Preliminary Minimum Reporting Requirements for In-Vivo Neural Interface Research: I. Implantable Neural Interfaces

The pace of research and development in neuroscience, neurotechnology, and neurorehabilitation is rapidly accelerating, with the number of publications doubling every 4.2 years. Maintaining this progress requires technological standards and scientific reporting guidelines to provide frameworks for communication and interoperability. The present lack of such neurotechnology standards limits the transparency, repro-ducibility, and meta-analysis of this growing body of literature, posing an ongoing barrier to research, clinical, and commercial objectives. Continued neurotechnological innovation requires the development of some minimal standards to promote integration between this broad spectrum of technologies and therapies. To preserve design freedom and accelerate the translation of research into safe and effective technologies with maximal user benefit, such standards must be collaboratively co-developed by the full range of neuroscience and neurotechnology stakeholders. This paper summarizes the preliminary recommendations of IEEE P2794 Standards Working Group, developing a Reporting Standard for in-vivo Neural Interface Research (RSNIR).


I. INTRODUCTION
N EURAL interfaces (NIs) are systems that record and/or modulate the activity of the nervous system (see Fig. 1).A broad spectrum of technological modalities for NIs has been developed over the last 50 years, including both invasive (implanted) and non-invasive systems (Fig. 1A).NIs have been shown to provide therapeutic benefit for a wide range of conditions, as well as providing powerful tools for studying nervous system physiology, improving human-machine interaction, and augmenting human capabilities [1].The rapid proliferation of neurotechnology in recent years (Fig. 1B) has produced a wealth of devices and systems with advanced neurosensing and neuromodulatory capacities, with a wide range of potential clinical and consumer applications.This diversity of NI technologies, applications, performance metrics, and experimental paradigms -along with the present lack of technological standards and reporting guidelines -has severely limited the transparency, reproducibility, and meta-analysis of this body of literature and hampered its translation into widely adopted and commercially available neurotechnologies.
The effective interpretation, aggregation, and meta-analysis of NI research thus requires more structured and extensive reporting standards to improve the overall 'information interoperability' of NI study reports and data.Several related reporting guidelines and initiatives have been enacted in recent years to improve scientific reproducibility and replicability 1 across the health and cognitive science domains [3].For example, the Enhancing the Quality and Transparency of Health Research (EQUATOR) network [4] has compiled a list of best-practice reporting guidelines specific to different types of studies, including the CONSORT guidelines for randomized clinical trials [5], the ARRIVE standard for pre-clinical animal trials [6], the PRISMA guidelines for systematic reviews and metaanalyses [7], and more.Collectively, these guidelines articulate the thorough scientific reporting of study protocols, research subjects, and outcomes, but they lack the technological specificity needed to ensure sufficiently detailed descriptions of NI 1 There remain no unified consensus definitions of "reproducibility" and "replicability" -while a majority of engineering and health science disciplines distinguish between reproducibility as the ability to derive the same results from the original data/code versus replicability as the ability to arrive at the same findings using new data [2], some fields reverse this distinction, with others use the terms interchangeably.This article shall observe the former convention.systems, methods, and results for accurate interpretation and replicability.
On a more technical level, regarding the sharing and interoperability of scientific data, the FAIR principles of findability, accessibility, interpretability, and re-usability [8] have been widely recognized in numerous neuroinformatics initiatives, including the International Neuroinformatics Coordinating Facility [9], Neurodata Without Borders [10], the NeuroImaging Data Model [11], ReproNim [12], and the Organization for Human Brain Mapping's Committee on Best Practices in Data Analysis and Sharing (COBIDAS) [13].However, these guidelines focus primarily on the formatting, analysis, and sharing of data, rather than on scientific reporting.
To address this 'standardization gap,' IEEE P2794 Standards Working Group (SWG) -spawned from the IEEE Industry Connections Activity on Neurotechnology for Brain-Machine Interfacing [14] -is currently developing a set of Reporting Standards for in vivo Neural Interface Research (RSNIR), with the primary objective of improving the scientific quality and transparency of NI research across a wide range of neurotechnological modalities.These Standards aim to establish the technical specificity necessary to achieve thorough interpretability and replicability of NI studies -and thus to improve the positive impact of NI research on the development of safe, effective, and human-beneficial neurotechnologies.
While a primary application of RSNIR will be to improve scientific publications (e.g., via adoption by high-impact journal editors), this standard is intended to serve as a reference for any neurotechnology stakeholder that seeks to improve the rigor and transparency of NI research, including regulatory bodies and funding agencies, as well as for translation of NI research into medical devices.This report previews one such set of guidelines under development, to be included in the forthcoming draft Standard.Constructive feedback is welcomed from all stakeholders, including scientific, commercial, clinical, regulatory, and end-user perspectives.

II. SCOPE
The official scope of IEEE P2794 (RSNIR) SWG is to "define the essential characteristics and parameters of in-vivo neural interface research studies (including clinical trials) to be reported in scientific and clinical literature, including both minimum reporting standards and best-practice guidelines."The RSNIR SWG has defined the scope of NIs to be addressed by the draft Standard to include all engineered systems that directly record bio-signals of neurological origin and/or directly modulate neural activity."NI research" is defined to include all studies where NI technologies are employed, either as the object of investigation or solely for recording data.More details regarding the scope and organization of SWG P2794 are provided in the Supplementary Materials, Section ( §) V.
This article specifically sets forth a preliminary minimum information standard (in the FAIR [8] sense, e.g., [15], [16]) for reporting research involving implanted NIs.The provided recommendations are the most specific regarding electrodebased NIs due to their advanced maturity and broad adoption 1 Lab bench and in vitro studies are beyond the official in vivo scope of RSNIR.Recommendations given here as reference, for complementarity to in vivo studies. 2 The PNS is classically divided into somatic and autonomic divisions, with the autonomic further delineated into parasympathetic and sympathetic sub-divisions.
relative to other NI modalities [17], [18].Because RSNIR is envisioned to complement existing Standards and consensus guidance documents, the scope of this module does not include aspects of NIs sufficiently specified by existing Standards, such as biocompatibility assessment [19] or characterization of research subjects [5], [6].

III. REPORTING TOPICS FOR IMPLANTABLE NEURAL INTERFACES
As a starting principle, RSNIR-compliant NI study reports should adhere to all applicable reporting guidelines (per the EQUATOR network [4]).The purpose of RSNIR is to expand upon these guidelines by identifying the technological and methodological details necessary to ensure clear, replicable NI reporting.Accordingly, requirements already covered in these 'parent' guidelines will not be exhaustively listed here but may be repeated for clarity and context.To promote findable, accessible reporting [8], NI research publications should specify the NI technology(s), neuroanatomic targets, use paradigms /applications, and overall study design in the publicly-accessible metadata (title, abstract, and keywords).

A. Neural Interfacing Context and Study Aims
To provide sufficient context and rationale, the background/introduction section of NI study reports should clearly identify the fundamental capabilities and limitations in the pertinent technological state-of-the-art and the scientific knowledge gaps addressed by the current study, with reference to authoritative works.Reports should specify the technological or methodological innovation(s) and scientific hypotheses proposed by the study.Testable hypotheses and additional qualitative/descriptive study aims should be stated in relation to the study's primary and secondary outcome measures.
The developmental stage of the study (technology development 1 vs. animal studies [20] vs. clinical validation [21]) should be identified per Table I.The report should indicate which NI system types and operation modes were investigated, per the IEEE NeuroEthics framework [22]: r Recording/sensing r Stimulation/neuromodulation r Closed-loop control of applications or prosthetic devices r Physical/biological modification r Neural augmentation and facilitation.
These loosely align with the brain-computer interface application scenarios originally proposed by Wolpaw and colleagues [23] and expanded upon in [1]: replacing, restoring, enhancing, supplementing, improving, and studying neurological function.Finally, the NI description should specify the target neuroanatomical structure(s) and device-tissue interface type/region.

B. NI Experimental Design and Outcome Measures
As a guiding principle, all aspects of experimental designs featuring NIs should be described in sufficient detail to permit replication by other researchers, provided use of the same NI

TABLE II REPORTING TOPICS FOR NI EXPERIMENTAL DESIGN AND OUTCOME MEASURES
1 This is important, as many drugs have effects on the nervous system which may influence NI performance, e.g., [45]. 2 For more complex stimuli, such as movies or sequences of spoken words, examples should be provided as supplementary data.system(s) and experimental setup.All NI studies must comply with consensus standards of ethical conduct, including local regulations, institutional review board approval, and the Declaration of Helsinki [24].Tables I and II list essential study characteristics to be reported, as outlined below.

1) High-Level Study Design:
The NI study description should first identify the overall experimental paradigm(s) using established study design taxonomy terminology such as prospective/retrospective cohort study, single/double-cohort, randomized controlled trial (RCT), or case-control study [25], [26].Within-subjects designs (where each participant serves as their own control, such as n-of-1 case studies [27]) are common for early clinical and pre-clinical NI research, with the main motivation to demonstrate proof-of-concept and/or subject-specific safety and efficacy of the NI prior to conducting large-scale clinical trials.Given the high tendency for individual variability, this approach demands a detailed description of the clinical and demographic characteristics of all subjects (Table II).Follow-up data collection to monitor the clinical evolution after experimental intervention is highly encouraged.
Subsequent pilot [28] and larger-scale clinical studies evaluating an intervention's effectiveness with respect to established standard therapy(s) for broader user populations typically employ between-subjects study designs, such as the "gold standard" RCT.Important for these types of experiments is the definition and recruitment of a representative control group.Blinded assessment of outcomes is strongly encouraged.In "crossover" designs featuring multiple interventions administered in serial, randomization of intervention sequence between subjects is advised, with a sufficiently long "washout" period to combat carryover effects (such as improved performance due to longer exposure to the NI).Such experimental designs can also be used in animal studies.
For all study designs, subjects should be characterized in detail per Table II, including all inclusion/exclusion criteria, recruitment methods, and group allocation.Baseline outcome measures should be noted before the start of intervention, along with other relevant clinical and demographic characteristics.
2) Description of Intervention(s): All interventions, including procedures, NI devices, treatment programs, and surgical procedures, must be described in detail to ensure reproducibility.Stimulation and recording protocols, including the conditions under which the experiment was conducted, must be reported.If visual, auditory, tactile, or other sensory stimuli were used in either experimental or control conditions, these stimuli must be described per Table II.Whenever the experimental design involves behavioral assessments, potential behavioral biases and mitigation strategies (whenever applicable) should be reported (e.g., human handedness, education, expectations about the study).
3) Outcome Measures and Statistical Analysis: All outcome and performance assessment measures -both NI-derived and otherwise -must be precisely defined.The selection and relevance of all such measures to the study aims and hypotheses should be justified.Basic signal quality metrics for NI data (e.g, signal to noise ratio) are recommended, as are usability and user satisfaction scores.
All statistical analyses should be reported according to pertinent reporting guidelines and best practices (e.g., [4], [29], [30]).Reporting of data-processing and statistical methods must be sufficient to reproduce the presented results from raw data.The data set(s) between which each statistical comparison was conducted (e.g., between vs. within-subjects) must be clearly reported and justified.Where feasible, intended analyses of outcome measures should be documented and disclosed prior to data collection in order to maximize transparency and the statistical validity of the results obtained and minimize the opportunity for so-called 'p-hacking' [3].

C. Description of the Neural Interface
Insufficiently detailed reporting of NI device/system characteristics is the biggest barrier to the interpretability, replicability, and meta-analyzability of NI research -especially clinical studies.To overcome this barrier, researchers must provide a thorough description of the NI (per Table III), including specification of the applied stimuli and/or recording procedures.These parameters are critical to comparing NI performance across technologies, devices, and cohorts.
Fig. 2 shows a block diagram of a generic closed-loop NI system architecture which includes transducers (electrodes), signal acquisition and processing for neural recording, and stimulus generation and delivery for neuromodulation.The characteristics of all of these modules are essential for interpreting NI performance; essential reporting parameters for NI transducers and hardware are given in Table III, and essential reporting parameters for NI signal acquisition and processing is given in table IV.Diagrams such as Fig. 2 are essential for communicating the overall plan for a given NI approach and application, and we encourage their use for describing both the NI under test and the experimental context in which the NI is deployed.For custom experimental devices (including modified devices), authors should also provide a labelled diagram showing electrode / transducer sizes and locations. 2 The placement and positioning of the NI are critical to NI effectiveness (see [31]- [33]) and must be carefully reported (including the transducer, connectors, and any implanted electronics).Anatomical structures should be specified with reference to a widely-accepted formal vocabulary such as the Federative International Program for Anatomical Terminology 2 Electrode names like 'anode' and 'cathode' lead to confusion in the context of biphasic charge-balanced electrical stimulation to and should be avoided.Similarly, the labels 'active' or 'reference' imply assumptions about where activity or activation is occurring which may not be satisfied.For current-controlled stimuli, the term 'return' is clearer and should be preferred.For recording, 'reference' is to be preferred to other terms as this is the potential connected to a galvanically isolated recording device.As patients must never be connected to earth, by any means, terms such as 'earth', 'neutral', 'safety ground', or 'building ground' must not be used.We suggest using electrode / transducer labels like E1, E2, E3, etc.

TABLE III REPORTING TOPICS FOR NI PHYSICAL DEVICE PROPERTIES
1 ISO 10993 [19] loosely defines the terms "sub-acute" (> 24hr, <14d) and "sub-chronic" (14-28d) in the context of systemic toxicity evaluation. 2 Here, we use "minimally-invasive" to describe implanted NIs for which tissue or organ barriers such as the meninges or perineurium are not breached. 3See main text footnote 2 regarding the use of the term 'reference' vs 'ground', and also Supplementary Materials, Section VI.C. 4 Reporting of other details can be referenced to literature, provided those details have been measured in an equivalent (intraoperative) environment. 5The mechanical and electrochemical properties of NIs are critical to their long-term safety & efficacy and influenced by fabrication techniques.See [50], [57], [58].[34] or recognized anatomical atlases (e.g., [35], [36]).Implantation and device positioning procedures must be described, including the location of each component relative to anatomical landmarks, expected error margins, and any criterion for surgical re-positioning or exclusion.Describe any procedures carried-out to confirm device position during or after concluding the experiment (e.g., histology, CT imaging).Finally, for research concerning entire implanted NI systems (as opposed to investigations of NI components), expected and observed implant lifespans should be reported, as well as any observed or predicted failure modes (e.g., [37]).
From a clinician, end-user, or regulatory perspective, the algorithms used for signal-processing, stimulus generation and closed-loop control are as much a part of a NI as the underlying hardware.Reporting of these aspects of NI systems must be conducted to the same level of rigor as reporting of the physical interface; essential reporting parameters are given in Table IV.For neuro-sensing NIs, an unambiguous description of how signals from the electrodes / transducers are processed into recording channels is necessary.For novel NIs using recording approaches which might not be familiar to the wider NI community, the biophysical basis for the observed signals and measurement approach should be justified.Similarly, for novel neuromodulation NI approaches, the mechanism of the modulation of nerve activity should be described.
Algorithms used for signal conditioning, pre-processing, and analysis must be clearly reported and referenced.Providing public repositories containing implementations with representative

TABLE IV REPORTING TOPICS FOR NI SIGNAL PROCESSING PROPERTIES
1 For recording physiological signals, transient artifacts can mimic physiological signals when filtered through high-pass filters higher than first-order.To avoid this, characterizing any such filter by a single time constant value can ensure this good practice has been enforced. 2If recording signals reach saturation during regular use (e.g., due to stimulation artifacts), this should be noted along with the expected duration of invalid signal. 3Adequate filter specification is necessary to extract useful signals from noisy neurodata but is frequently underreported.Quite often, filters are used with non-linear group velocity transfer functions, and analysis methods are applied afterwards which assume linear group frequency transfer functions.We have also seen cases where non-causal filtering (e.g., MATLAB's filtfilt) is applied in such a way as to cause responses to precede stimulus, which is obviously nonsensical. 4In principle, the order of linear signal processing steps is not important.In practice, malfunction, artefacts, and other sources of confusion are more easily identified in the frame of an orderly description.See Supplemental Materials, Section VII.data sets is recommended.Inputs and outputs should be clearly specified, including confidence interval estimates (e.g., via bootstrap analysis of noisy input data, [38]).Existing standards for signal-processing research (e.g., [39]) should be applied.

D. Neural Interface Results and Discussion
NI research reports should clearly and succinctly present the results of all analyses described in the methods (including primary and secondary outcomes), plus any additional posthoc analyses (identified as such), in a manner that accurately summarizes and represents the full data set(s) analyzed, according to established biostatistical best practices [29], [30].
Graphical data representation (figures and tables) is preferred to text.Numerical values displayed in figures should be either incorporated in the figure, given in a corresponding table, or in the supplemental materials.Wherever applicable (including aggregated measures and descriptive statistics), measurement variability and uncertainty should be quantified with standard measures (standard deviations, confidence intervals, etc.).Likewise, all comparisons conducting using inferential statistics should report statistical significance (or non-significance) and effect size.Where parametric statistics are used, the normality of data distribution should be reported.Rationale should be provided for the exclusion from presented analysis of any data collected within the same protocol.Measures of NI signal quality (e.g., signal-to-noise ratio) or essential performance are strongly recommended, along with presentation of example raw data.
All unexpected or adverse events (e.g., device failures or explantations, subject withdrawal, unplanned animal deaths, etc.) should be reported.Observed technical issues and complications should also be reported, including all mechanical, electrical, or software failures (broken electrodes, connections, etc.).
Discussion of results should address the following topics: r To what extent do the results confirm the study hypothe- sis/es, and how do they fulfill the study objectives?
r The distinction between statistical and clinical/functional significance, with reference to the observed effect size, uncertainty, and minimal clinically important difference.r Key challenges to the future development and application of the presented technologies, including usability considerations and open questions for further investigation.

IV. DISCUSSION
As a preview of the proposed IEEE P2794 (RSNIR) draft Standard, this document has outlined minimum reporting requirements to ensure adequate transparency, reproducibility, and replicability of in vivo research involving implantable NIs, in line with recommendations from the National Academies of Sciences [64].In this way, RSNIR aims to complement existing scientific and clinical reporting guidelines by adding a layer of specificity to implantable NI technology.Most of these recommendations apply to all NI technology (including non-invasive modalities), and the RSNIR SWG is currently working to adapt these requirements to such technologies, including EEG-based BCIs.
In addition to high-level scientific reporting guidelines (EQUATOR etc.), RSNIR will be supported by a network of complementary NI-relevant Standards under current development, including IEEE P2731 (Unified Technology for Brain-Computer Interfaces) and P2792 (Therapeutic Electrical Stimulation Waveforms).For medical NI technologies, RSNIR also aims to facilitate compliance with foundational medical device standards such as ISO 14971 (risk management), ISO 13485 (quality management systems), and IEC 60601 (safety and essential performance requirements), as well as neurotechnologyspecific standards such as ISO 14708 and EN 45502 (active implantable medical devices for surgery).
The future impact of RSNIR in promoting high-quality neuroscience and neurotechnology development will depend critically on its widespread adoption by a variety of institutions that define incentives across academic, commercial, and clinical domains, including high-impact scientific publications, funding agencies, regulatory bodies (including clinical trial registries [40]), and/or medical payers.To promote such adoption, the draft Standard will seek to support an 'ecosystem of information interoperability' that serves the needs and objectives of all neurotechnology stakeholders, including aforementioned institutions as well as researchers, developers, clinicians, and end users.
To facilitate adoption at different levels of technological maturity (e.g., Technology Readiness Level [41]), the draft RSNIR will apply the principle of indirect reporting, whereby reporting requirements may be fulfilled via reference to previous publications or documents, provided that all required details are contained in the primary publication (including supplemental materials) and all others directly cited therein.
Regarding potential adoption by commercial entities, the draft RSNIR will seek to accommodate the proprietary nature of some NI system design details, by allowing the study replicability criterion to be fulfilled on a system-dependent basis, requiring the use of commercial hardware or software.In such cases, public assurance of the NI system's basic safety and performance may be achieved via third-party certification according to official testing Standards (UL, ASTM, CE-marking, etc.).To make the official RSNIR usable and useful at all stages of research & development (technological maturity), feedback to this article and participation in the RSNIR SWG are welcomed from all such stakeholders.

Fig. 2 .
Fig. 2. Block diagram of a prototypical NI system architecture.Sensors and effectors may interface invasively or non-invasively with the central or peripheral nervous system (CNS / PNS).Neural sensing components will almost always include hardware signal conditioning, digital-to-analog conversion, digital signal processing, and feature extraction.Neuromodulation components include waveform selection and generation and the output drive to the stimulus end effector.Sensors, from top to bottom: high-density intracortical (Utah) array, ECoG array, EEG.Effectors: deep brain stimulation, peripheral nerve array (FINE, [46]), and transcutaneous stimulation.

TABLE I
REPORTING TOPICS FOR NI STUDY AIMS AND CONTEXT

r
The fundamental novelty and/or significance of the find- ings with respect to the current state of the art, scientific body of knowledge, and/or field of potential applications.