{"id":"a2185afb-0173-4287-98f7-1433dd5840fd","arxiv_id":"2411.14923","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":16,"one_line_summary":"An unvalidated conceptual proposal for wearable and cloud monitoring of muscular dystrophy, with a hand-weighted RiskScore formula and no test data.","lead":"The paper proposes an IoT-based remote monitoring system for muscular dystrophy patients using wearable sensors, point-of-care blood tests, and cloud machine learning. It is a conceptual design with no data or validation, so its value is limited to sketching how such a system might be assembled.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The RiskScore in Eq. (1) is internally miscalibrated: its weights sum to 1.1 and the stated 3/10, 6/10 alert thresholds do not correspond to the formula's output range, so the system's early-warning mechanism cannot work as written.","rationale":"The reader's weakest-assumption analysis identifies exactly the load-bearing weakness: the RiskScore weights and thresholds in Eq. (1) are unsupported and internally inconsistent. I stress-tested this claim and found it is not merely a calibration concern but a formal inconsistency. The weights sum to 1.1, so the score is not normalized; the threshold notation '3/10' and '6/10' has no consistent mapping to the formula's output; and a normal-state example (CPK=200, ALT=40, AST=40, EMG=0.5) already produces a score above the moderate-risk threshold. The central claim promises complete, real-time monitoring that enables early, reliable intervention; that entire promise depends on the alerting logic being clinically correct. Since the only quantitative alerting logic fails on its own terms, the paper's strongest claim is not supported.\n\nIt is important to be fair: the paper is explicitly labeled conceptual, and it does not claim to present patient data or a deployed system. The absence of empirical validation alone would be a reason to treat the proposal as unverified rather than as a demonstration. But the internal miscalibration goes further: it shows that even the paper's own minimal quantitative component cannot function as described without substantial revision. The remaining material on ARIMA, LSTM, Random Forest, and SVM is generic methodology description with no experiments, no preprocessing specifics, and no implementation evidence. The conclusion's claims about 'significantly improving clinical outcomes' are therefore overstatements relative to what the paper actually establishes.\n\nI do not see a separate load-bearing concern beyond the RiskScore inconsistency that would change the verdict. The lack of clinical data and validation is captured by the same structural critique: the central mechanism is unsubstantiated and inconsistent. One could also flag that several references are duplicated [17] and [18], and some are 'unpublished' or 'in press', but those are scholarly-hygiene issues rather than the deciding factor for the central claim. The reviewer instruction says not to manufacture concerns, and I do not think a distinctive alternative concern is needed. The reader's REJECT verdict remains appropriate because the paper's core promise—reliable real-time personalized early warning—rests on a formula that is internally inconsistent and clinically unvalidated. An UNCHANGED verdict is therefore correct.","tokens_in":10620,"tokens_out":3235,"duration_ms":33920,"concrete_test":"Evaluate Eq. (1) analytically across the normal and abnormal ranges stated in the paper: CPK from 20 to 1000 U/L, ALT/AST from 10 to 140 U/L, and EMG amplitude from 0.1 to 2.0 mV. Build a small table of RiskScore values for representative combinations, including all-normal values (CPK=200, ALT=40, AST=40, EMG=0.5) and clinically dangerous values (CPK=1000, ALT=140, AST=100, EMG=0.1). Then compare each score against the stated 3/10 and 6/10 thresholds. If any all-normal combination crosses 3/10, or if the high-risk threshold is not crossed by any plausible abnormal combination, the alert logic is falsified as written. This check requires no data, only the formula and the thresholds already given in Section III-D.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of the paper is that the proposed IoT system gives clinicians reliable, real-time, early-warning monitoring for MD patients. The only quantitative mechanism that implements this early warning is the RiskScore in Section III-D: RiskScore = 0.5 × (CPK/1000) + 0.25 × (ALT/100) + 0.25 × (AST/100) + 0.1 × EMG Amplitude (mV).\n\nThis formula has three concrete problems. First, the weights sum to 1.1, not 1.0, so the score is not a normalized risk measure. Second, the thresholds are stated as 'RiskScore > 3/10: Moderate Risk' and 'RiskScore > 6/10: High Risk', which reads as 0.3 and 0.6 if the score is a fraction, but the formula can exceed 1.0 and has no stated maximum or rescaling to a 0–10 scale. Third, using normal physiological values from the paper's own ranges, CPK = 200 U/L (upper normal), ALT = 40 U/L, AST = 40 U/L, and EMG amplitude = 0.5 mV (the stated threshold for functional decline), the RiskScore is 0.5×(0.2) + 0.25×(0.4) + 0.25×(0.4) + 0.1×(0.5) = 0.35, which crosses the 0.3 moderate-risk threshold even though every biomarker is at or near its normal limit. This means the alert tier would fire on normal or high-normal data, while the 'high risk' threshold may be unreachable in many clinically dangerous configurations, e.g., CPK = 500 U/L, ALT = 100 U/L, AST = 100 U/L, EMG = 0.1 mV gives 0.25 + 0.25 + 0.25 + 0.01 = 0.76, which is moderate rather than high under a 0.6 threshold interpretation.\n\nBecause every claimed benefit—timely intervention, personalized alerts, proactive care—depends on these thresholds being clinically correct, this miscalibration is load-bearing. The paper provides no calibration data, no derivation, and no external validation for the weights or thresholds. Without a consistent scoring scale, the system could either flood clinicians with false alerts or miss true deterioration, directly undermining the strongest claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This conceptual paper proposes an IoT-based remote monitoring system for managing Muscular Dystrophy (MD). The envisioned system combines wearable sensors (EMG, ECG/PPG, accelerometers), point-of-care testing for CPK/ALT/AST, environmental sensors, and cloud-based analytics using ARIMA, LSTM, Random Forest, and SVM models. The only concrete quantitative mechanism is a RiskScore formula in Section III-D, intended to trigger moderate-risk and high-risk alerts. The paper claims this will enable real-time, personalized interventions and improve outcomes, but it presents no implementation, no data, no experimental validation, and no clinical evidence for the risk thresholds.","tokens_in":11228,"tokens_out":2608,"duration_ms":24759,"significance":"If the system were realized and validated, the multi-modal monitoring concept could be clinically useful for a chronic, progressive condition like MD. The paper correctly identifies gaps in current remote monitoring and names relevant technologies. However, the paper's sole quantitative contribution, the RiskScore, is internally inconsistent, and the central claim that the system improves outcomes is entirely unsupported by data. The manuscript does not provide reproducible code, a prototype, a dataset, or falsifiable predictions beyond a hand-specified formula. As it stands, the paper is a system-design sketch rather than a validated predictive modeling study.","major_comments":[{"comment":"The RiskScore formula has a basic internal inconsistency: the weights 0.5, 0.25, 0.25, and 0.1 sum to 1.1, so the score is not a normalized 0-1 risk measure. The thresholds are written as 'RiskScore > 3/10' and 'RiskScore > 6/10', which would be 0.3 and 0.6 on a normalized scale, but the formula has no such range. Using the paper's own normal upper values (CPK = 200 U/L, ALT = 40 U/L, AST = 40 U/L, EMG = 0.5 mV), the score equals 0.5*(0.2) + 0.25*(0.4) + 0.25*(0.4) + 0.1*(0.5) = 0.35, which already exceeds the stated moderate-risk threshold of 0.3. The alert logic would therefore fire on entirely normal readings, undermining the system's primary early-warning function.","section":"Section III-D, Eq. (1)"},{"comment":"The paper states that 'These weights and risk thresholds are calibrated in line with a clinically aligned approach', but no calibration procedure, calibration dataset, or external source is provided. The weights and thresholds are free parameters chosen by the authors, and the same RiskScore is used both to define patient risk and to justify alert thresholds, creating a circular justification. No evidence is given that these particular weights or thresholds have any predictive or clinical validity, so the claim of 'calibration' is unsupported.","section":"Section III-D"},{"comment":"The central claim that the proposed system 'would significantly improve clinical outcomes and the quality of life for the patients' (Section V) has no supporting evidence. Section IV describes model training and validation in general terms (e.g., cross-validation, accuracy, MAE/RMSE), but the manuscript reports no dataset, no experiments, no performance metrics, and no comparison against baseline monitoring. The abstract itself labels the paper as 'conceptual', yet the conclusion makes empirical causal claims that cannot be supported without implementation and evaluation.","section":"Sections IV and V"},{"comment":"The specific alert thresholds (e.g., CPK rise > 500 U/L in 24 hours, ALT > 140 U/L, AST > 100 U/L, heart rate variability < 20 ms, SpO2 < 90%) are presented as derived from 'evidence-based guidelines from AANEM, ACR, ESC, ATS, and CDC', but no specific guideline or reference is cited for any of these cutoffs. References [7]-[11] are general society guidelines and do not contain these particular numerical thresholds. Since the system's entire value proposition is timely, reliable alerts, the provenance of these thresholds is load-bearing and currently unsubstantiated.","section":"Section III-D, 'Biochemical Markers' and threshold lists"}],"minor_comments":[{"comment":"There is a typo in the paragraph on interpretability: 'which biomarkers—Thai is, CPK levels' should read 'that is, CPK levels'.","section":"Section IV, 'Model Selection'"},{"comment":"References [17] and [18] are identical duplicates (Brown, T., & Green, M., 2019).","section":"References"},{"comment":"The caption reads 'Fig 3: This table summarizes...' but the object is a table, not a figure; the caption should be adjusted for consistency with IEEE style.","section":"Figure 3"},{"comment":"The paper uses both 'Creatine Kinase (CK)' and 'Creatine Phosphokinase (CPK)' interchangeably; the terminology should be unified throughout.","section":"Section I"},{"comment":"The sentence 'The higher its value is, the more severe the injury to muscles, all the way to 1,000 U/L and above' is ambiguous: it is unclear whether values above 1,000 U/L are considered a separate alert threshold or an extension of the normal-to-damage continuum.","section":"Section III-D"},{"comment":"Some references are incomplete or provisional, such as [4] 'unpublished', [5] 'in press', and [12] lacking a year and venue; these should be completed before submission.","section":"References"}],"recommendation":"reject","confidential_remarks":"This is a conceptual design paper with no validation. The RiskScore formula's internal inconsistency (weights sum to 1.1; thresholds incompatible with the formula's output range; normal values trigger an alert) is a load-bearing flaw that cannot be fixed by local edits because the entire alerting mechanism rests on it. The absence of any data or implementation further distances the central claims from evidence. The paper might be suitable for a position or vision track after major revision, but as a regular research paper it does not meet the bar."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague —\n\nThis is a conceptual paper, and it should be judged as one. The authors say so up front, and they do a reasonable job of assembling the standard IoT-plus-wearables-plus-ML architecture for remote monitoring and applying it to MD. To their credit, they recognize that wearables cannot measure CPK/ALT/AST and build in point-of-care testing for those biomarkers. The device/transmission table is sensible, and they cite relevant clinical guidelines for thresholds like SpO2 below 90% and heart rate above 120 bpm. As a broad system sketch, it is readable and organized.\n\nBut the actual contribution is thin. There is no data, no implementation, no validation, and no new algorithm. The only novel-looking item is the RiskScore in Eq. (1), and it does not work as written. The weights sum to 1.1 rather than 1.0; the stated thresholds (3/10 and 6/10) do not correspond to the formula's output range; and plugging in the paper's own normal values (CPK=200, ALT/AST=40, EMG=0.5 mV) gives a score above 0.3, so the moderate-risk alert would fire on normal data. The stress-test note is right: this is load-bearing, because the entire early-warning claim depends on these thresholds being clinically correct, and no calibration data or derivation is provided. Saying the weights are \"calibrated\" and \"evidence-based\" without showing a calibration procedure is not enough.\n\nThe conclusion overstates what has been shown. The system \"will\" do many things, but nothing has been built or tested. The reference list has duplicate entries and some weak items, though that is secondary to the missing evidence.\n\nI would not send this to a full peer review as is. A desk reject with detailed feedback is appropriate: the authors would need at least a fixed and justified risk score, a pilot dataset or simulation, and a validation plan before this becomes refereeable. The current audience is someone looking for a broad architecture overview, not a clinician or researcher needing evidence.","headline":"A clear conceptual outline of an IoT-plus-ML monitoring system for muscular dystrophy, but the one quantitative piece, the RiskScore, is internally inconsistent and the paper has no data or implementation to back its claims.","tokens_in":11758,"tokens_out":2444,"would_cite":false,"duration_ms":24208,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proposes an IoT-based system that streams muscular dystrophy biomarkers into personalized real-time risk alerts.","keywords":["muscular dystrophy","IoT health monitoring","wearable sensors","predictive modeling","RiskScore","remote patient monitoring","point-of-care testing","machine learning healthcare"],"falsifier":"Calculate the RiskScore for a patient with CPK 1200 U/L, ALT 120 U/L, AST 120 U/L, and EMG amplitude 0.3 mV; the formula gives 1.23, while the text says scores above 6/10 are high risk, so the comparison only makes sense if the thresholds are read as 0.6 on a unit scale—which the paper never states. A clinician testing the system would need a rule for converting the raw weighted sum to the 0-to-10 scale before the moderate and high alerts can fire.","tokens_in":10443,"feed_emoji":"🩺","tokens_out":9328,"duration_ms":78028,"temperature":0.7,"pith_summary":"This conceptual paper proposes an Internet-of-Things care-management system for muscular dystrophy (MD), a group of progressive muscle disorders. It argues that wearable sensors for muscle, cardiac, respiratory, and environmental data, combined with point-of-care blood tests for CPK and ALT/AST, can feed cloud machine-learning models that issue real-time, personalized alerts before health crises develop. The intended payoff is a shift from periodic clinic visits to continuous home monitoring, giving clinicians earlier warning of muscle damage and letting patients adjust activity, hydration, and medication day by day. The paper presents the architecture, data flows, and a composite RiskScore formula, but does not report a working implementation or clinical validation.","feed_headline":"IoT plan sends real-time risk alerts for muscular dystrophy","feed_subtitle":"Wearable sensors and pinprick blood tests would spot rising muscle damage in time for earlier treatment.","key_machinery":"The load-bearing object is the RiskScore equation, a weighted composite of four inputs CPK, ALT, AST, and EMG amplitude designed so that CPK contributes half the weight. The system pairs this score with a set of fixed clinical thresholds (CPK above 1000 U/L, ALT above 140 U/L, AST above 100 U/L, SpO2 below 90%, heart rate above 120 bpm, HRV below 20 ms, temperature above 30°C, humidity above 70%) and with machine-learning models assigned by data type: ARIMA and LSTM for CPK and other time-series trends, Random Forest and SVM for risk classification. A gateway performs local preprocessing, including a Fast Fourier Transform on EMG, before encrypted upload to the cloud, and that local filtering is what makes the real-time alert loop plausible. The RiskScore is the single number that decides whether caregiver and provider are notified at moderate or high risk.","core_discovery":"The paper's central claim is that a complete real-time picture of an MD patient's health can be assembled from a body-worn sensor network plus frequent finger-prick biomarker measurements, and that machine-learning models (ARIMA, LSTM, Random Forest, SVM) can turn this stream into predictive warnings. The proposed RiskScore is the mechanism that condenses the stream into an actionable number: $RiskScore = 0.5 \\times \\frac{CPK}{1000} + 0.25 \\times \\frac{ALT}{100} + 0.25 \\times \\frac{AST}{100} + 0.1 \\times EMG\\ amplitude\\ (mV)$, with moderate-risk alerts above 3/10 and high-risk alerts above 6/10. The paper argues that CPK, as the most sensitive muscle-damage marker, should dominate the score, and that thresholds such as CPK above 1000 U/L, SpO2 below 90%, and EMG amplitude below 0.5 mV should trigger provider notifications. If the design works as described, clinicians would see deteriorating trends such as a multi-day CPK climb before an acute episode, enabling earlier changes to medication or physiotherapy.","pith_inferences":["Before deployment, the RiskScore's 0-to-10 alert thresholds would need to be reconciled with a formula whose weights sum to 1.1, since as written the numeric scales do not match.","Because CPK rises after any muscle injury or exertion, the system would need a personal baseline and activity context; otherwise a benign post-exercise spike could be scored as moderate risk.","A testable extension would be to run the RiskScore on retrospective MD patient data with known adverse events and measure whether alert timing precedes hospitalizations."],"forward_implications":["Care could move from snapshot clinic visits to continuous home surveillance, with alerts triggered by trends rather than by scheduled tests.","A clinician could be warned when predicted CPK is about to cross 200 U/L or 1000 U/L, creating a window for medication or therapy changes before muscle damage accumulates.","Patients would receive day-specific activity, hydration, and exercise guidance based on the current RiskScore and environmental conditions.","The same architecture of wearables, cloud analytics, risk score, and alert thresholds could be reused for other progressive chronic diseases."],"supporting_citations":[{"why":"Establishes wearable sensor systems as the monitoring base for chronic disease management.","marker":"[2]"},{"why":"Motivates the IoT platform for personalized healthcare in chronic conditions.","marker":"[3]"},{"why":"Supplies clinical guidelines tying electrodiagnostic testing and biomarkers to neuromuscular monitoring.","marker":"[7]"},{"why":"Provides the clinical standard for using liver enzyme measurements to flag systemic stress.","marker":"[8]"},{"why":"Supports the heart rate variability threshold used as an autonomic dysfunction alert.","marker":"[9]"},{"why":"Supports the SpO2 and oxygen saturation thresholds used in the alert logic.","marker":"[10]"},{"why":"Provides the environmental heat and humidity guidance behind the fatigue-related alerts.","marker":"[11]"},{"why":"Grounds the choice of machine-learning models for real-time health monitoring.","marker":"[19]"},{"why":"Provides the dynamic risk-scoring concept the RiskScore extends.","marker":"[20]"},{"why":"Frames the predictive-modeling methodology for medical applications.","marker":"[25]"}],"fun_headline_variants":["IoT and wearables predict muscular dystrophy flare-ups in real time","ML-driven RiskScore alerts for muscular dystrophy monitoring","Wearable sensors and blood tests forecast muscular dystrophy crises","Real-time risk alerts for muscular dystrophy via IoT and ML"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire alert system rests on hand-picked RiskScore weights and clinical thresholds that the paper states but never derives from patient data or published calibration; if those numbers misrepresent real muscular dystrophy physiology, the system will either cry wolf or miss the deterioration it promises to catch.","fun_headline_variants_meta":{"raw":{"variants":["IoT and wearables predict muscular dystrophy flare-ups in real time","ML-driven RiskScore alerts for muscular dystrophy monitoring","Wearable sensors and blood tests forecast muscular dystrophy crises","Real-time risk alerts for muscular dystrophy via IoT and ML"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000808,"raw_usage":{"total_tokens":3526,"prompt_tokens":907,"completion_tokens":2619,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":523,"completion_tokens_details":{"reasoning_tokens":2561}},"tokens_in":523,"tokens_out":2619,"duration_ms":17326,"temperature":1.0,"reasoning_tokens":2561,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:42:57.313837+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Calculate the RiskScore for a patient with CPK 1200 U/L, ALT 120 U/L, AST 120 U/L, and EMG amplitude 0.3 mV; the formula gives 1.23, while the text says scores above 6/10 are high risk, so the comparison only makes sense if the thresholds are read as 0.6 on a unit scale—which the paper never states. A clinician testing the system would need a rule for converting the raw weighted sum to the 0-to-10 scale before the moderate and high alerts can fire.","supporting_citations":[{"cited_title":"Dynamic Risk Scoring in Chronic Disease: An IoT Approach","cited_arxiv_id":null,"evidence_quote":"Provides the dynamic risk-scoring concept the RiskScore extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Frames the predictive-modeling methodology for medical applications."},{"cited_title":"Wearable sensor systems for chronic disease management: A review,","cited_arxiv_id":null,"evidence_quote":"Establishes wearable sensor systems as the monitoring base for chronic disease management."},{"cited_title":"The application of IoT in personalized healthcare for chronic diseases,","cited_arxiv_id":null,"evidence_quote":"Motivates the IoT platform for personalized healthcare in chronic conditions."},{"cited_title":"Guidelines on the Use of Electrodiagnostic Testing and Biomarkers in Neuromuscular Disorders","cited_arxiv_id":null,"evidence_quote":"Supplies clinical guidelines tying electrodiagnostic testing and biomarkers to neuromuscular monitoring."},{"cited_title":"Standards for Monitoring Liver Enzymes and Systemic Inflammation in Chronic Conditions","cited_arxiv_id":null,"evidence_quote":"Provides the clinical standard for using liver enzyme measurements to flag systemic stress."},{"cited_title":"Heart Rate Variability: Standards of Measurement, Physiological Interpretation, and Clinical Use","cited_arxiv_id":null,"evidence_quote":"Supports the heart rate variability threshold used as an autonomic dysfunction alert."},{"cited_title":"Guidelines for the Six- Minute Walk Test and Oxygen Saturation Measurements","cited_arxiv_id":null,"evidence_quote":"Supports the SpO2 and oxygen saturation thresholds used in the alert logic."},{"cited_title":"Heat- Related Illness: Guidelines for Prevention and Response","cited_arxiv_id":null,"evidence_quote":"Provides the environmental heat and humidity guidance behind the fatigue-related alerts."},{"cited_title":"Using Machine Learning for Real- Time Health Monitoring","cited_arxiv_id":null,"evidence_quote":"Grounds the choice of machine-learning models for real-time health monitoring."}],"review_version":1}