Cyto R-CNN and CytoNuke Dataset: Towards reliable whole-cell segmentation in bright-field histological images.

BACKGROUND AND OBJECTIVE: Cell segmentation in bright-field histological slides is a crucial topic in medical image analysis. Having access to accurate segmentation allows researchers to examine the relationship between cellular morphology and clinical observations. Unfortunately, most segmentation methods known today are limited to nuclei and cannot segment the cytoplasm. METHODS: We present a new network architecture Cyto R-CNN that is able to accurately segment whole cells (with both the nucleus and the cytoplasm) in bright-field images. We also present a new dataset CytoNuke, consisting of multiple thousand manual annotations of head and neck squamous cell carcinoma cells. Utilizing this dataset, we compared the performance of Cyto R-CNN to other popular cell segmentation algorithms, including QuPath's built-in algorithm, StarDist, Cellpose and a multi-scale Attention Deeplabv3+. To evaluate segmentation performance, we calculated AP50, AP75 and measured 17 morphological and staining-related features for all detected cells. We compared these measurements to the gold standard of manual segmentation using the Kolmogorov-Smirnov test. RESULTS: Cyto R-CNN achieved an AP50 of 58.65% and an AP75 of 11.56% in whole-cell segmentation, outperforming all other methods (QuPath 19.46/0.91%; StarDist 45.33/2.32%; Cellpose 31.85/5.61%, Deeplabv3+ 3.97/1.01%). Cell features derived from Cyto R-CNN showed the best agreement to the gold standard (D¯=0.15) outperforming QuPath (D¯=0.22), StarDist (D¯=0.25), Cellpose (D¯=0.23) and Deeplabv3+ (D¯=0.33). CONCLUSION: Our newly proposed Cyto R-CNN architecture outperforms current algorithms in whole-cell segmentation while providing more reliable cell measurements than any other model. This could improve digital pathology workflows, potentially leading to improved diagnosis. Moreover, our published dataset can be used to develop further models in the future.

CellViT: Vision Transformers for precise cell segmentation and classification.

Nuclei detection and segmentation in hematoxylin and eosin-stained (H&E) tissue images are important clinical tasks and crucial for a wide range of applications. However, it is a challenging task due to nuclei variances in staining and size, overlapping boundaries, and nuclei clustering. While convolutional neural networks have been extensively used for this task, we explore the potential of Transformer-based networks in combination with large scale pre-training in this domain. Therefore, we introduce a new method for automated instance segmentation of cell nuclei in digitized tissue samples using a deep learning architecture based on Vision Transformer called CellViT. CellViT is trained and evaluated on the PanNuke dataset, which is one of the most challenging nuclei instance segmentation datasets, consisting of nearly 200,000 annotated nuclei into 5 clinically important classes in 19 tissue types. We demonstrate the superiority of large-scale in-domain and out-of-domain pre-trained Vision Transformers by leveraging the recently published Segment Anything Model and a ViT-encoder pre-trained on 104 million histological image patches - achieving state-of-the-art nuclei detection and instance segmentation performance on the PanNuke dataset with a mean panoptic quality of 0.50 and an F1-detection score of 0.83. The code is publicly available at https://github.com/TIO-IKIM/CellViT.

Histology-Based Prediction of Therapy Response to Neoadjuvant Chemotherapy for Esophageal and Esophagogastric Junction Adenocarcinomas Using Deep Learning.

PURPOSE: Quantifying treatment response to gastroesophageal junction (GEJ) adenocarcinomas is crucial to provide an optimal therapeutic strategy. Routinely taken tissue samples provide an opportunity to enhance existing positron emission tomography-computed tomography (PET/CT)-based therapy response evaluation. Our objective was to investigate if deep learning (DL) algorithms are capable of predicting the therapy response of patients with GEJ adenocarcinoma to neoadjuvant chemotherapy on the basis of histologic tissue samples. METHODS: This diagnostic study recruited 67 patients with I-III GEJ adenocarcinoma from the multicentric nonrandomized MEMORI trial including three German university hospitals TUM (University Hospital Rechts der Isar, Munich), LMU (Hospital of the Ludwig-Maximilians-University, Munich), and UME (University Hospital Essen, Essen). All patients underwent baseline PET/CT scans and esophageal biopsy before and 14-21 days after treatment initiation. Treatment response was defined as a ≥35% decrease in SUVmax from baseline. Several DL algorithms were developed to predict PET/CT-based responders and nonresponders to neoadjuvant chemotherapy using digitized histopathologic whole slide images (WSIs). RESULTS: The resulting models were trained on TUM (n = 25 pretherapy, n = 47 on-therapy) patients and evaluated on our internal validation cohort from LMU and UME (n = 17 pretherapy, n = 15 on-therapy). Compared with multiple architectures, the best pretherapy network achieves an area under the receiver operating characteristic curve (AUROC) of 0.81 (95% CI, 0.61 to 1.00), an area under the precision-recall curve (AUPRC) of 0.82 (95% CI, 0.61 to 1.00), a balanced accuracy of 0.78 (95% CI, 0.60 to 0.94), and a Matthews correlation coefficient (MCC) of 0.55 (95% CI, 0.18 to 0.88). The best on-therapy network achieves an AUROC of 0.84 (95% CI, 0.64 to 1.00), an AUPRC of 0.82 (95% CI, 0.56 to 1.00), a balanced accuracy of 0.80 (95% CI, 0.65 to 1.00), and a MCC of 0.71 (95% CI, 0.38 to 1.00). CONCLUSION: Our results show that DL algorithms can predict treatment response to neoadjuvant chemotherapy using WSI with high accuracy even before therapy initiation, suggesting the presence of predictive morphologic tissue biomarkers.

Identification of subject-specific responses to footwear during running.

Placing a stronger focus on subject-specific responses to footwear may lead to a better functional understanding of footwear's effect on running and its influence on comfort perception, performance, and pathogenesis of injuries. We investigated subject-specific responses to different footwear conditions within ground reaction force (GRF) data during running using a machine learning-based approach. We conducted our investigation in three steps, guided by the following hypotheses: (I) For each subject x footwear combination, unique GRF patterns can be identified. (II) For each subject, unique GRF characteristics can be identified across footwear conditions. (III) For each footwear condition, unique GRF characteristics can be identified across subjects. Thirty male subjects ran ten times at their preferred (self-selected) speed on a level and approximately 15 m long runway in four footwear conditions (barefoot and three standardised running shoes). We recorded three-dimensional GRFs for one right-foot stance phase per running trial and classified the GRFs using support vector machines. The highest median prediction accuracy of 96.2% was found for the subject x footwear classification (hypothesis I). Across footwear conditions, subjects could be discriminated with a median prediction accuracy of 80.0%. Across subjects, footwear conditions could be discriminated with a median prediction accuracy of 87.8%. Our results suggest that, during running, responses to footwear are unique to each subject and footwear design. As a result, considering subject-specific responses can contribute to a more differentiated functional understanding of footwear effects. Incorporating holistic analyses of biomechanical data is auspicious for the evaluation of (subject-specific) footwear effects, as unique interactions between subjects and footwear manifest in versatile ways. The applied machine learning methods have demonstrated their great potential to fathom subject-specific responses when evaluating and recommending footwear.

Modeling biological individuality using machine learning: A study on human gait.

Human gait is a complex and unique biological process that can offer valuable insights into an individual's health and well-being. In this work, we leverage a machine learning-based approach to model individual gait signatures and identify factors contributing to inter-individual variability in gait patterns. We provide a comprehensive analysis of gait individuality by (1) demonstrating the uniqueness of gait signatures in a large-scale dataset and (2) highlighting the gait characteristics that are most distinctive to each individual. We utilized the data from three publicly available datasets comprising 5368 bilateral ground reaction force recordings during level overground walking from 671 distinct healthy individuals. Our results show that individuals can be identified with a prediction accuracy of 99.3% by using the bilateral signals of all three ground reaction force components, with only 10 out of 1342 recordings in our test data being misclassified. This indicates that the combination of bilateral ground reaction force signals with all three components provides a more comprehensive and accurate representation of an individual's gait signature. The highest accuracy was achieved by (linear) Support Vector Machines (99.3%), followed by Random Forests (98.7%), Convolutional Neural Networks (95.8%), and Decision Trees (82.8%). The proposed approach provides a powerful tool to better understand biological individuality and has potential applications in personalized healthcare, clinical diagnosis, and therapeutic interventions.

Valuing vicinity: Memory attention framework for context-based semantic segmentation in histopathology.

The segmentation of histopathological whole slide images into tumourous and non-tumourous types of tissue is a challenging task that requires the consideration of both local and global spatial contexts to classify tumourous regions precisely. The identification of subtypes of tumour tissue complicates the issue as the sharpness of separation decreases and the pathologist's reasoning is even more guided by spatial context. However, the identification of detailed tissue types is crucial for providing personalized cancer therapies. Due to the high resolution of whole slide images, existing semantic segmentation methods, restricted to isolated image sections, are incapable of processing context information beyond. To take a step towards better context comprehension, we propose a patch neighbour attention mechanism to query the neighbouring tissue context from a patch embedding memory bank and infuse context embeddings into bottleneck hidden feature maps. Our memory attention framework (MAF) mimics a pathologist's annotation procedure - zooming out and considering surrounding tissue context. The framework can be integrated into any encoder-decoder segmentation method. We evaluate the MAF on two public breast cancer and liver cancer data sets and an internal kidney cancer data set using famous segmentation models (U-Net, DeeplabV3) and demonstrate the superiority over other context-integrating algorithms - achieving a substantial improvement of up to 17% on Dice score. The code is publicly available at https://github.com/tio-ikim/valuing-vicinity.

Gutenberg Gait Database, a ground reaction force database of level overground walking in healthy individuals.

The Gutenberg Gait Database comprises data of 350 healthy individuals recorded in our laboratory over the past seven years. The database contains ground reaction force (GRF) and center of pressure (COP) data of two consecutive steps measured - by two force plates embedded in the ground - during level overground walking at self-selected walking speed. The database includes participants of varying ages, from 11 to 64 years. For each participant, up to eight gait analysis sessions were recorded, with each session comprising at least eight gait trials. The database provides unprocessed (raw) and processed (ready-to-use) data, including three-dimensional GRF and two-dimensional COP signals during the stance phase. These data records offer new possibilities for future studies on human gait, e.g., the application as a reference set for the analysis of pathological gait patterns, or for automatic classification using machine learning. In the future, the database will be expanded continuously to obtain an even larger and well-balanced database with respect to age, sex, and other gait-specific factors.

Revealing the unique features of each individual's muscle activation signatures.

There is growing evidence that each individual has unique movement patterns, or signatures. The exact origin of these movement signatures, however, remains unknown. We developed an approach that can identify individual muscle activation signatures during two locomotor tasks (walking and pedalling). A linear support vector machine was used to classify 78 participants based on their electromyographic (EMG) patterns measured on eight lower limb muscles. To provide insight into decision-making by the machine learning classification model, a layer-wise relevance propagation (LRP) approach was implemented. This enabled the model predictions to be decomposed into relevance scores for each individual input value. In other words, it provided information regarding which features of the time-varying EMG profiles were unique to each individual. Through extensive testing, we have shown that the LRP results, and by extent the activation signatures, are highly consistent between conditions and across days. In addition, they are minimally influenced by the dataset used to train the model. Additionally, we proposed a method for visualizing each individual's muscle activation signature, which has several potential clinical and scientific applications. This is the first study to provide conclusive evidence of the existence of individual muscle activation signatures.

Sex-Specific Brain Responses to Imaginary Dance but Not Physical Dance: An Electroencephalography Study of Functional Connectivity and Electrical Brain Activity.

To date, most neurophysiological dance research has been conducted exclusively with female participants in observational studies (i.e., participants observe or imagine a dance choreography). In this regard, the sex-specific acute neurophysiological effect of physically executed dance can be considered a widely unexplored field of research. This study examines the acute impact of a modern jazz dance choreography on brain activity and functional connectivity using electroencephalography (EEG). In a within-subject design, 11 female and 11 male participants were examined under four test conditions: physically dancing the choreography with and without music and imagining the choreography with and without music. Prior to the EEG measurements, the participants acquired the choreography over 3 weeks with one session per week. Subsequently, the participants conducted all four test conditions in a randomized order on a single day, with the EEG measurements taken before and after each condition. Differences between the male and female participants were established in brain activity and functional connectivity analyses under the condition of imagined dance without music. No statistical differences between sexes were found in the other three conditions (physically executed dance with and without music as well as imagined dance with music). Physically dancing and music seem to have sex-independent effects on the human brain. However, thinking of dance without music seems to be rather sex-specific. The results point to a promising approach to decipher sex-specific differences in the use of dance or music. This approach could further be used to achieve a more group-specific or even more individualized and situationally adapted use of dance interventions, e.g., in the context of sports, physical education, or therapy. The extent to which the identified differences are due to culturally specific attitudes in the sex-specific contact with dance and music needs to be clarified in future research.

Fatigue-Related and Timescale-Dependent Changes in Individual Movement Patterns Identified Using Support Vector Machine.

The scientific and practical fields-especially high-performance sports-increasingly request a stronger focus be placed on individual athletes in human movement science research. Machine learning methods have shown efficacy in this context by identifying the unique movement patterns of individuals and distinguishing their intra-individual changes over time. The objective of this investigation is to analyze biomechanically described movement patterns during the fatigue-related accumulation process within a single training session of a high number of repeated executions of a ballistic sports movement-specifically, the frontal foot kick (mae-geri) in karate-in expert athletes. The two leading research questions presented for consideration are (1) Can characteristics of individual movement patterns be observed throughout the entire training session despite continuous changes, i.e., even as fatigue-related processes increase? and (2) How do intra-individual movement patterns change as fatigue-related processes increase throughout a training session? Sixteen expert karatekas performed 606 frontal foot kicks directed toward an imaginary target. The kicks were performed in nine sets at 80% (K-80) of the self-experienced maximal intensity. In addition, six kicks at maximal intensity (K-100) were performed after each of the nine sets. Between the sets, the participants took a 90-s break. Three-dimensional full-body kinematic data of all kicks were recorded with 10 infrared cameras. The normalized waveforms of nine upper- and lower-body joint angles were classified using a supervised machine learning method (support vector machine). The results of the classification revealed a disjunct distinction between the kinematic movement patterns of individual athletes. The identification of unique movement patterns of individual athletes was independent of the intensity and the degree of fatigue-related processes. In other words, even with the accumulation of fatigue-related processes, the unique movement patterns of an individual athlete can be clearly identified. During the training session, changes in intra-individual movement patterns could also be detected, indicating the occurrence of adaptations in individual movement patterns throughout the fatigue-related accumulation process. The results suggest that these adaptations can be modeled in terms of changes in patterns rather than increasing variance. Practical consequences are critically discussed.

Can Individual Movement Characteristics Across Different Throwing Disciplines Be Identified in High-Performance Decathletes?

Although the individuality of whole-body movements has been suspected for years, the scientific proof and systematic investigation that individuals possess unique movement patterns did not manifest until the introduction of the criteria of uniqueness and persistence from the field of forensic science. Applying the criteria of uniqueness and persistence to the individuality of motor learning processes requires complex strategies due to the problem of persistence in the learning processes. One approach is to examine the learning process of different movements. For this purpose, it is necessary to differentiate between two components of movement patterns: the individual-specific component and the discipline-specific component. To this end, a kinematic analysis of the shot put, discus, and javelin throwing movements of seven high-performance decathletes during a qualification competition was conducted. In total, joint angle waveforms of 57 throws formed the basis for the recognition task of individual- and discipline-specific throwing patterns using a support vector machine. The results reveal that the kinematic throwing patterns of the three disciplines could be distinguished across athletes with a prediction accuracy of up to 100% (57 of 57 throws). In addition, athlete-specific throwing characteristics could also be identified across the three disciplines. Prediction accuracies of up to 52.6% indicated that up to 10 out of 19 throws of a discipline could be assigned to the correct athletes, based on only knowing these athletes from the kinematic throwing patterns in the other two disciplines. The results further suggest that individual throwing characteristics across disciplines are more pronounced in shot put and discus throwing than in javelin throwing. Applications for training and learning practice in sports and therapy are discussed. In summary, the chosen approach offers a broad field of application related to the search of individualized optimal movement solutions in sports.

Acute Effects of an Incremental Exercise Test on Psychophysiological Variables and Their Interaction.

Besides neurophysiological effects, the potential influence of exercise induced strains in terms of peripheral physiology or subjectively perceived stress as well as their possible reciprocal relation is not clearly understood yet. This study aimed to analyze effects of increasing exercise intensity on brain activity (spontaneous EEG), heart rate variability (HRV) and rating of perceived exertion (RPE) by means of a graded exercise test (GXT). Fifteen participants performed an open-loop GXT on a bicycle ergometer beginning at 50W and an increment of 50W every three minutes. Rest measurements were conducted pre- (5 min) and especially post-exercise (15 min) to analyze (neuro-) physiological prolonged effects. EEG and HRV were measured in parallel before, during (including RPE) and after GXT. Brain activity showed next to already determined effects (e.g. increased (pre)frontal theta, alpha and beta power) a particular activation of the temporal lobe after GXT compared to pre-resting state. HRV frequency parameters significantly decreased following GXT. Recovery process revealed a significant alteration of EEG and HRV towards pre-resting state with prolonged effects in the temporal lobe. Correlation analysis during GXT led to moderately negative effects of EEG total spectrum power and HRV frequency parameters. Frontopolar and temporal lobe revealed noteworthy negative correlated effects with HRV. Referring to RPE, solely temporal gamma activity correlated moderately positive with RPE. Recovery exposed only in the temporal cortex a moderately negative correlation to HF power. Thus, further analysis of the temporal brain lobe in context with exhausting physical exercise comprising induced regulation of cardiovascular stress and perceived exertion is promoted. These results indicate a brain lobe specific relation to peripheral physiology as well as perceived strain with a dependency of rest or exercise condition. Therefore, enough incentives are given to encourage further analysis of a connection between the (neuro-) physiological system as well as subjectively perceived exertion.

Systematic Comparison of the Influence of Different Data Preprocessing Methods on the Performance of Gait Classifications Using Machine Learning.

Human movements are characterized by highly non-linear and multi-dimensional interactions within the motor system. Therefore, the future of human movement analysis requires procedures that enhance the classification of movement patterns into relevant groups and support practitioners in their decisions. In this regard, the use of data-driven techniques seems to be particularly suitable to generate classification models. Recently, an increasing emphasis on machine-learning applications has led to a significant contribution, e.g., in increasing the classification performance. In order to ensure the generalizability of the machine-learning models, different data preprocessing steps are usually carried out to process the measured raw data before the classifications. In the past, various methods have been used for each of these preprocessing steps. However, there are hardly any standard procedures or rather systematic comparisons of these different methods and their impact on the classification performance. Therefore, the aim of this analysis is to compare different combinations of commonly applied data preprocessing steps and test their effects on the classification performance of gait patterns. A publicly available dataset on intra-individual changes of gait patterns was used for this analysis. Forty-two healthy participants performed 6 sessions of 15 gait trials for 1 day. For each trial, two force plates recorded the three-dimensional ground reaction forces (GRFs). The data was preprocessed with the following steps: GRF filtering, time derivative, time normalization, data reduction, weight normalization and data scaling. Subsequently, combinations of all methods from each preprocessing step were analyzed by comparing their prediction performance in a six-session classification using Support Vector Machines, Random Forest Classifiers, Multi-Layer Perceptrons, and Convolutional Neural Networks. The results indicate that filtering GRF data and a supervised data reduction (e.g., using Principal Components Analysis) lead to increased prediction performance of the machine-learning classifiers. Interestingly, the weight normalization and the number of data points (above a certain minimum) in the time normalization does not have a substantial effect. In conclusion, the present results provide first domain-specific recommendations for commonly applied data preprocessing methods and might help to build more comparable and more robust classification models based on machine learning that are suitable for a practical application.

Electrical Brain Activity and Its Functional Connectivity in the Physical Execution of Modern Jazz Dance.

Besides the pure pleasure of watching a dance performance, dance as a whole-body movement is becoming increasingly popular for health-related interventions. However, the science-based evidence for improvements in health or well-being through dance is still ambiguous and little is known about the underlying neurophysiological mechanisms. This may be partly related to the fact that previous studies mostly examined the neurophysiological effects of imagination and observation of dance rather than the physical execution itself. The objective of this pilot study was to investigate acute effects of a physically executed dance with its different components (recalling the choreography and physical activity to music) on the electrical brain activity and its functional connectivity using electroencephalographic (EEG) analysis. Eleven dance-inexperienced female participants first learned a Modern Jazz Dance (MJD) choreography over three weeks (1 h sessions per week). Afterwards, the acute effects on the EEG brain activity were compared between four different test conditions: physically executing the MJD choreography with music, physically executing the choreography without music, imaging the choreography with music, and imaging the choreography without music. Every participant passed each test condition in a randomized order within a single day. EEG rest-measurements were conducted before and after each test condition. Considering time effects the physically executed dance without music revealed in brain activity analysis most increases in alpha frequency and in functional connectivity analysis in all frequency bands. In comparison, physically executed dance with music as well as imagined dance with music led to fewer increases and imagined dance without music provoked noteworthy brain activity and connectivity decreases at all frequency bands. Differences between the test conditions were found in alpha and beta frequency between the physically executed dance and the imagined dance without music as well as between the physically executed dance with and without music in the alpha frequency. The study highlights different effects of a physically executed dance compared to an imagined dance on many brain areas for all measured frequency bands. These findings provide first insights into the still widely unexplored field of neurological effects of dance and encourages further research in this direction.

Increased gait variability during robot-assisted walking is accompanied by increased sensorimotor brain activity in healthy people.

BACKGROUND: Gait disorders are major symptoms of neurological diseases affecting the quality of life. Interventions that restore walking and allow patients to maintain safe and independent mobility are essential. Robot-assisted gait training (RAGT) proved to be a promising treatment for restoring and improving the ability to walk. Due to heterogenuous study designs and fragmentary knowlegde about the neural correlates associated with RAGT and the relation to motor recovery, guidelines for an individually optimized therapy can hardly be derived. To optimize robotic rehabilitation, it is crucial to understand how robotic assistance affect locomotor control and its underlying brain activity. Thus, this study aimed to investigate the effects of robotic assistance (RA) during treadmill walking (TW) on cortical activity and the relationship between RA-related changes of cortical activity and biomechanical gait characteristics. METHODS: Twelve healthy, right-handed volunteers (9 females; M = 25 ± 4 years) performed unassisted walking (UAW) and robot-assisted walking (RAW) trials on a treadmill, at 2.8 km/h, in a randomized, within-subject design. Ground reaction forces (GRFs) provided information regarding the individual gait patterns, while brain activity was examined by measuring cerebral hemodynamic changes in brain regions associated with the cortical locomotor network, including the sensorimotor cortex (SMC), premotor cortex (PMC) and supplementary motor area (SMA), using functional near-infrared spectroscopy (fNIRS). RESULTS: A statistically significant increase in brain activity was observed in the SMC compared with the PMC and SMA (p < 0.05), and a classical double bump in the vertical GRF was observed during both UAW and RAW throughout the stance phase. However, intraindividual gait variability increased significantly with RA and was correlated with increased brain activity in the SMC (p = 0.05; r = 0.57). CONCLUSIONS: On the one hand, robotic guidance could generate sensory feedback that promotes active participation, leading to increased gait variability and somatosensory brain activity. On the other hand, changes in brain activity and biomechanical gait characteristics may also be due to the sensory feedback of the robot, which disrupts the cortical network of automated walking in healthy individuals. More comprehensive neurophysiological studies both in laboratory and in clinical settings are necessary to investigate the entire brain network associated with RAW.

Current State and Future Prospects of EEG and fNIRS in Robot-Assisted Gait Rehabilitation: A Brief Review.

Gait and balance impairments are frequently considered as the most significant concerns among individuals suffering from neurological diseases. Robot-assisted gait training (RAGT) has shown to be a promising neurorehabilitation intervention to improve gait recovery in patients following stroke or brain injury by potentially initiating neuroplastic changes. However, the neurophysiological processes underlying gait recovery through RAGT remain poorly understood. As non-invasive, portable neuroimaging techniques, electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) provide new insights regarding the neurophysiological processes occurring during RAGT by measuring different perspectives of brain activity. Due to spatial information about changes in cortical activation patterns and the rapid temporal resolution of bioelectrical changes, more features correlated with brain activation and connectivity can be identified when using fused EEG-fNIRS, thus leading to a detailed understanding of neurophysiological mechanisms underlying motor behavior and impairments due to neurological diseases. Therefore, multi-modal integrations of EEG-fNIRS appear promising for the characterization of neurovascular coupling in brain network dynamics induced by RAGT. In this brief review, we surveyed neuroimaging studies focusing specifically on robotic gait rehabilitation. While previous studies have examined either EEG or fNIRS with respect to RAGT, a multi-modal integration of both approaches is lacking. Based on comparable studies using fused EEG-fNIRS integrations either for guiding non-invasive brain stimulation or as part of brain-machine interface paradigms, the potential of this methodologically combined approach in RAGT is discussed. Future research directions and perspectives for targeted, individualized gait recovery that optimize the outcome and efficiency of RAGT in neurorehabilitation were further derived.

Explaining the unique nature of individual gait patterns with deep learning.

Machine learning (ML) techniques such as (deep) artificial neural networks (DNN) are solving very successfully a plethora of tasks and provide new predictive models for complex physical, chemical, biological and social systems. However, in most cases this comes with the disadvantage of acting as a black box, rarely providing information about what made them arrive at a particular prediction. This black box aspect of ML techniques can be problematic especially in medical diagnoses, so far hampering a clinical acceptance. The present paper studies the uniqueness of individual gait patterns in clinical biomechanics using DNNs. By attributing portions of the model predictions back to the input variables (ground reaction forces and full-body joint angles), the Layer-Wise Relevance Propagation (LRP) technique reliably demonstrates which variables at what time windows of the gait cycle are most relevant for the characterisation of gait patterns from a certain individual. By measuring the time-resolved contribution of each input variable to the prediction of ML techniques such as DNNs, our method describes the first general framework that enables to understand and interpret non-linear ML methods in (biomechanical) gait analysis and thereby supplies a powerful tool for analysis, diagnosis and treatment of human gait.

Fatigue-related changes in technique emerge at different timescales during repetitive training.

Training consisting of numerous repetitions performed as closely as possible to ideal techniques is common in sports and every-day tasks. Little is known about fatigue-related technique changes that emerge at different timescales when repeating complex actions such as a karate front kick. Accordingly, 15 karatekas performed 600 kicks (1 pre-block and 9 blocks). The pre-block comprised 6 kicks (3 with each leg) at maximum intensity (K-100%). Each block comprised 60 kicks (10 with each leg) at 80% of their self-perceived maximum intensity (K-80%) plus 6 K-100%. In between blocks, the participants rested for 90 seconds. Right leg kinematics (peak joint angles, peak joint angular velocities, peak joint linear resultant velocities, and time of occurrence of peaks) and kick duration corresponding to the K-80% were measured resulting in numerous variations with fatigue. At the timescale of tens of seconds, the changes involved variables that were related to velocity of execution (slowed down), while variables related to movement form were hardly affected. At the timescale of tens of minutes, the opposite results were observed. These findings challenge the long-standing rationale underlying repetitive training, suggesting instead that such involuntary variations in technique might play a crucial role in motor skill training.

Intra-individual gait patterns across different time-scales as revealed by means of a supervised learning model using kernel-based discriminant regression.

OBJECTIVE: Traditionally, gait analysis has been centered on the idea of average behavior and normality. On one hand, clinical diagnoses and therapeutic interventions typically assume that average gait patterns remain constant over time. On the other hand, it is well known that all our movements are accompanied by a certain amount of variability, which does not allow us to make two identical steps. The purpose of this study was to examine changes in the intra-individual gait patterns across different time-scales (i.e., tens-of-mins, tens-of-hours). METHODS: Nine healthy subjects performed 15 gait trials at a self-selected speed on 6 sessions within one day (duration between two subsequent sessions from 10 to 90 mins). For each trial, time-continuous ground reaction forces and lower body joint angles were measured. A supervised learning model using a kernel-based discriminant regression was applied for classifying sessions within individual gait patterns. RESULTS AND DISCUSSION: Discernable characteristics of intra-individual gait patterns could be distinguished between repeated sessions by classification rates of 67.8 ± 8.8% and 86.3 ± 7.9% for the six-session-classification of ground reaction forces and lower body joint angles, respectively. Furthermore, the one-on-one-classification showed that increasing classification rates go along with increasing time durations between two sessions and indicate that changes of gait patterns appear at different time-scales. CONCLUSION: Discernable characteristics between repeated sessions indicate continuous intrinsic changes in intra-individual gait patterns and suggest a predominant role of deterministic processes in human motor control and learning. Natural changes of gait patterns without any externally induced injury or intervention may reflect continuous adaptations of the motor system over several time-scales. Accordingly, the modelling of walking by means of average gait patterns that are assumed to be near constant over time needs to be reconsidered in the context of these findings, especially towards more individualized and situational diagnoses and therapy.