{"id":"e8d4f76f-d9f1-4666-9db9-bf28fdaff5c6","arxiv_id":"2411.15504","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A passive shoulder exoskeleton did not change the number of muscle synergies during overhead screwing, reduced activation of the primary synergy, and altered the secondary synergy from pectoralis-dominant to deltoid-dominant.","lead":"Eight healthy men performed an overhead screwing task with and without a passive shoulder exoskeleton while muscle activity was recorded. The exoskeleton did not change the number of muscle coordination patterns, but it lowered activation of the main pattern and changed the secondary pattern, suggesting less effort but possible new coordination.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'unchanged primary synergy' result may be an artifact of joint k-means clustering across conditions; S2 comparison also relies on only 6-7 of 8 subjects.","rationale":"The paper's headline contribution is that the exoskeleton does not alter the number of synergies and existing major synergies but may induce new synergies. The strongest_claim emphasizes the unchanged primary synergy and altered secondary synergy. Both of these conclusions hinge on how synergies are labeled and aligned across conditions. The method pools all Normal and Intervention synergy vectors and runs one k-means with k=2; this joint clustering necessarily places one synergy from each condition in each cluster, so the subsequent statement that Normal S1 and Intervention S1 are highly correlated (r=0.94) is not an independent test of preservation. A condition-independent alignment could yield different assignments, especially for the secondary synergy, which is already variable across subjects (only 6/8 Normal and 7/8 Intervention subjects show the group S2 pattern). The entropy and activation-reduction results are less affected by this alignment issue, but they are secondary to the synergy-composition claim. Therefore the manuscript should not be accepted as definitive; a re-analysis with condition-independent synergy matching is required. The reader's CONDITIONAL verdict correctly identifies the synergy pipeline as the weak point; our stress-test sharpens the specific circularity and the subset-dependence of the S2 result. We do not recommend rejection because the study is an exploratory case study, the authors state limitations, and the re-analysis is straightforward.","tokens_in":14781,"tokens_out":9206,"duration_ms":82557,"concrete_test":"Perform condition-independent synergy alignment: (1) cluster the 16 Normal-condition W columns alone and label the AD/MD-dominant cluster S1; (2) assign each Intervention-condition W column to the nearest Normal cluster centroid by cosine similarity without re-clustering; (3) recompute the S1 correlation and the S2 muscle-weight profiles. If the S1 correlation falls below ~0.8, or if the S2 assignment changes for more than one subject, the claim that the primary synergy is unaltered should be revised.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Methods II.D.1, NMF is run separately per subject and condition, then all W columns are pooled and k-means clustered into two groups (Results III.A.2). Because clustering is performed on both conditions simultaneously, Normal S1 and Intervention S1 are placed in the same cluster by construction, so the reported r=0.94 between them is not independent evidence that the primary synergy is unchanged. Additionally, the S2 comparison (PM in Normal vs. MD in Intervention, r=-0.45) is based on only 6/8 (Normal) and 7/8 (Intervention) subjects whose S2 matched the group pattern; for the remaining subjects the S2 label is unreliable. These two issues directly affect the paper's headline claims about synergy preservation and induction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a case study (n=8 healthy male participants) on the effects of a passive shoulder exoskeleton (HIT-POSE) on muscle synergies during a simulated overhead screwing task. Using non-negative matrix factorization (NMF) and EMG topographic maps, the authors report that the exoskeleton does not change the number of synergies (n = 2), leaves the first synergy (AD/MD) unchanged, changes the second synergy (from PM to MD), significantly reduces activation of the first synergy, and increases the entropy of the EMG topographic map. The paper is framed as a first investigation of shoulder-exoskeleton effects on muscle synergy during overhead work.","tokens_in":14905,"tokens_out":5184,"duration_ms":46843,"significance":"If the central conclusions were robust, this would be a useful contribution to exoskeleton assessment, being among the first to combine NMF-based synergy analysis and EMG topographic entropy for a shoulder exoskeleton in an occupational task. The decrease in first-synergy activation is supported by several metrics (activation profile, Recr, Ad) with p-values below 0.05, and the topographic mean and entropy differences are highly significant (p < 0.001 and p < 0.01, respectively). The authors also explicitly enumerate limitations (small sample, case-study scope, limited neurophysiological correlation). However, the main preservation/induction claims about synergies are weakened by the analysis pipeline, as described in the major comments.","major_comments":[{"comment":"The claim that Normal S1 and Intervention S1 are identical (r = 0.94) is not supported as independent evidence because the synergy identification step pools all W columns from both conditions and clusters them jointly with k-means. Since the clustering is performed on both conditions simultaneously, Normal S1 and Intervention S1 are placed in the same cluster by construction, making the subsequent correlation between them partly an artifact of the assignment rule. To test preservation of the primary synergy, the synergies should be identified independently within each condition (or matched after independent clustering) and then compared, reporting the distribution of similarity across subjects. As it stands, the headline claim that the exoskeleton 'does not alter existing major synergies' rests on a circular step.","section":"II.D.1 and III.A.2"},{"comment":"The comparison of the second synergy is based on only 6/8 participants in Normal and 7/8 in Intervention, with the remaining participants excluded after 'visually checking.' The reported negative correlation (r = -0.45) between Normal S2 and Intervention S2 is computed on this subset and is not accompanied by a statistical test (e.g., whether r differs from zero). Because the S2 change is the only evidence for the claimed 'induction of a new synergy,' the authors should either include all subjects with a principled labeling procedure or explicitly present the subset analysis as exploratory and temper the corresponding conclusion.","section":"III.A.2 and Fig. 4(d)-(e)"},{"comment":"The statistical analysis applies many paired tests across activation profiles, Recr, Ad, topographic mean, CoGx/CoGy, and entropy, without multiple-comparison correction. Several of the reported p-values (p = 0.0371, p = 0.03, p = 0.04) are close to the 0.05 threshold and would not survive a Bonferroni correction for the number of comparisons. In addition, the conclusions depend on several arbitrary thresholds (VAF global >90% and per-muscle >75% in II.D.1, activation duration >0.5 in Eq. (5), and the acceleration variance threshold in II.C.1); no sensitivity analysis is reported. The authors should at least state the number of comparisons and report adjusted or exact p-values, and ideally show that the main results (synergy count = 2, S1 unchanged, S2 different) are robust to reasonable variations of these thresholds.","section":"II.E, III.A.3-4, and III.B"}],"minor_comments":[{"comment":"The notation for At is inconsistent with the text: the equation uses 'XOR' while the text says 'OR operation,' and the symbol At is later called Ad in the Results. Please define the operation and unify the symbol.","section":"Eq. (5)"},{"comment":"The caption refers to 'the mean of the normalized activation profiles (Wnorm)' but the figure displays Hnorm activation profiles; Wnorm is the synergy weight matrix. Please correct the caption.","section":"Fig. 4(f) caption"},{"comment":"The sentence 'The objective of this study was to systematically how shoulder exoskeleton...' is missing a verb (e.g., 'investigate'). Please revise.","section":"Introduction, last paragraph"},{"comment":"The normalization in Eq. (2) divides each element by the sum across k (synergies), but the index i runs over muscles and the matrix is m x s; this is not the standard column normalization. Please clarify the intended normalization.","section":"II.D.1, Eq. (2)"},{"comment":"The null-distribution construction for the non-negative regression is described very briefly; it is unclear how the p-values were obtained and what exact comparison was made. Please provide more detail or move the merging/fractionation interpretation to the Discussion as clearly exploratory.","section":"III.A.5"},{"comment":"Reference [43] (Li and Qin, 'Evolutionary LSTM... sleep prediction') appears unrelated to the acceleration variance threshold for task detection; please verify the citation.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The study is a case series from the group that developed the HIT-POSE device, and no conflict-of-interest statement is included. Given the small sample size and the circularity in the synergy matching, the manuscript would be better positioned as a preliminary/exploratory study. The journal should ensure that the device-evaluation claims do not exceed the evidence, and should consider whether the present statistical treatment is sufficient for the stated conclusions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the HIT-POSE overhead work paper. It's a legitimate first look at whether a passive shoulder exoskeleton changes muscle synergy, not just single-muscle EMG. The main solid finding: activation of the dominant AD/MD synergy drops significantly with the exoskeleton, and the topographic mean decreases too. That is consistent and useful.\n\nThe paper also does some things well: it uses both NMF and topographic entropy, reports p-values for the main comparisons, and is honest about being a case study with eight healthy men and one task.\n\nThe problems are in the synergy-sorting and the 'new synergy' conclusion. They run NMF per subject per condition, but then pool all W columns and k-means cluster into two groups before labeling S1/S2 by AD/MD dominance. So the r=0.94 between Normal S1 and Intervention S1 is partly built into the clustering step; it is not independent evidence that the primary synergy is unchanged. The S2 comparison is even thinner: only 6/8 Normal and 7/8 Intervention subjects had a reliable S2 at all, and on that subset the correlation is -0.45. That's a weak basis for 'induces new synergies.' Also, the topographic-map methods are ambiguous: the text says RMS features are averaged across individuals to build one map per condition, yet the results report subject-level means and p-values. That needs clarification. The VAF and activation thresholds are arbitrary, and there's no multiple-comparison correction, but for a case study those are less serious than the sorting issue.\n\nThe central activation-reduction finding survives these concerns; the synergy-number and synergy-identity claims are mostly descriptive and should be read as hypothesis-generating.\n\nWho is this for? Someone working on exoskeleton evaluation or muscle synergy methods. It is not a definitive answer. It deserves a serious referee because the question is timely and the data, while small, were collected with a real device and a real task. A good reviewer will ask for the per-subject synergy plots, a rerun of the sorting without pooling conditions, and a clear description of the topographic statistics.\n\nMy recommendation: engage with it, but treat the 'new synergy' language as a suggestion, not a result. If I were the editor, I'd send it to review and ask for those clarifications.","headline":"A plausible first case study that shoulder exoskeletons reduce activation of the primary muscle synergy, but the 'induces new synergy' claim rests on a fragile and partly circular synergy-sorting step.","tokens_in":15487,"tokens_out":4634,"would_cite":false,"duration_ms":40375,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Overhead exoskeleton keeps two-synergy plan, cuts muscle effort","keywords":["muscle synergy","passive shoulder exoskeleton","overhead work","non-negative matrix factorization","EMG topographic map","surface electromyography","screwing task","motor coordination"],"falsifier":"Re-run the same overhead screwing protocol with a stricter or alternative synergy-selection rule, for example requiring total VAF above 95% or per-muscle VAF above 90%, and with synergies aligned by similarity rather than by AD/MD dominance. If the optimal count becomes three in either condition, or if the first-synergy correlation between conditions drops below the values reported, the paper's central claim of unchanged synergy number and preserved primary synergy would be falsified. A second decisive check would be a sham-exoskeleton condition with no assistive torque: if the second-synergy shift from PM to MD still appears, the assistance itself is not what induces the new synergy.","tokens_in":14561,"feed_emoji":"💪","tokens_out":7756,"duration_ms":61957,"temperature":0.7,"pith_summary":"This case study asks whether a passive shoulder exoskeleton changes how the nervous system coordinates muscles during overhead work. Eight healthy men performed the same overhead screwing task with and without the HIT-POSE exoskeleton while eight shoulder and trunk muscles were recorded by surface EMG. The paper argues that the exoskeleton leaves the basic organization of movement intact: the number of muscle synergies stays at two and the dominant shoulder-flexion synergy (anterior and middle deltoid, $r=0.94$) is unchanged. What changes is the secondary synergy, whose leading muscle shifts from pectoralis major to middle deltoid, and the amount of neural drive, since the first synergy's activation profile, average recruitment, and activation duration all decrease significantly. The EMG topographic maps add that the exoskeleton lowers overall muscle activation and makes the spatial distribution more uniform, which the authors read as a sign of reduced effort without added motor complexity.","feed_headline":"Overhead exoskeleton keeps two-synergy plan, cuts muscle effort","feed_subtitle":"Wearing the exoskeleton lowered primary-synergy activation and swapped the second synergy's main muscle.","key_machinery":"The analysis rests on non-negative matrix factorization (NMF) of eight-channel surface EMG envelopes, written as $E=W\\times H$, where $W$ holds each muscle's weight in each synergy and $H$ holds each synergy's activation over time. The number of synergies is chosen by variance accounted for: the smallest $s$ with total VAF above 90% and each muscle's VAF above 75%. Because NMF output ordering is arbitrary, the paper aligns synergies across subjects and conditions with k-means clustering followed by manual labeling based on anterior/middle deltoid dominance, naming the AD/MD module the first synergy. A second tool, the EMG topographic map, stacks averaged RMS features across the eight channels and over time, and quantifies the resulting image by mean value, center-of-gravity coordinates, and Shannon entropy; these indices capture the spatial uniformity and timing of muscle activation. The regression-based similarity measures ($R$, $S_s$, $\\sigma_{S_s}$) are used to test whether assisted-condition synergies arise by merging or fractionating the normal-condition synergies.","core_discovery":"The paper's central claim is that wearing the HIT-POSE passive shoulder exoskeleton during overhead screwing does not change the number or the primary structure of muscle synergies, but it does change the secondary synergy and the magnitude of neural activation. With non-negative matrix factorization, two synergies explained the eight-channel EMG in both conditions under the variance-accounted-for criterion. The first synergy, dominated by anterior and middle deltoid, was essentially identical across conditions ($r=0.94$); its activation profile, average recruitment level, and activation duration fell significantly in the assisted condition ($p<0.05$). The second synergy was not the same: its highest-weight muscle changed from pectoralis major to middle deltoid and the two were weakly negatively correlated ($r=-0.45$), which the authors interpret as the exoskeleton inducing a new synergy rather than merely scaling the old one. In the EMG topographic maps, the mean value dropped ($p<0.001$) and entropy rose ($p<0.01$), while the center of gravity did not shift, supporting the conclusion that the exoskeleton reduces and homogenizes muscle activation without changing which muscle groups dominate or when they are active.","pith_inferences":["A direct test of the 'new synergy' interpretation would be to record the same screwing task with a passive exoskeleton that applies no torque; if the second-synergy shift from PM to MD persists in a sham condition, it may stem from the device's physical interface rather than from assistance itself.","The entropy increase may partly reflect the normalization step, which divides all RMS values by the global maximum across conditions; re-running the topographic analysis on per-condition normalization or on raw scaled amplitudes would clarify whether the exoskeleton truly homogenizes activation or simply reduces its overall amplitude.","Because the study enrolled only eight healthy right-handed men, the key invariance claims (same synergy count, preserved first synergy) should be treated as hypotheses about the general worker population; extending the protocol to women, left-handed workers, and fatigued states would show whether the motor adaptation generalizes.","If the synergy-count invariance holds for other overhead tasks, it would suggest that passive shoulder exoskeletons act by scaling and rerouting existing motor modules rather than by recruiting new ones, giving device designers a concrete target: minimize unintended shifts in non-primary synergies through interface and torque-profile design."],"forward_implications":["For overhead screwing tasks, wearing the HIT-POSE exoskeleton does not add a new muscle-synergy module, so the motor system appears able to perform the task with the same coordination complexity as without assistance.","The primary shoulder-flexion synergy (AD and MD) is preserved under assistance, indicating that the dominant neural command for the task is not rewritten by the exoskeleton.","The change of the second synergy from a pectoralis-major-dominant to a middle-deltoid-dominant pattern implies that assistance can shift the stabilizing role among muscles, possibly creating a new synergy without increasing the synergy count.","Significant reductions in the first synergy's activation profile, average recruitment level, and activation duration suggest that the exoskeleton lowers the neural drive to shoulder agonist muscles and may delay fatigue during overhead work.","The topographic-map results (lower mean, higher entropy, unchanged center of gravity) indicate that the exoskeleton reduces the overall level of monitored muscle activation and spreads it more uniformly while preserving the timing and dominant muscle groups of the task."],"supporting_citations":[{"why":"Supplies the HIT-POSE exoskeleton's design, assistive-torque profiles, and the normalization paradigm used in the experiment; the intervention device comes from this work.","marker":"[41]"},{"why":"Establishes non-negative matrix factorization as the most appropriate method for extracting muscle synergies, justifying the central decomposition used here.","marker":"[35]"},{"why":"Provides the k-means clustering approach used to sort the unordered NMF synergies into consistent first and second synergies across subjects and conditions.","marker":"[33]"},{"why":"Previous exoskeleton-assisted walking study in multiple sclerosis that reported unchanged existing synergies with an induced new synergy; the paper's main comparative reference for its synergy findings.","marker":"[30]"},{"why":"Introduces the EMG topographic map metrics (mean, center of gravity, entropy) for exoskeleton evaluation that the spatial analysis directly applies.","marker":"[39]"},{"why":"Defines the plasticity of muscle synergies through fractionation and merging, the framework used to interpret the regression and similarity results.","marker":"[47]"},{"why":"Together with [45], supplies the VAF thresholds (global above 90%, per-muscle above 75%) used to select the number of synergies.","marker":"[44]"},{"why":"Together with [44], supplies the VAF thresholds used for synergy-number selection and contextualizes synergy count as a complexity measure in neurological conditions.","marker":"[45]"}],"fun_headline_variants":["Shoulder exo cuts neural activation, swaps second muscle synergy","Passive exo lowers effort, keeps two synergies, remixes secondary","Exo reduces muscle drive, leaves synergy count unchanged","Overhead exo: less neural effort, same synergy plan, new secondary"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusions depend on the procedural choices of how many synergies to count (VAF above 90% globally and 75% per muscle) and how to label which synergy is which (k-means clustering plus manual AD/MD dominance); if those choices were different, the claims that the count stays at two and the first synergy is identical could change.","fun_headline_variants_meta":{"raw":{"variants":["Shoulder exo cuts neural activation, swaps second muscle synergy","Passive exo lowers effort, keeps two synergies, remixes secondary","Exo reduces muscle drive, leaves synergy count unchanged","Overhead exo: less neural effort, same synergy plan, new secondary"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000263,"raw_usage":{"total_tokens":1682,"prompt_tokens":1108,"completion_tokens":574,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":724,"completion_tokens_details":{"reasoning_tokens":499}},"tokens_in":724,"tokens_out":574,"duration_ms":5421,"temperature":1.0,"reasoning_tokens":499,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:12:54.673201+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same overhead screwing protocol with a stricter or alternative synergy-selection rule, for example requiring total VAF above 95% or per-muscle VAF above 90%, and with synergies aligned by similarity rather than by AD/MD dominance. If the optimal count becomes three in either condition, or if the first-synergy correlation between conditions drops below the values reported, the paper's central claim of unchanged synergy number and preserved primary synergy would be falsified. A second decisive check would be a sham-exoskeleton condition with no assistive torque: if the second-synergy shift from PM to MD still appears, the assistance itself is not what induces the new synergy.","supporting_citations":[{"cited_title":"A novel passive occupational shoulder exoskeleton with adjustable peak assistive torque angle for overhead tasks,","cited_arxiv_id":null,"evidence_quote":"Supplies the HIT-POSE exoskeleton's design, assistive-torque profiles, and the normalization paradigm used in the experiment; the intervention device comes from this work."},{"cited_title":"Non-negative matrix factorisation is the most appropriate method for extraction of muscle synergies in walking and running,","cited_arxiv_id":null,"evidence_quote":"Establishes non-negative matrix factorization as the most appropriate method for extracting muscle synergies, justifying the central decomposition used here."},{"cited_title":"Muscle synergy assessment during single-leg stance,","cited_arxiv_id":null,"evidence_quote":"Provides the k-means clustering approach used to sort the unordered NMF synergies into consistent first and second synergies across subjects and conditions."},{"cited_title":"Evaluation of muscle synergy during exoskeleton- assisted walking in persons with multiple sclerosis,","cited_arxiv_id":null,"evidence_quote":"Previous exoskeleton-assisted walking study in multiple sclerosis that reported unchanged existing synergies with an induced new synergy; the paper's main comparative reference for its synergy findings."},{"cited_title":"Effect analysis of wearing an lumbar exoskeleton on coordinated activities of the low back muscles using semg topographic maps,","cited_arxiv_id":null,"evidence_quote":"Introduces the EMG topographic map metrics (mean, center of gravity, entropy) for exoskeleton evaluation that the spatial analysis directly applies."},{"cited_title":"Plasticity of muscle synergies through fractionation and merging during development and training of human runners,","cited_arxiv_id":null,"evidence_quote":"Defines the plasticity of muscle synergies through fractionation and merging, the framework used to interpret the regression and similarity results."},{"cited_title":"Common muscle synergies for balance and walking,","cited_arxiv_id":null,"evidence_quote":"Together with [45], supplies the VAF thresholds (global above 90%, per-muscle above 75%) used to select the number of synergies."},{"cited_title":"Muscle synergies and complexity of neuromuscular control during gait in cerebral palsy,","cited_arxiv_id":null,"evidence_quote":"Together with [44], supplies the VAF thresholds used for synergy-number selection and contextualizes synergy count as a complexity measure in neurological conditions."}],"review_version":1}