{"schema_version":1,"research_id":"eoa_inverse_innovation_exp04_external_evaluation_20260802","source_assessment_id":"negative_space_design__neuroscience:PROPOSAL_FIRST:v0","cell_id":"negative_space_design__neuroscience","search_queries":["closed-loop neural stimulation inter-stimulus interval carryover adaptation baseline response overlap study","adaptive experimental design neuroscience null trials catch trials no stimulus baseline","deep brain stimulation washout period carryover closed loop stimulation protocol","NIH BRAIN Initiative closed loop stimulation adaptive research funding need","site:pmc.ncbi.nlm.nih.gov electrical stimulation artifact recovery inter stimulus interval neural response carryover","site:pubmed.ncbi.nlm.nih.gov neural stimulation interstimulus interval adaptation evoked potential randomized","closed-loop stimulation refractory period algorithm neural stimulation blanking interval patent product documentation","responsive neurostimulation detection stimulation refractory period therapy limit manual","fMRI event related design jitter null events efficiency baseline primary study Dale 1999","rapid event related fMRI null events jittered interstimulus interval overlap deconvolution","optogenetic stimulation inter trial interval recovery adaptation neuroscience protocol study","closed loop neuroscience adaptive stimulus selection intertrial interval constraint software","site:fda.gov investigational device exemption changes investigational plan IRB approval neurostimulator study","site:hhs.gov OHRP changes approved research IRB approval before implementation 45 CFR 46","site:olaw.nih.gov IACUC significant changes animal activity approval stimulation protocol","site:braininitiative.nih.gov neuroethics invasive noninvasive neural stimulation research safety","site:bls.gov OEWS May 2025 software developers median annual wage medical scientists data scientists","site:bls.gov occupational outlook handbook medical scientists pay 2025 software developers pay","NIH salary cap 2026 official notice","university research computing recharge rates 2026 programmer hourly rate neuroscience","BIDS specification events.tsv trial_type omitted stimulus response time official","Brain Imaging Data Structure events.tsv value n/a onset duration official standard","NWB TimeIntervals trials omitted stimuli intervals specification","\"Learning to control the brain through adaptive closed-loop patterned stimulation\" full text","\"Influence of inter-stimulus interval on auditory evoked potentials\" full text","\"Stochastic designs in event-related fMRI\" full text"],"sources":[{"source_id":"S01","title":"Learning to Control the Brain through Adaptive Closed-Loop Patterned Stimulation","publisher":"Journal of Neural Engineering / IOP Publishing","url":"https://www.timbuschman.com/_files/ugd/568973_dda3daa9a6ea4ad8bd4b580f543adaab.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2020-10-13","accessed_at":"2026-08-02","claims_supported":["Adaptive closed-loop stimulation was implemented in vivo with real-time recording and stimulation.","The learned electrical-stimulation response was affected by preceding visual stimulation, directly demonstrating adaptation/history dependence.","A custom MATLAB interface provided a concrete implementation precedent."]},{"source_id":"S02","title":"Influence of inter-stimulus interval on auditory evoked potentials","publisher":"IEEE Engineering in Medicine and Biology Society / PubMed","url":"https://pubmed.ncbi.nlm.nih.gov/17282637/","source_class":"PRIMARY_RESEARCH","publication_date":"2005","accessed_at":"2026-08-02","claims_supported":["In paired-stimulus human recordings, the succeeding auditory evoked potential was completely inhibited when the inter-stimulus interval was shorter than 150 ms.","Neural responses can materially depend on preceding-stimulus timing."]},{"source_id":"S03","title":"Stochastic Designs in Event-Related fMRI","publisher":"NeuroImage / Academic Press","url":"https://web.mit.edu/swg/ImagingPubs/experimental-design/friston_stochastic.1999.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"1999-11","accessed_at":"2026-08-02","claims_supported":["Trial-free periods that allow baseline attainment were already formalized as null events in rapid event-related neuroscience designs.","Null-event frequency, minimum stimulus-onset asynchrony, and the estimand jointly affect design efficiency.","The optimal timing design differs for absolute evoked responses and differential responses."]},{"source_id":"S04","title":"The ERP omitted stimulus response to “no-stim” events and its implications for fast-rate event-related fMRI designs","publisher":"NeuroImage / Elsevier","url":"https://www.sciencedirect.com/science/article/abs/pii/S1053811903000120","source_class":"PRIMARY_RESEARCH","publication_date":"2003-04","accessed_at":"2026-08-02","claims_supported":["No-stim events are established tools for addressing overlap from adjacent trials.","An omitted expected stimulus can itself evoke a measurable neural response, so silence cannot automatically be treated as a neutral baseline.","Omission percentage and stimulus-onset asynchrony affect this risk."]},{"source_id":"S05","title":"RNS System User Manual","publisher":"NeuroPace","url":"https://www.neuropace.com/wp-content/uploads/2021/02/rns-system-manual-300m.pdf","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"2021-02","accessed_at":"2026-08-02","claims_supported":["A commercial responsive neurostimulator implements a post-episode interval during which a newly detected episode does not trigger stimulation.","The system implements daily therapy limits, configurable pulse intervals, testing, stored electrocorticography, and an operator halt control.","The manual warns that stimulation settings should be adjusted if afterdischarge activity appears."]},{"source_id":"S06","title":"NWB Format Specification 2.10.0: Format Overview","publisher":"Neurodata Without Borders","url":"https://nwb-schema.readthedocs.io/en/stable/format.html","source_class":"STANDARD","publication_date":"2026-07","accessed_at":"2026-08-02","claims_supported":["NWB represents trials, epochs, invalid intervals, continuous electrical recordings, and experimental event tables.","Custom interval and event metadata can encode intentional pauses, faults, provenance, and resumption states without stopping acquisition.","The standard distinguishes trial intervals from invalid times, supporting explicit meaning-of-absence coding."]},{"source_id":"S07","title":"Award Information: Simplified, Scalable 24-hour adaptive deep brain stimulation for Parkinson’s disease","publisher":"U.S. Department of Health and Human Services TAGGS / NINDS","url":"https://taggs.hhs.gov/Detail/AwardDetail?arg_AwardNum=UH3NS140730&arg_ProgOfficeCode=137","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2026-05-19","accessed_at":"2026-08-02","claims_supported":["NINDS is an identifiable funder of adaptive stimulation work and UCSF is an identifiable adopter.","The funded project expressly identifies multi-timescale neural-signal fluctuation, nonstationarity, programming complexity, and data-driven optimization as problems.","The FY2026 award action was $2,095,406, demonstrating material institutional commitment to adaptive stimulation research, though not demand for this exact pause mechanism."]},{"source_id":"S08","title":"Investigational Device Exemption (IDE)","publisher":"U.S. Food and Drug Administration","url":"https://www.fda.gov/medical-devices/premarket-submissions-selecting-and-preparing-correct-submission/investigational-device-exemption-ide","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"undated; current page","accessed_at":"2026-08-02","claims_supported":["Nonexempt clinical evaluation of an investigational device requires an approved IDE before initiation.","Investigational device studies require IRB-approved plans; significant-risk studies also require FDA approval.","Informed consent, monitoring, records, and reports are required."]},{"source_id":"S09","title":"IDE Application","publisher":"U.S. Food and Drug Administration","url":"https://www.fda.gov/medical-devices/investigational-device-exemption-ide/ide-application","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"undated; current page","accessed_at":"2026-08-02","claims_supported":["Changes to an investigational plan generally require appropriate supplemental and IRB approval before implementation.","Protocol changes must be assessed for effects on scientific soundness, planned analysis, risk-benefit, rights, safety, and welfare.","Software, firmware, control-mechanism, and clinical-protocol changes may be significant depending on their extent."]},{"source_id":"S10","title":"Investigator Responsibilities FAQs","publisher":"HHS Office for Human Research Protections","url":"https://www.hhs.gov/ohrp/regulations-and-policy/guidance/faq/investigator-responsibilities/index.html","source_class":"OFFICIAL_GUIDANCE","publication_date":"undated; current page","accessed_at":"2026-08-02","claims_supported":["Investigators must obtain IRB approval before nonexempt human-subject research begins.","An approved study modification requires prior IRB approval unless needed to eliminate an immediate hazard.","The principal investigator generally bears overall study responsibility."]},{"source_id":"S11","title":"Protocol Review Resources: Significant Changes to Animal Activities","publisher":"NIH Office of Laboratory Animal Welfare","url":"https://grants.nih.gov/policy-and-compliance/policy-topics/animal-welfare/compliance-topics/protocol-review","source_class":"OFFICIAL_GUIDANCE","publication_date":"2026-04-13","accessed_at":"2026-08-02","claims_supported":["IACUC review applies to significant changes in approved animal activities.","Changes in the duration, frequency, type, or number of procedures can require institutionally defined review or veterinary consultation.","Animal-welfare and personnel-safety effects remain approval considerations."]},{"source_id":"S12","title":"National employment and wage data by occupation, May 2025","publisher":"U.S. Bureau of Labor Statistics","url":"https://www.bls.gov/news.release/ocwage.t01.htm?mod=article_inline","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"2026-05","accessed_at":"2026-08-02","claims_supported":["The May 2025 mean annual wage was $148,100 for software developers, $126,800 for data scientists, and $116,840 for medical scientists excluding epidemiologists.","These figures provide 2026 planning anchors for labor-equivalent costs."]},{"source_id":"S13","title":"Center for Research Informatics Price Guide, FY26","publisher":"University of Chicago Center for Research Informatics","url":"https://cri.uchicago.edu/wp-content/uploads/2025/07/CRI-BIO.Recharge_Rates.FY26.pdf","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"2025-07-01","accessed_at":"2026-08-02","claims_supported":["Internal FY26 application-development and bioinformatics-analysis rates were $160 per hour.","External academic application-development rates were $246–$321 per hour depending on service line.","The listed services include application development, data warehousing, de-identification, IRB-writing assistance, storage, and project support."]},{"source_id":"S14","title":"A Deep Brain Stimulation Trial Period for Treating Chronic Pain","publisher":"Journal of Clinical Medicine / PubMed Central","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7600449/","source_class":"PRIMARY_RESEARCH","publication_date":"2020-10","accessed_at":"2026-08-02","claims_supported":["DBS trial methodology already treats wash-in and wash-out periods as functions of expected stimulation effects.","Longer wash-out periods increase confidence that later observations are not carryover from earlier stimulation conditions.","Spacing therefore has established scientific value but imposes time and trial-efficiency tradeoffs."]}],"problem_evidence":{"support":"STRONG","rationale":"Multiple primary studies visibly establish the underlying phenomenon: preceding stimuli and inter-stimulus timing alter later neural responses, rapid schedules create response overlap, and an adaptive closed-loop experiment demonstrated stimulation-response adaptation. The issue matters because it can confound attribution of an observed response to the current stimulus. The prevalence of harmful over-scheduling across closed-loop laboratories is not quantified, and the evidence spans different modalities and timescales rather than validating one universal recovery rule.","source_ids":["S01","S02","S03","S04","S14"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"NINDS is a concrete funder and UCSF a concrete adopter working on adaptive stimulation; their funded project expressly identifies neural nonstationarity, programming complexity, and data-driven optimization needs. PIs control scientific scheduling, while IRBs, FDA, or IACUCs may authorize live changes. No source expresses demand for the candidate's exact package of state-triggered, non-backfillable, explicitly typed null epochs.","source_ids":["S07","S08","S09","S10","S11"]},"prior_art":{"proximity":"ESTABLISHED_PRACTICE","closest_analogues":[{"name":"Stochastic null events and trial-free baseline periods","similarity":"Rapid event-related neuroscience designs have long inserted probabilistic no-stimulus periods so baseline and event-specific responses can be estimated despite overlap.","remaining_difference":"The candidate applies this established timing device inside a closed-loop stimulation generator, adds state-dependent entry criteria, prevents automatic backfilling, and preserves a typed audit trail.","source_ids":["S03","S04"]},{"name":"DBS wash-out periods","similarity":"DBS protocols already pause stimulation between conditions to reduce carryover and improve attribution.","remaining_difference":"Conventional wash-out is commonly fixed or block-level; the candidate proposes bounded recovery-triggered selection by an online experiment generator.","source_ids":["S14"]},{"name":"NeuroPace post-episode interval and therapy limits","similarity":"A first-party responsive-neurostimulation product already implements a protected waiting interval, suppresses bursts during that interval, limits daily therapies, stores recordings, and supplies halt controls.","remaining_difference":"The product feature is a clinical episode-management/safety behavior, not an experimental null-trial policy optimized and evaluated for stimulus-versus-history identifiability.","source_ids":["S05"]},{"name":"History-dependent modeling and deconvolution","similarity":"Established analysis models use timing, convolution, randomization, and history structure to separate overlapping responses without surrendering every stimulation opportunity.","remaining_difference":"The candidate creates observed no-stimulus intervals and recovery measurements rather than relying entirely on statistical separation of dense overlapping responses.","source_ids":["S03","S04"]}],"distinctive_claim_remaining":"At fixed elapsed time and unchanged safety limits, a recovery-triggered, explicitly typed, non-backfillable null-epoch policy improves held-out separation of current-stimulus effects from recent-history effects more than (a) dense scheduling plus a history/deconvolution model, (b) a fixed longer inter-trial interval, and (c) randomized null events, by enough to offset its loss of stimulated-condition coverage. This narrow claim is contrastive and falsifiable; the broad idea of using null or wash-out periods is not distinctive.","confidence":"HIGH"},"implementation_evidence":{"support":"STRONG","rationale":"Real-time adaptive stimulation, configurable protected waiting intervals, therapy limits, logging, manual halt, and standardized interval/event metadata all exist. A single-lab prototype is therefore technically credible. Workflow and data feasibility depend on synchronized acquisition continuing throughout pauses and on separately coding intentional, recovery-triggered, operator-triggered, and fault states. Human or animal deployment is conditional on applicable IRB, FDA/IDE, IACUC, sponsor, PI, and device-control authority. Safety risks remain: an omission can evoke activity, recovery-triggered sampling can condition observations on endogenous state, a wrong recovery feature can suppress needed conditions, and replacing omitted trials can increase session burden.","source_ids":["S01","S04","S05","S06","S08","S09","S10","S11"]},"scores":{"meaningful_impact":{"score":4,"rationale":"If the problem is present, better attribution of stimulus effects could materially improve causal interpretation and reproducibility; realized impact and cross-modality prevalence remain unmeasured.","source_ids":["S01","S02","S03","S14"]},"stakeholder_pull":{"score":3,"rationale":"An identifiable NIH funder and adaptive-DBS adopter express related nonstationarity and optimization needs, but not demand for this exact intervention.","source_ids":["S07"]},"incremental_advantage":{"score":2,"rationale":"Null events, wash-out intervals, refractory/post-episode intervals, and history models already cover most functional ingredients; advantage depends on an untested head-to-head result.","source_ids":["S03","S04","S05","S14"]},"distinctiveness_plausibility":{"score":2,"rationale":"Only the integrated state-triggered, typed, non-backfillable policy and its comparative identifiability claim remain plausibly distinctive after close prior art.","source_ids":["S03","S04","S05","S14"]},"technical_implementability":{"score":4,"rationale":"Adaptive closed-loop stimulation, timers, limits, logging, and interval metadata have concrete implementations; modality-specific recovery estimation and validated fault classification remain work.","source_ids":["S01","S05","S06"]},"adoption_authority_feasibility":{"score":3,"rationale":"A PI can sponsor an offline analysis and may control a laboratory schedule, but live human or animal changes require institutional and possibly FDA approval.","source_ids":["S08","S09","S10","S11"]},"evidence_readiness":{"score":3,"rationale":"A bounded archival analysis is feasible if suitable synchronized data and schedule logs exist, but the decisive incremental claim requires a randomized live comparison or equivalent prospective data.","source_ids":["S01","S03","S06"]},"safety_net_benefit":{"score":3,"rationale":"Pauses can reduce delivered stimulation and support recovery observation, and established systems show limits and halt controls; omission-evoked responses and longer-session burden prevent assuming a net safety benefit.","source_ids":["S04","S05","S14"]},"scalability":{"score":3,"rationale":"The software and metadata pattern is transferable, but recovery markers, pause bounds, scientific estimands, devices, and approvals are modality- and protocol-specific.","source_ids":["S01","S06","S08","S11"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"UNDER_10K","scope":"One archival-session feasibility analysis: data audit, lagged response model, recovery-curve estimation, three emulated timing policies, blocked cross-validation, and a short go/no-go report.","confidence":"MODERATE","assumptions":["Existing synchronized neural data, event logs, and computing access are available without acquisition or license charges.","Approximately 30–50 hours of analyst or research-software effort at an internal resource-equivalent rate near $160 per hour.","The session contains enough naturally occurring interval variation to estimate recovery; otherwise the step terminates as not data-ready."],"source_ids":["S12","S13"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Single-laboratory research prototype adding null-epoch state logic, non-backfill constraints, event typing, operator displays, replay testing, audit logs, and protocol documentation to an existing generator.","confidence":"MODERATE","assumptions":["Approximately 400–1,000 hours of application-development, data-science, QA, PI, and operator effort.","Existing acquisition and stimulation hardware are reused.","This excludes participant recruitment, implantation, device purchase, and a live trial."],"source_ids":["S01","S05","S06","S12","S13"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"One-site approved prospective pilot with software verification, protocol and regulatory submissions, operator training, monitoring, data management, analysis, and rollback readiness.","confidence":"LOW","assumptions":["Existing research device and clinical or animal infrastructure are reused; new implants, scanners, or major hardware are excluded.","One software/controls engineer plus fractional neuroscientist, PI, statistician, regulatory, QA, and operator effort over roughly 6–18 months.","The range varies sharply between noninvasive, animal, and significant-risk implanted-device studies."],"source_ids":["S08","S09","S10","S11","S12","S13"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Single-site maintenance after launch: 0.25–0.75 blended FTE for software maintenance, calibration, audit review, data stewardship, refresher training, and protocol reporting.","confidence":"MODERATE","assumptions":["No device replacement, implantation, participant-care, or added session-time cost is included.","Existing institutional storage and compute remain available.","Major algorithm or protocol revisions would return to startup or launch scope."],"source_ids":["S06","S09","S10","S12","S13"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Primary evidence shows short-interval inhibition, adaptation by preceding stimulation, overlap, and the established need for baseline/wash-out periods.","source_ids":["S01","S02","S03","S04","S14"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"UCSF and NINDS are identifiable adopter/funder entities for adaptive stimulation; PIs and the applicable IRB, FDA, or IACUC are identifiable authorizers for implementation.","source_ids":["S07","S08","S09","S10","S11"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim specifies a state-triggered protected policy and three comparators, with held-out identifiability and condition-coverage outcomes.","source_ids":["S03","S04","S05","S14"]},"bounded_next_evidence_step":{"status":"YES","reason":"A one-session archival analysis can be capped by dataset, policy count, elapsed time, metrics, and explicit stopping rules without changing live stimulation.","source_ids":["S01","S03","S06","S12","S13"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The immediate archival step introduces no stimulation or new subject exposure. Any live step remains explicitly gated by PI, IRB/IACUC, sponsor, and FDA authority as applicable.","source_ids":["S08","S09","S10","S11"]},"credible_cost_scope_and_range":{"status":"YES","reason":"Each band is scoped to a single site and anchored to 2026-relevant labor and institutional recharge data, with hardware, implantation, and participant-care exclusions stated.","source_ids":["S12","S13"]}},"next_evidence_step":"Obtain one previously collected, synchronized session with complete stimulus proposals, delivered-event codes, fault logs, continuous neural acquisition, and enough naturally occurring interval variation. Before analysis, preregister: (1) a stimulus-only model; (2) a dense-schedule history/deconvolution model; (3) a fixed-interval policy; (4) randomized null events; and (5) no more than three recovery-triggered protected-null policies. Use blocked time-series cross-validation and interval-matched resampling under the original elapsed-time and safety bounds. Primary outcomes are held-out predictive log score and reduction in covariance or condition number between stimulus and history effects; secondary outcomes are baseline stability and condition coverage. Stop as unsupported if recent-history and prestimulus-state terms improve held-out score by less than 5%, no recovery plateau lies within the permissible pause, the state-triggered policy fails to beat both fixed and randomized spacing by the preregistered margin, or projected condition-coverage loss exceeds 10%. This archival step is only a feasibility filter: it cannot establish counterfactual post-gap responses, so a later approved randomized live pilot remains necessary for causal validation.","blocking_evidence":["No public estimate establishes how often closed-loop neuroscience generators actually overfill eligible moments or how much inference is harmed across modalities.","No head-to-head study was found comparing recovery-triggered protected null epochs with fixed spacing, randomized nulls, and dense history-model correction under equal resource constraints.","A validated, modality-specific recovery signal and maximum safe/useful pause have not been supplied.","Archival dense data cannot reveal neural responses to gaps that never occurred; decisive efficacy evidence therefore requires prospective randomized timing or equivalent natural variation.","Recovery-triggered selection may introduce state-dependent sampling bias, which needs a prespecified causal/statistical treatment.","The effect of omissions themselves must be measured because no-stim events can evoke neural responses.","Live human or animal adoption requires protocol-specific authority determinations and approvals.","World novelty, patentability, freedom to operate, market size, and realized impact remain unmeasured."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"This evaluation establishes only that null events, wash-out periods, protected post-episode intervals, adaptive stimulation, and interval/event metadata already exist, while the narrow state-triggered comparative claim remains untested. It does not measure world novelty, patentability, freedom to operate, market size, or realized impact, and the bounded web search is not a patent or exhaustive literature search.","arm":"PROPOSAL_FIRST","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":true,"material_progress_observed":true,"progress_targets":["Secure one eligible archival session and its complete schedule, acquisition, omission, and fault data dictionary.","Preregister the recovery feature, permissible pause range, comparators, blocked-validation method, identifiability metric, coverage-loss ceiling, and all falsifiers before inspecting outcomes.","Demonstrate material serial dependence and a recovery plateau within the permissible pause; otherwise stop the candidate.","Show that the state-triggered policy exceeds fixed and randomized null spacing and dense history modeling by a prespecified margin without more than 10% condition-coverage loss.","Obtain a PI and oversight determination defining whether a later live pilot requires IRB, IACUC, sponsor, or FDA/IDE review.","If archival criteria pass, run an approved randomized pilot that equalizes elapsed time and directly measures post-gap responses, omission-evoked activity, safety, and operator fault recognition."],"reason":"Web evidence strongly validates timing-related carryover but also shows that null events, wash-outs, protected waiting intervals, limits, and history modeling are established. The remaining advantage is a narrow empirical comparison that cannot be resolved by more bounded web research: it requires suitable proprietary session data and ultimately prospective live testing."}}