furthermore, blood flow restriction (BFR) is effective in therapy for weakened patients, so combining BFR and exergames might be promising. Objective:As part of an iterative process of user-centered development, this mixed method study investigates the acceptability and feasibility of the Ghostly game as a stand-alone added therapy or combined with BFR in strength training of hospitalized older adults. Methods:A mixed methods study was conducted on 15 hospitalized older adults. Participants were randomized into 3 groups and received daily interventions from the moment they were included in the geriatric ward, until discharge from the hospital. The Ghostly group received daily conventional therapy with the Ghostly game as added therapy, the Ghostly + BFR group received daily conventional therapy with Ghostly in combination with BFR as added therapy and last, the control group received daily conventional therapy with dose-matched isometric exercises as added therapy. The primary outcome, user experience, was assessed before discharge from the hospital using the Usefulness, Satisfaction, and Ease of Use questionnaire and through expert observations. Clinical outcomes such as muscle strength, muscle architecture, and segmental body composition were assessed at baseline and before discharge from the hospital to test the feasibility of the research protocol in preparation for future randomized controlled trials. Results:A total of 15 hospitalized older adults (11 female participants, 73.33%) were included in this study with an average age of 84.53 (range: 78-94) years. Participants received an average of 3.47 (range: 3-5) intervention sessions after transferring to the geriatric ward of the hospital. Results on user experience revealed high scores on all subcategories of the Usefulness, Satisfaction, and Ease of Use questionnaire (usefulness: 78.93%, ease of use: 82.99%, ease of learning: 85.36%, and satisfaction: 87.55%). Furthermore, expert observations identified issues with color contrast, reaction time speed, and the need to tailor the game to accommodate the diverse requirements of different patient populations. All outcomes and procedures were found feasible for a future randomized controlled trial. Conclusions:This mixed methods study combines the innovative aspects of an electromyography-driven exergame with strength training principles of BFR and reveals the acceptability and feasibility of the Ghostly game as a stand-alone added therapy modality for strength training in hospitalized older adults and in combination with BFR. Future improvements of the exergame could focus on addressing expert-identified issues, including optimizing color contrast, adjusting reaction time speeds, and tailoring the game to meet the needs of different patient populations.

Three-dimensional imaging technologies are increasingly used in breast reconstructive and plastic surgery due to their potential for efficient and accurate preoperative assessment and planning. This study systematically evaluates the accuracy and consistency of six commercially available 3D scanning applications (apps)—Structure Sensor, 3D Scanner App, Heges, Polycam, SureScan, and Kiri—in reconstructing the female torso. To avoid variability introduced by human subjects, a silicone breast mannequin model was scanned, with fiducial markers placed at known anatomical landmarks. Manual distance measurements were obtained using calipers by two independent evaluators and compared to digital measurements extracted from 3D reconstructions in Blender software. Each scan was repeated six times per application to ensure reliability. SureScan demonstrated the lowest mean error (2.9 mm), followed by Structure Sensor (3.0 mm), Heges (3.6 mm), 3D Scanner App (4.4 mm), Kiri (5.0 mm), and Polycam (21.4 mm), which showed the highest error and variability. Even the app using an external depth sensor (Structure Sensor) showed no statistically significant accuracy advantage over those using only the iPad’s built-in camera (except for Polycam), underscoring that software is the primary driver of performance, not hardware (alone). This work provides practical insights for selecting mobile 3D scanning tools in clinical workflows and highlights key limitations, such as scaling errors and alignment artifacts. Future work should include patient-based validation and explore deep learning to enhance reconstruction quality. Ultimately, this study lays the foundation for more accessible and cost-effective 3D imaging in surgical practice, showing that smartphone-based tools can produce clinically useful scans.

BACKGROUND:Multiple sclerosis (MS) is the most common neurological disease in young adults. Virtual reality (VR) offers a promising rehabilitation tool by providing controllable, personalised environments for safe, adaptable and engaging training. Virtual reality can be tailored to patients' motor and cognitive skills, enhancing motivation through exciting scenarios and feedback. OBJECTIVES:Primary objective To assess the effects of virtual reality interventions compared with an alternative or no intervention on lower limb and gait function, and balance and postural control in people with MS. Secondary objective To assess the effects of virtual reality interventions compared with an alternative or no intervention on upper limb function, cognitive function, fatigue, global motor function, activity limitation, participation restriction and quality of life, and adverse events in people with MS. SEARCH METHODS:We identified relevant articles through electronic searches of CENTRAL, MEDLINE, Embase, PEDro, CINAHL and Scopus. We also searched trials registries (ClinicalTrials.gov and the WHO ICTRP search portal) and checked reference lists. We carried out all searches up until August 2022. SELECTION CRITERIA:We included only (quasi-)randomised controlled trials (RCTs) that assessed virtual reality interventions, defined as "an artificial, computer-generated simulation or creation of a real-life environment or situation allowing the user to navigate through and interact with", in people with MS. The primary outcomes were lower limb and gait function, and balance and postural control. Secondary outcome measures were upper limb function, cognitive function, fatigue, global motor function, activity limitation, participation and quality of life, and adverse events. Eligible participants were people with MS who were 18 years or older. DATA COLLECTION AND ANALYSIS:Two review authors independently screened the studies based on pre-specified criteria, extracted study data and assessed the risk of bias of the included studies. We used the risk of bias 2 tool (RoB 2). A third review author was consulted to resolve conflicts. MAIN RESULTS:We included 33 RCTs with 1294 people with MS. The sample sizes of the included studies were relatively small and there was considerable heterogeneity between studies regarding the virtual reality devices and the outcome measures used. The control group either received no intervention, conventional therapy or an alternative intervention (an intervention that does not fit the description of conventional therapy for the rehabilitation of people with MS). We most frequently judged the risk of bias as 'some concerns' across domains, leading to an overall high risk of bias in the majority of included studies for all outcome measures. Primary outcomes When compared with no intervention, virtual reality interventions may result in no difference in lower limb and gait function (Timed Up and Go, mean difference (MD) -0.43 sec, 95% confidence interval (CI) -0.85 to 0.00; 6 studies, 264 participants; low-certainty evidence) or balance and postural control (Berg Balance Scale, MD 0.29 points, 95% CI -0.1 to 0.68; 4 studies, 137 participants; very low-certainty evidence). When virtual reality interventions are compared to conventional therapy, results for lower limb and gait function probably do not differ between interventions (Timed Up and Go, MD -0.2 sec, -1.65 to 1.25; 4 studies, 107 participants; moderate-certainty evidence). However, virtual reality interventions probably improve balance and postural control (Berg Balance Scale, MD 2.39 points, 95% CI 1.22 to 3.57; 7 studies, 201 participants; moderate-certainty evidence), almost reaching the clinically important difference (3 points). Secondary outcomes Compared to no intervention, the use of virtual reality may also improve upper limb function (9-Hole Peg Test, MD -4.19 sec, 95% CI -5.86 to -2.52; 2 studies, 84 participants; low-certainty evidence), almost reaching the clinically important difference (4.38 points) and participation and quality of life, but the evidence is very uncertain (MS International QoL, MD 9.24 points, 95% CI 5.76 to 12.73; 2 studies, 82 participants; very low-certainty evidence). Compared to conventional therapy, virtual reality interventions may improve participation and quality of life (Falls Efficacy Scale-1, MD -3.07 points, 95% CI -5.99 to -0.15; 3 studies, 101 participants; low-certainty evidence), but not upper limb function (9-Hole Peg Test, MD 0.10 sec, 95% CI -1.70 to 1.89; 3 studies, 93 participants; low-certainty evidence). For other key secondary outcome measures, i.e. global motor function and adverse events, there were no data available as these were not measured in the studies. AUTHORS' CONCLUSIONS:We found evidence that the use of virtual reality may be more effective than no intervention in improving upper limb function and participation and quality of life. Training with virtual reality may be superior to conventional therapy for improving balance and postural control, and participation and quality of life. For the other outcomes, there was no clear difference between virtual reality and conventional therapy. There was insufficient evidence to reach conclusions about the effect of virtual reality on global motor function, activity limitations and adverse events. Additional high-quality, large-scale studies are needed to expand and confirm these findings.

BACKGROUND AND OBJECTIVES:External ventricular drain (EVD) placement is often performed freehand, a technique subpar to accurate yet impractical image-guided methods, yielding optimal placement in only 70%. The aim of this study was to address shortcomings in EVD placement and image guidance technologies by implementing high-accuracy augmented reality (AR) guidance. METHODS:We conducted a prospective clinical pilot study to assess feasibility, safety, and clinical performance of EVD placement using a standalone AR headset equipped with high-accuracy inside-out infrared tracking and software addressing EVD placement. Placement quality was reported using a newly defined extended modified Kakarla scale, and dichotomized into clinically relevant outcome parameters. Results were compared with a nonconcurrent freehand control group using one-sided Fisher exact tests. RESULTS:Eleven AR-guided EVD placements were performed, achieving functional placement in all cases on the first attempt, vs 7 (64%) in the control group ( P = .045); successful placement in 9 (82%) vs 5 (45%); optimal in 8 (73%) vs 3 (27%) ( P = .043); suboptimal in 2 (18%) vs 5 (45%); and failed in 0 vs 1 (9%). No AR-guided placements required revision, whereas the freehand group had a 36% reintervention rate ( P = .045). Procedure-related complications occurred in 2 AR-guided cases (18%), vs 5 (45%) freehand (all post-reintervention). CONCLUSION:This study presents the first clinical use case of EVD placement using high-accuracy AR guidance contained in a standalone head-worn navigation system. Safe and reliable outcomes using a validated pipeline were demonstrated, eliminating stick-and-poke attempts and resulting in improved quality, increased single attempt success rates, and reduced revision and complication rates. Based on these results, a multicenter randomized controlled trial will be initiated.

Low-cost, portable motion capture (MoCap) systems struggle to achieve the same accuracy as the marker-based gold standard, and often fail to provide real-time feedback on patients' motion parameters. To address these challenges, we present HoloMoCap, a novel marker-based MoCap system enabling clinicians to track and visualize human movements in real-time through a head-mounted Augmented Reality (AR) display. HoloMoCap is a HoloLens 2 stand-alone application, requiring no external tracking systems or additional servers for motion analysis. The application utilizes the HoloLens' depth sensor, operating at 5 frames per second (fps), to perform inside-ut tracking of infrared markers attached to the patient's skin. At each frame, the system detects reflective markers, uses an on-evice Deep Learning (DL) model to associate each marker with its corresponding body landmark, and calculates anatomical joint angles (hip and knee flexion, abduction, and rotation). Validation against Vicon was performed during rehabilitation exercises (squats and hip abduction), showing that estimated joint angles maintain root-mean-square error (RMSE) and mean absolute error (MAE) below 2 degrees for most angles. HoloMoCap accurately estimated the range of motion (ROM) for hip abduction and knee flexion, with average MAEs of 0.4 degrees and 1.2 degrees, respectively. However, for hip flexion, the MAE can exceed 10 degrees at maximum flexion during squats. HoloMoCap shows promise as a portable and cost-effective solution for motion capture, although further improvements in accuracy and frame rate are necessary to broaden its clinical applications.

BACKGROUND:Breast microcalcifications (MCs) are considered to be a robust marker of breast cancer. A machine learning model can provide breast cancer diagnosis based on properties of individual MCs - if their characteristics are captured at high resolution and in 3D.PURPOSE:The main purpose of the study was to explore the impact of image resolution (8 µm, 16 µm, 32 µm, 64 µm) when diagnosing breast cancer using radiomics features extracted from individual high resolution 3D micro-CT MC images.METHODS:Breast MCs extracted from 86 female patients were analyzed at four different spatial resolutions: 8 µm (original resolution) and 16 µm, 32 µm, 64 µm (simulated image resolutions). Radiomic features were extracted at each image resolution in an attempt, to find a compact feature signature allowing to distinguish benign and malignant MCs. Machine learning algorithms were used for classifying individual MCs and samples (i.e., patients). For sample diagnosis, a custom-based thresholding approach was used to combine individual MC results into sample results. We conducted classification experiments when using (a) the same MCs visible in 8 µm, 16 µm, 32 µm, and 64 µm resolution; (b) the same MCs visible in 8 µm, 16 µm, and 32 µm resolution; (c) the same MCs visible in 8 µm and 16 µm resolution; (d) all MCs visible in 8 µm, 16 µm, 32 µm, and 64 µm resolution. Accuracy, sensitivity, specificity, AUC, and F1 score were computed for each experiment.RESULTS:The individual MC results yielded an accuracy of 77.27%, AUC of 83.83%, F1 score of 77.25%, sensitivity of 80.86%, and specificity of 72.2% at 8 µm resolution. For the individual MC classifications we report for the F1 scores: a 2.29% drop when using 16 µm instead of 8 µm, a 4.01% drop when using 32 µm instead of 8 µm, a 10.69% drop when using 64 µm instead of 8 µm. The sample results yielded an accuracy and F1 score of 81.4%, sensitivity of 80.43%, and specificity value of 82.5% at 8 µm. For the sample classifications we report for F1 score values: a 6.3% drop when using 16 µm instead of 8 µm, a 4.91% drop when using 32 µm instead of 8 µm, and a 6.3% drop when using 64 µm instead of 8 µm.CONCLUSIONS:The highest classification results are obtained at the highest resolution (8 µm). If breast MCs characteristics could be visualized/captured in 3D at a higher resolution compared to what is used nowadays in digital mammograms (approximately 70 µm), breast cancer diagnosis will be improved.

In several orthopedic procedures, the accurate use of surgical power tools is critical to avoid damage to surrounding tissues. As such, various guidance techniques and safety measures were developed. Augmented reality (AR) guidance shows promise but requires validation. We evaluated a new approach using an inside-out infrared tracking solution for the HoloLens to compensate for its limited tracking performance. Eighteen participants with varying levels of experience (student, trainee, expert) each drilled twelve trajectories (six perpendicular, six oblique) in equidimensional wooden logs. Three different techniques were evaluated: freehand drilling; proprioception-guided drilling towards the contralateral index finger; and AR-guided drilling using a tracked drill and a virtual overlay of the log with predefined guidance vectors. The angular errors between planned and performed trajectories were compared using a mixed-design ANOVA. The results demonstrated that guidance technique (p < 0.001) and drilling direction (p < 0.001) significantly affected drilling accuracy, while experience (p = 0.75) did not. AR outperformed both other techniques, particularly for oblique trajectories (p < 0.001). For perpendicular trajectories, it only outperformed proprioception guidance (p = 0.04). Target plots revealed an important scatter perpendicular to the longitudinal axis of the log during freehand and proprioception-guided drilling, especially for oblique trajectories. This inaccuracy disappeared during AR-guided drilling. As such, we were able to conclude that AR guidance using inside-out infrared tracking reduced angular uncertainty during directional drilling, resulting in improved drilling accuracy. This improvement was particularly noticeable for complex trajectories and angles. The benefits of AR guidance were observed across all experience levels, highlighting its potential for orthopedic applications. We believe this study opens the way for the methodical evaluation of AR guidance in specific orthopedic use cases.