Bilateral Deep Brain Stimulation (DBS) of the subthalamic nucleus is commonly used for treating motor symptoms in patients with advanced Parkinson’s Disease (PD). The aim of this study is to quantitatively and non-invasively evaluate motor control changes in PD patients following DBS through an approach based on the combination of graph theory and frequency-domain electromyography (EMG) analysis. Instrumented gait analysis was carried out on a group of 30 PD patients and 30 age-matched controls. PD patients were longitudinally followed up, with assessments pre-DBS implant ( $\mathrm{T}_{{0}}\text {)}$ , 3 months post-DBS implant ( $\mathrm{T}_{{1}}\text {)}$ , and 12 months post-DBS implant ( $\mathrm{T}_{{2}}\text {)}$ . EMG signals from 12 lower-limb and trunk muscles were acquired, calculating Inter-Muscular Coherence (IMC) for each muscle pair. Adjacency matrices derived from IMC were used to generate 3D muscle networks through a force-based algorithm. Two families of network parameters were extracted: global metrics (modularity and density) and local metrics (node strength and local clustering coefficient). Muscle network modularity of PD patients at T0 was significantly lower than that of controls ( $0.34{\,}\pm{\,}0.07$ vs. $0.41{\,}\pm{\,}0.07$ ; ${p} =0.003$ ) and this difference persisted at T1 ( $0.35{\,}\pm{\,}0.01$ ; ${p} =0.037$ ), but not at T2 ( $0.38{\,}\pm{\,}0.01$ ; ${p} =1.00$ ). Analogously, muscle network density was higher in PD patients at T0 (T0: $0.69{\,}\pm{\,}0.10$ vs. Controls: $0.56{\,}\pm{\,}0.10$ ; ${p} =0.004$ ), decreased at T1 ( $0.65{\,}\pm{\,}0.14$ ; ${p} =0.034$ ), and was comparable to that of controls at T2 ( $0.60{\,}\pm{\,}0.14$ ; ${p} =1.00$ ). Node-level analyses similarly showed that PD patients values moved toward control-group reference levels after DBS surgery, reflecting reduced individual muscle connectivity and a more structured pattern of muscle coordination. Global metrics showed a good agreement with respect to the clinical score UPDRS-III. Graph theory applied to EMG analysis opens new perspectives in the study of motor control strategies during gait and confirms the efficacy of DBS in alleviating motor symptoms of PD patients.

Bilateral Deep Brain Stimulation (DBS) of the subthalamic nucleus is commonly used for treating motor symptoms in patients with advanced Parkinson's Disease (PD). The aim of this study is to quantitatively and non-invasively evaluate motor control changes in PD patients following DBS through an approach based on the combination of graph theory and frequency-domain electromyography (EMG) analysis. Instrumented gait analysis was carried out on a group of 30 PD patients and 30 age-matched controls. PD patients were longitudinally followed up, with assessments pre-DBS implant ( $\mathrm{T}_{{0}}\text {)}$ , 3 months post-DBS implant ( $\mathrm{T}_{{1}}\text {)}$ , and 12 months post-DBS implant ( $\mathrm{T}_{{2}}\text {)}$ . EMG signals from 12 lower-limb and trunk muscles were acquired, calculating Inter-Muscular Coherence (IMC) for each muscle pair. Adjacency matrices derived from IMC were used to generate 3D muscle networks through a force-based algorithm. Two families of network parameters were extracted: global metrics (modularity and density) and local metrics (node strength and local clustering coefficient). Muscle network modularity of PD patients at T0 was significantly lower than that of controls ( $0.34{\,}\pm{\,}0.07$ vs. $0.41{\,}\pm{\,}0.07$ ; ${p} =0.003$ ) and this difference persisted at T1 ( $0.35{\,}\pm{\,}0.01$ ; ${p} =0.037$ ), but not at T2 ( $0.38{\,}\pm{\,}0.01$ ; ${p} =1.00$ ). Analogously, muscle network density was higher in PD patients at T0 (T0: $0.69{\,}\pm{\,}0.10$ vs. Controls: $0.56{\,}\pm{\,}0.10$ ; ${p} =0.004$ ), decreased at T1 ( $0.65{\,}\pm{\,}0.14$ ; ${p} =0.034$ ), and was comparable to that of controls at T2 ( $0.60{\,}\pm{\,}0.14$ ; ${p} =1.00$ ). Node-level analyses similarly showed that PD patients values moved toward control-group reference levels after DBS surgery, reflecting reduced individual muscle connectivity and a more structured pattern of muscle coordination. Global metrics showed a good agreement with respect to the clinical score UPDRS-III. Graph theory applied to EMG analysis opens new perspectives in the study of motor control strategies during gait and confirms the efficacy of DBS in alleviating motor symptoms of PD patients.

Accurate localization of the SubThalamic Nucleus (STN) during Deep Brain Stimulation (DBS) surgery is critical for therapeutic efficacy and is commonly supported by intraoperative MicroElectrode Recordings (MERs). While deep learning approaches have shown promising performance in automatic STN identification, their limited transparency hinders their clinical adoption. In this work, we present an interpretable deep learning pipeline for MERs classification coupled with a structured validation framework aimed at assessing the clinical relevance of the model reasoning. To this aim, the classification output from a patch-based convolutional neural network with self-attention is combined with Grad-CAM relevance maps. Alignment between relevance maps and manual annotations from 3 expert neurologists with 27, 20, and 11 years of experience was quantified using overlap-based metrics, while perceived transparency, usefulness, and trustworthiness were assessed through Likert-scale questionnaires. Results show competitive classification performance (0.92 ± 0.07 AUC) and consistent agreement between automatic explanations and expert reasoning, with a maximum Dice Score of 0.77 ± 0.11, alongside high clinician acceptance and perceived interpretability. These findings suggest that structured, expert-centred validation of XAI can provide a meaningful contribution toward trustworthy AI-assisted decision support systems in intraoperative neurosurgery.

INTRODUCTION:Anterior cruciate ligament (ACL) tears disrupt the neural structures within the ligament, impairing the neuromuscular control of knee-stabilizing muscles. Consequently, muscle activity patterns are a crucial area of research in return-to-sport evaluation after ACL reconstruction (ACL-R). Artificial intelligence-based methods may help identify and extract meaningful parameters from these patterns. This study aims at accurately and reliably quantifying knee muscle preactivation and timing-based cocontraction indexes (CCIs) in athletes with and without ACL-R during sports-specific landing tasks using an artificial intelligence approach. METHODS:Eleven athletes with ACL-R and 18 control athletes (CA) performed two landing tasks: single-leg hop and single-leg cross drop landing. EMG signals were recorded bilaterally from four knee-stabilizing muscles: biceps femoris (BF), semitendinosus (ST), vastus lateralis, and vastus medialis. Muscle preactivation onsets before landing and timing-based CCIs were computed through a pretrained deep-learning-based muscle activity detector (LSTM-MAD). To ensure comparability with state-of-the-art approaches, amplitude-based EMG CCIs were also computed. RESULTS:LSTM-MAD estimated muscle preactivation onset with an error under 23 ms compared with manual segmentations by three experts. During single-leg hops, ACL-R athletes exhibited significantly greater BF (ACL-R: -193 ± 12 ms; CA: -152 ± 5 ms; P = 0.002) and ST (ACL-R: -189 ± 11 ms; CA: -140 ± 4 ms; P < 0.001) preactivations and longer cocontraction durations (ACL-R: 69% ± 2%; CA: 54% ± 2%; P < 0.001) compared with CA. During single-leg cross drop landings, ACL-R athletes showed greater BF (ACL-R: -234 ± 25 ms; CA: -150 ± 11 ms; P = 0.02) preactivations and longer BF and ST cocontractions (ACL-R: 55% ± 5%; CA: 35% ± 2%; P = 0.003) compared with CA. At return to sport, ACL-R athletes demonstrated greater BF and ST preactivation and cocontraction during landing tasks. CONCLUSIONS:Integrating artificial intelligence-based methods to assess neuromuscular control could improve the effectiveness of rehabilitation protocols, facilitating safer return-to-sport decisions.

This work aims to evaluate how Deep Brain Stimulation (DBS) impacts the motor-cognitive dual-task performance of Parkinson's Disease (PD) patients. We analyzed the muscle synergies of 27 PD patients at T0 (pre-surgery), T1(3 months post-surgery), and T2 (12 months post-surgery), compared to a control group of 30 age-matched individuals, during a walking task and a motor-cognitive dual task (walking while performing a phonemic fluency task). To evaluate dual-task interference, the Dual Task Effect (DTE) of both motor and cognitive metrics was analyzed. Our findings demonstrate that DBS significantly enhances dual-task capacity, with PD patients transitioning from a detrimental "posture-second" strategy at T0 to a more efficient attentional allocation pattern post-surgery. More specifically, the average motor metric mDTEFWHM (DTE of the Full-Width-at-Half-Maximum of the muscle synergy activation coefficients) of PD patients changed from ${}{{-}}\text {}$ 12.5 $~\pm ~$ 11.5 % (T0 to ${}{{-}}\text {}$ 3.7 $~\pm ~$ 10.2 % (T1 and ${}{{-}}\text {}$ 4.5 $~\pm ~$ 8 % (T2, becoming not different from that of controls ( ${}{{-}}\text {}$ 1.1 $~\pm ~$ 12.7 %). On the other hand, the PD cognitive DTE (cDTE) at T0 was ${}{{-}}\text {}$ 12.4 $~\pm ~$ 23.2 %, not significantly different from that of controls ( ${}{{-}}\text {}$ 23.0 $~\pm ~$ 21.0 %), and remained unchanged 1 year after the DBS implant (T2: ${}{{-}}\text {}$ 11.2 $~\pm ~$ 25.9). The reduced impact of cognitive loading on motor function without compromising cognitive performance suggests enhanced attentional resource management of PD patients after DBS that may translate to improved dynamic balance and reduced fall risk in daily activities.

Objective.Deep Brain Stimulation (DBS) of the SubThalamic Nucleus (STN) is effective in alleviating motor symptoms in medication-refractory patients with Parkinson's Disease (PD). Intraoperative identification of the STN relies on MicroElectrode Recordings (MERs), typically analyzed by trained operators. However, this approach is time-consuming and subject to variability. For this reason, this study proposes ML-STIM (Machine Learning for SubThalamic nucleus Intraoperative Mapping), a ML pipeline designed to automate STN classification from MERs, ensuring high accuracy and real-time performance.Approach.ML-STIM consists of MERs pre-processing, feature extraction, and classification using a MultiLayer Perceptron. An adaptive artifact removal algorithm was optimized to balance artifacts identification and STN signal preservation, and the features were selected among those recommended in literature through correlation analysis and ReliefF ranking. The pipeline was trained and validated on a public dataset (Dataset A, 46 patients) and tested on an independent dataset (Dataset B, 36 patients), from a different surgical center, to assess generalizability. Dataset B is made publicly available as well.Main Results.ML-STIM achieved 87.8 ± 1.7% accuracy on Dataset A and 83.8 ± 1.6% accuracy on Dataset B, significantly outperforming a state-of-the-art deep learning model (ResNet-AT,p< 0.01). The artifact removal step significantly improved classification specificity (p< 0.001). ML-STIM processed raw 10-second recordings in 139.4 ± 2.1 ms, demonstrating real-time feasibility.Significance.These results confirm ML-STIM as an accurate, interpretable, and computationally efficient solution for intraoperative STN identification in DBS surgeries.

The detection of gait subphases is pivotal for a comprehensive assessment of gait quality, playing a key role in different applications such as rehabilitation programs, movement disorder diagnostics, and fall prevention strategies. However, few methods provide dynamic subphase segmentation relying solely on plantar pressure signals in real-life, unsupervised conditions. This work aims to present an open-source, flexible toolbox for the automatic detection of gait subphases, and to introduce novel digital gait biomarkers derived from subphase analysis, enabling effective monitoring of frail patients in real-world, challenging environments. A novel MATLAB toolbox for decoding gait subphases from plantar pressure signals (PIN2GPI – from Pressure INsoles to Gait Phase Identification) is described and made publicly available. To test our algorithm, the open database provided by the Mobilise-D consortium is used, focusing on walking bouts recorded through pressure insoles in an unsupervised setting during free activities of daily living (lasting approximately 2.5 h). We extracted relevant gait parameters from a population of 32 elderly subjects: 14 frail patients after Proximal Femur Fracture (PFF) and 18 older Healthy Adults (HA). On average, PFF patients showed, with respect to HA, a reduced number of gait cycles (1059 ± 201 vs. 2076 ± 246; p = 0.006), percentage of time spent walking (9.1 ± 1.7

Fall-risk assessment of frail individuals is pivotal to implement fall prevention campaigns. Within the framework of the ongoing project MOVEWISE (Mobility Observation Via Wearable Integrated Sensor Evaluation), the aim of this work is to introduce digital gait biomarkers for monitoring frail patients in an ecological (but challenging) scenario. The open database provided by the Mobilise-D consortium was used to test our algorithm. We analyzed walking bouts recorded through pressure insoles in an unsupervised setting, during free activities of daily living (lasting approximately 2.5 hours). We extracted relevant gait parameters from a population of 32 elderly subjects (14 frail patients after Proximal Femur Fracture (PFF) and 18 older Healthy Adults (HA)). On average, PFF patients showed a reduced number of gait cycles (PFF: 524 ± 100, HA: 1030 ± 123, $\boldsymbol{p}=\boldsymbol{0.006})$, a reduced cadence (PFF: 35.5 ± 1.8 cycles/min, HA: 42.3 ± 1.3 cycles/min, p = 0.005), an increased percentage of atypical gait cycles (PFF: 0.90 $\pm$ 0.23 %/cycles/min, HA: 0.38 ± 0.06 %/cycles/min, $\boldsymbol{p}=\boldsymbol{0.046})$, and more asymmetrical gait phases, significantly different for Flat-Foot contact (PFF: 6.5 $\pm \boldsymbol{1.3}$ % of gait cycle, HA: 2.5 ± 0.4 % of gait cycle, $\boldsymbol{p}=\boldsymbol{0.003})$ and Swing (PFF: 6.6 ± 1.6 % of gait cycle, HA: 1.7 ± 0.3 % of gait cycle, $\boldsymbol{p}=\boldsymbol{0.002})$. The proposed pipeline was able to extract informative gait parameters although the recordings were performed out-of-lab in an unsupervised environment, efficiently pinpointing key factors related to fall risk.

Deep Brain Stimulation (DBS) of the SubThalamic Nucleus (STN) is an effective electroceutical therapy for treating motor symptoms in patients with Parkinson’s disease. Accurate placement of the stimulating electrode within the STN is essential for achieving optimal therapeutic outcomes. To this end, MicroElectrode Recordings (MERs) are acquired during surgery to provide intraoperative visual and auditory confirmation of the electrode position. This work introduces a machine learning-based pipeline for real-time classification of MERs to identify the STN during DBS procedures. The pipeline, designed for high classification accuracy and real-time applicability, incorporates interpretable machine learning techniques to ensure compatibility with clinical practices. The performance of a multi-layer perceptron was evaluated both with and without an intermediate artifact removal step applied during data pre-processing. The artifact removal step significantly enhanced classification accuracy from 84.4

Digital gait monitoring is increasingly used to assess locomotion and fall risk. The aim of this work is to analyze the changes in the foot–floor contact sequences of Parkinson’s Disease (PD) patients in the year following the implantation of Deep Brain Stimulation (DBS). During their best-ON condition, 30 PD patients underwent gait analysis at baseline (T0), at 3 months after subthalamic nucleus DBS neurosurgery (T1), and at 12 months (T2) after subthalamic nucleus DBS neurosurgery. Thirty age-matched controls underwent gait analysis once. Each subject was equipped with bilateral foot-switches and a 5 min walk was recorded, including both straight-line and turnings. The walking speed, turning time, stride time variability, percentage of atypical gait cycles, stance, swing, and double support duration were estimated. Overall, the gait performance of PD patients improved after DBS, as also confirmed by the decrease in their UPDRS-III scores from 19.4 ± 1.8 to 10.2 ± 1.0 (T0 vs. T2) (p < 0.001). In straight-line walking, the percentages of atypical cycles of PD on the more affected side were 11.1 ± 1.5% (at T0), 3.1 ± 1.5% (at T1), and 5.1 ± 2.4% (at T2), while in controls it was 3.1 ± 1.3% (p < 0.0005). In turnings, this percentage was 13.7 ± 1.1% (at T0), 7.8 ± 1.1% (at T1), and 10.9 ± 1.8% (at T2), while in controls it was 8.1 ± 1.0% (p < 0.001). Therefore, in straight-line walking, the atypical cycles decreased by 72% at T1, and by 54% at T2 (with respect to baseline), while, in turnings, atypical cycles decreased by 43% at T1, and by 20% at T2. The percentage of atypical gait cycles proved an informative digital biomarker for quantifying PD gait changes after DBS, both in straight-line paths and turnings.

Study Objectives:Polysomnography (PSG) currently serves as the benchmark for evaluating sleep disorders. Its discomfort makes long-term monitoring unfeasible, leading to bias in sleep quality assessment. Hence, less invasive, cost-effective, and portable alternatives need to be explored. One promising contender is the in-ear-electroencephalography (EEG) sensor. This study aims to establish a methodology to assess the similarity between the single-channel in-ear-EEG and standard PSG derivations. Methods:The study involves 4-hour signals recorded from 10 healthy subjects aged 18-60 years. Recordings are analyzed following two complementary approaches: (1) a hypnogram-based analysis aimed at assessing the agreement between PSG and in-ear-EEG-derived hypnograms; and (2) a feature- and analysis-based on time- and frequency-domain feature extraction, unsupervised feature selection, and definition of Feature-based Similarity Index via Jensen-Shannon Divergence (JSD-FSI). Results:We find large variability between PSG and in-ear-EEG hypnograms scored by the same sleep expert according to Cohen's kappa metric, with significantly greater agreements for PSG scorers than for in-ear-EEG scorers (p < .001) based on Fleiss' kappa metric. On average, we demonstrate a high similarity between PSG and in-ear-EEG signals in terms of JSD-FSI-0.79 ± 0.06-awake, 0.77 ± 0.07-nonrapid eye movement, and 0.67 ± 0.10-rapid eye movement-and in line with the similarity values computed independently on standard PSG channel combinations. Conclusions:In-ear-EEG is a valuable solution for home-based sleep monitoring; however, further studies with a larger and more heterogeneous dataset are needed.

Bipedal locomotion was a major functional change during hominin evolution, yet, our understanding of this gradual and complex process remains strongly debated. Based on fossil discoveries, it is possible to address functional hypotheses related to bipedal anatomy, however, motor control remains intangible with this approach. Using comparative models which occasionally walk bipedally has proved to be relevant to shed light on the evolutionary transition toward habitual bipedalism. Here, we explored the organization of the neuromuscular control using surface electromyography (sEMG) for six extrinsic muscles in two baboon individuals when they walk quadrupedally and bipedally on the ground. We compared their muscular coordination to five human subjects walking bipedally. We extracted muscle synergies from the sEMG envelopes using the non-negative matrix factorization algorithm which allows decomposing the sEMG data in the linear combination of two non-negative matrixes (muscle weight vectors and activation coefficients). We calculated different parameters to estimate the complexity of the sEMG signals, the duration of the activation of the synergies, and the generalizability of the muscle synergy model across species and walking conditions. We found that the motor control strategy is less complex in baboons when they walk bipedally, with an increased muscular activity and muscle coactivation. When comparing the baboon bipedal and quadrupedal pattern of walking to human bipedalism, we observed that the baboon bipedal pattern of walking is closer to human bipedalism for both baboons, although substantial differences remain. Overall, our findings show that the muscle activity of a non-adapted biped effectively fulfills the basic mechanical requirements (propulsion and balance) for walking bipedally, but substantial refinements are possible to optimize the efficiency of bipedal locomotion. In the evolutionary context of an expanding reliance on bipedal behaviors, even minor morphological alterations, reducing muscle coactivation, could have faced strong selection pressure, ultimately driving bipedal evolution in hominins.

We introduce an open-access tool capable of automatically extracting the timing of gait events during unconstrained locomotion across different neuromotor impairments. The gait analysis interactive tool is conceived as an assistant for gait assessment studies, both in healthy participants or in people with motor impairments affecting gait symmetry, regularity, or balance, as usually encountered in patients with neurological disorders. Our open-access pipeline makes it possible to automatically identify the time of key gait events (heel strike, toe off) from a single gyroscope axis (lateral mid-axis), simplifying experimental protocols, and can easily be used in everyday life conditions. The code is user-friendly and interactive. At each stage of analysis, it allows for possible adjustments and manual corrections of undetected or mismatched events. To implement, test, and validate our algorithm, we used three different databases of gait recordings that span from healthy subjects to patients affected by Parkinson’s disease. The pipeline consists of three main sections that allow us to segment, identify, and eventually correct the events within the gait cycle. The algorithm achieved an average accuracy of 99.23

Objective. The accurate temporal analysis of muscle activations is of great importance in several research areas spanning from the assessment of altered muscle activation patterns in orthopaedic and neurological patients to the monitoring of their motor rehabilitation. Several studies have highlighted the challenge of understanding and interpreting muscle activation patterns due to the high cycle-by-cycle variability of the sEMG data. This makes it difficult to interpret results and to use sEMG signals in clinical practice. To overcome this limitation, this study aims at presenting a toolbox to help scientists easily characterize and assess muscle activation patterns during cyclical movements. Approach. CIMAP (Clustering for the Identification of Muscle Activation Patterns) is an open-source Python toolbox based on agglomerative hierarchical clustering that aims at characterizing muscle activation patterns during cyclical movements by grouping movement cycles showing similar muscle activity. Main results. From muscle activation intervals to the graphical representation of the agglomerative hierarchical clustering dendrograms, the proposed toolbox offers a complete analysis framework for enabling the assessment of muscle activation patterns. The toolbox can be flexibly modified to comply with the necessities of the scientist. CIMAP is addressed to scientists of any programming skill level working in different research areas such as biomedical engineering, robotics, sports, clinics, biomechanics, and neuroscience. CIMAP is freely available on GitHub (https://github.com/Biolab-PoliTO/CIMAP). Significance. CIMAP toolbox offers scientists a standardized method for analyzing muscle activation patterns during cyclical movements.