Nanowear

A regulator cleared the algorithms on their own, which settles this axis more definitively than any argument could.

The 2021 clearance was software only, covering the platform as an end to end digital system and enabling standalone artificial intelligence and deep learning algorithms to operate as a medical device in their own right rather than as a feature of a cleared instrument. That is the strongest possible evidence that the models are the regulated article, and very few vendors in this index hold a clearance of that shape.

The underlying necessity is physical. Cloth nanosensors produce raw multi modal signal across electrophysiological, hemodynamic, acoustic, metabolic and activity channels at a stated volume exceeding 100 million data points per patient per day. No clinician reads that; the diagnostics exist only because models reduce it.

The blood pressure product makes the same point in a different way, since a cuffless continuous measurement is not a sensor reading at all but an inference from other signals, and the company names the model family behind it.

The company's own framing is consistent rather than promotional, describing itself as building deep learning based diagnostics and its platform as getting smarter with each patient.

Graded A without qualification. The garment is an instrument for feeding models, and the models carry their own regulatory identity.

The oversight boundary is published in regulatory language, which almost nothing else in this index does.

The company publishes its indication for use on a customer facing page, and that indication states the platform is intended for use under the direction of a licensed medical professional, to record, display and store specified physiological data. Every element of that sentence does work. A named clinician directs use, so the system does not operate autonomously. The verbs are record, display and store, which describe an instrument rather than an interpreter. And the data types are enumerated rather than described in general terms.

Publishing an indication for use is itself the creditable act. Across this session, vendor after vendor has been graded down for leaving the scope of its product to be inferred from marketing, and clearance summaries are public records that companies rarely surface themselves. This one puts the binding language where a buyer will read it.

The blood pressure product adds a second bounded statement, since its validated performance is expressed as tracking changes beyond a stated threshold rather than as absolute measurement accuracy, which is an honest description of what a cuffless method can support.

What is missing is operational calibration. No alert protocol, no escalation path, no notification timing, no false alarm rate and no sensitivity figure for any diagnostic function was located, and a home based cardiopulmonary platform ultimately depends on someone acting on what it records.

Ask who monitors output in home deployments, what triggers escalation, and the detection performance of the cardiopulmonary algorithms.

Both the sensing physics and one model architecture are described specifically, which is an unusual combination.

The sensing side is characterised in detail: cloth based nanosensors woven into a gender neutral, size adjustable undergarment, capturing electrophysiological, hemodynamic, acoustic, metabolic and activity signals simultaneously and time synchronously at millisecond resolution, producing a stated volume exceeding 100 million data points per patient per day across more than 85 biomarker points.

The model side carries a disclosure that is genuinely rare. The blood pressure product is described as a four decision tree algorithmic model, which names an architecture family rather than referring to proprietary artificial intelligence, and the validated performance boundary is published as tracking changes beyond roughly 15 millimetres of mercury systolic and 10 diastolic. A stated boundary is more useful than an accuracy percentage here, because it tells a clinician what magnitude of change the method can be trusted to follow.

Intellectual property is enumerated by domain across 13 awarded and 12 pending patents covering nanotechnology, sensor manufacturing, the wearable, the platform, the ingestion pipeline and the algorithms.

What is missing is performance for everything except blood pressure. No sensitivity, specificity or operating point exists for the cardiopulmonary diagnostics, no model card, and no description of the deep learning architecture behind them.

The learning claim is also unexplained, since a platform that gets smarter with each patient implies a training loop that is nowhere described.

Ask for performance on the cardiopulmonary algorithms, and how the learning loop operates under change control.

The distinctive components are made in house and the partner layer is named, with no register behind either.

What is established is unusually favourable in one respect. The sensing technology is the company's own, protected across 13 awarded and 12 pending patents that explicitly cover nanotechnology, scaled nanosensor manufacturing, the multi parametric wearable, the software platform, the ingestion pipeline and the algorithms. Enumerating patent coverage by domain tells a reader which parts of the stack the company controls, and for a materials based sensing product that manufacturing claim is the load bearing one, since cloth nanosensors are not a component that can be bought.

The partner layer is also named openly, covering a continuous glucose monitoring manufacturer under a licensing and data arrangement, a blood collection patch developer, and a surgical device manufacturer as a distribution partner. Naming the parties whose data crosses into the platform is a genuine disclosure.

The hosting provider is named on the security page.

What is absent is enumeration and provenance. No sub processor register exists, no machine learning framework or component is identified, no contract manufacturer is named for garment production at scale, and no statement addresses training data.

That last question is sharpened by the company's own learning claim. A platform described as getting smarter with each patient implies deployed data feeding development, and no consent basis, governance or segregation between clinical and research data is described.

Ask for the sub processor register, the garment manufacturing arrangement, and the provenance and consent basis of training data.

One well described validation programme, research partnerships with real institutions, and no published commercial outcome.

The blood pressure validation is described with more methodological detail than most vendors offer. Three study arms, training and validation performed in independent geographies, demographic populations characterised against national census composition, and a stated performance boundary of tracking changes beyond roughly 15 millimetres of mercury systolic and 10 diastolic across all classes of hypertension. Separating training and validation geographically is a genuine guard against overfitting to one site's population and equipment.

Institutional relationships are named and substantial, including a remote diagnostic research alliance with the largest hospital system in New Jersey and a Brooklyn medical centre, alongside a commercial distribution partnership with a surgical device manufacturer for post surgical recovery. Recognition from a major medical device manufacturer's wearables challenge and from an industry analyst firm adds outside assessment.

What is missing is publication and deployment. No peer reviewed paper was located for any of the four cleared products, no named health system runs this at scale, and no patient or study volume is stated.

One marketing figure needs the usual caution. A claimed 60 percent reduction in provider workflow and patient time carries no method, comparator or setting, and appears in an executive quotation rather than in any analysis.

Ask for peer reviewed validation of the cardiopulmonary algorithms, current deployment scale, and the basis of the workflow reduction figure.

A named host and a published privacy notice, with one claim that raises a question the company does not answer.

What is disclosed is modest and real. A dedicated security page names Amazon Web Services as the hosting environment, states adherence to industry best practices for developing and maintaining the software and systems, and links a privacy notice covering protected health information under federal and state law. Naming the cloud provider is something a majority of vendors in this index decline to do.

What is missing is the operational layer: no encryption description, no retention schedule, no access control model, no audit logging and no deletion process.

The unanswered question comes from the company's own words. Executive material describes the platform as getting smarter with each patient, which is a claim about learning from deployed patient data. If models improve from clinical use, then patient physiological recordings feed development, and nothing states the consent basis, whether customers can decline, or whether individuals are informed. It also sits awkwardly against device regulation, since a cleared algorithm that changes in the field is a different regulatory object from one that is locked, and no change control position is published.

The data volume raises the stakes. More than 100 million data points per patient per day, captured continuously from a garment worn against the body, is among the densest physiological records any vendor here holds.

Ask whether models update from deployed patient data, under what consent and change control, and what is retained centrally.

A published commitment with a linked privacy notice, and no contractual position behind it.

What exists is better than an unattributed badge. A dedicated security page states a commitment to maintaining the privacy of protected health information as directed by applicable federal and state law, and links a full privacy notice rather than referring to one. Naming state law alongside federal law is a small but meaningful precision, since state health privacy regimes frequently impose obligations beyond the federal baseline and most vendors mention only the latter.

What is absent is the agreement layer. No business associate agreement template, execution requirement, negotiation stance or subcontractor flow down position was located.

The dual setting indication creates a question that nothing addresses. The platform is cleared for use both at home and in a healthcare facility, and those produce different contracting pictures: a facility deployment sits squarely inside the covered entity relationship, while a home deployment may run through a provider, a research programme or a distribution partner, each with a different chain. Nothing states which arrangements the company enters.

The partnership model adds another. Ingesting continuous glucose data from a third party device under a licensing and data partnership means patient data crosses a corporate boundary, and no material describes the terms.

Ask for the agreement template, which entity contracts in home deployments, and what governs data shared under device partnerships.

A standing security page exists, which is more than most of this session offered, and no external attestation sits behind it.

The page is a real customer facing resource rather than a footer badge. It names the hosting provider, states a commitment to protected health information privacy under federal and state law, links a full privacy notice, and publishes the platform's indication for use alongside. A prospective buyer can begin diligence without contacting sales, which is the behaviour this axis exists to reward.

What the page does not contain is verification. Adherence to industry best practices is asserted without naming which practices, and no information security management certification, service organisation controls report, penetration testing statement or vulnerability disclosure policy was located.

The surface merits more. This is a connected medical device worn against the body in patients' homes, transmitting a dense continuous physiological record, and now ingesting data from a third party continuous glucose system across a corporate boundary. Device authentication, firmware practice, garment decommissioning between patients and the security of the partner data channel are all unaddressed.

The cleared software only status raises the bar rather than lowering it, since a regulated algorithm operating as a device carries cybersecurity expectations in its own right under current device guidance, and nothing describes how those are met.

One pre emptive note: restating the privacy commitment cannot move this grade. Only an external attestation, or documentation available under agreement, will.

Ask whether an external security assessment exists, how the device authenticates and updates, and how the partner data channel is secured.

Four clearances spanning eight years, including one that most vendors in this index never attempt.

The sequence runs from an initial wearable clearance in 2016, through clearance of the diagnostic undergarment with its machine learning platform in 2020, to a software only clearance in 2021, and a cuffless continuous blood pressure diagnostic in January 2024.

The 2021 software only clearance is the significant one. It authorised the digital platform end to end and permitted standalone artificial intelligence and deep learning algorithms to function as software as a medical device, which means the models were assessed on their own terms rather than riding on a cleared sensor. That is a harder submission and a materially different regulatory asset, and this index has repeatedly recorded vendors whose algorithms sit in an undeclared space beside cleared hardware.

The 2024 clearance is notable on its own merits as a cuffless continuous blood pressure diagnostic, an area where regulators have been cautious, and the accompanying validation description is specific about study design and population.

The company also publishes its indication for use, which is uncommon and which this record credits.

One discrepancy belongs here rather than buried. The published indication is narrower than the marketed capability, covering two leads of electrocardiogram, respiration by thoracic impedance, heart sounds and activity, against marketing describing 85 or more biomarkers and replacement of four separate instruments. Both are the company's own statements.

Ask for the clearance numbers and the indication for each, and how the marketed biomarker count maps to cleared claims.

A representativeness claim tied to an external benchmark, which is a stronger form of bias disclosure than most vendors attempt.

The blood pressure model's validation is described as performed across three study arms with training and validation conducted in independent geographies, and in demographic populations characterised as mirroring the most recent national census with generalisability to the entire hypertensive population. Two things in that are substantive. Anchoring population composition to a census benchmark is an external, checkable standard rather than a self assessment that the sample was diverse. And separating training and validation geographically guards against a model that has learned one region's population, equipment and clinical practice.

That matters particularly for blood pressure, where measurement disparities across ethnic groups are well documented and where hypertension prevalence and presentation differ substantially by ancestry.

What holds this short of the top is that the claim carries no numbers and covers one model. No per subgroup performance is published, so representative enrolment is demonstrated while equivalent performance across those subgroups is not, and the other three cleared products carry no comparable statement.

One exposure specific to this form factor is unaddressed. Sensing depends on cloth contact against skin, so signal quality plausibly varies with body habitus, breast tissue, chest hair, skin condition and garment fit, and a size adjustable design mitigates without eliminating that.

No fairness framework, drift monitoring or external audit was located.

Ask for performance by ethnicity, sex and body habitus, and whether the census anchored approach was applied to the cardiopulmonary models.

A precisely bounded claim and no commitment behind it.

A dedicated pass located no service level agreement, no accuracy warranty, no uptime commitment, no indemnity and no remediation position.

The published indication for use does real work here and should be recognised without being mistaken for recourse. Stating that the platform records, displays and stores specified data under a licensed professional's direction limits what the company is claiming and places interpretation with the clinician, which is an honest allocation. It is a limitation of scope rather than an undertaking, and it tells a buyer nothing about what happens when the system fails within that scope.

The failure modes are consequential. A garment that stops recording mid session, a transmission that does not arrive, or an algorithm that misreads a cardiopulmonary signal in a patient being managed at home all produce clinical exposure, and no detection rate, availability figure or delivery commitment exists for any of them.

The blood pressure product is the one place a boundary is published, expressed as tracking changes beyond a stated magnitude, and even there nothing states the consequence of a missed change.

One allocation question is specific to this vendor. Where diagnostics combine the company's own biomarkers with glucose data ingested from a partner device, an erroneous combined assessment involves two manufacturers, and nothing describes how responsibility divides.

One pre emptive note: further clearances cannot move this grade. Only contractual terms, or published detection and availability performance, will.

Ask what is warranted on recording continuity and diagnostic accuracy, and how liability divides across partner data.

Device level interoperability is real and named; clinical system interoperability is absent.

The device side has concrete evidence. A licensing and data partnership ingests continuous glucose readings from a named third party monitoring system, time synchronously alongside the company's own cardiovascular biomarkers, which is a genuine integration rather than a co marketing arrangement and produces something neither party offers alone. A further initiative integrates third party blood collection patches. The ingestion pipeline is patented, indicating the intake layer is engineered rather than improvised, and interoperability was named as a guiding principle of the 2021 platform clearance.

Combining continuous glucose with continuous blood pressure, electrocardiography, respiration and hemodynamics in one time synchronous record is a meaningful capability for cardiometabolic assessment, and it is the strongest interoperability fact here.

What is absent is the clinical record. No electronic health record is named, no interface standard such as HL7 or FHIR is described, no marketplace listing was located, and nothing states whether recordings, diagnostics or events reach a patient chart.

That gap matters given the cleared indication, which describes recording, displaying and storing physiological data under a clinician's direction. A clinician directing use needs the output where they work, and nothing describes how it gets there.

Ask which record systems are integrated in production, through what standard, and how diagnostic output reaches the chart.

The cloud provider is named and nothing else about the deployment is described.

Naming Amazon Web Services on a dedicated security page is a real disclosure and this record credits it, since a substantial majority of vendors graded this session declined to identify their hosting provider at all. The architecture is otherwise described only in outline: a wearable garment, a mobile layer and a cloud environment receiving and analysing the data.

What is absent is everything specific. No region is stated, no residency commitment is made, no tenancy or segregation model is described for a platform serving both clinical and research users, and no backup, continuity or recovery position was located.

The dual setting indication makes the omission more consequential than usual. The platform is cleared for use both at home and inside a healthcare facility, and those involve different networks, different security expectations and potentially different data paths, with nothing describing whether deployment differs between them or whether a facility can keep data on premises.

Continuity deserves attention for the home case in particular, since a patient wearing a diagnostic garment at home depends on connectivity the vendor does not control, and nothing states whether the device buffers locally, how long it can operate offline, or what happens to a recording interrupted mid session.

The research context adds a further question about whether study data is segregated from clinical data.

Ask which regions host data, whether facility deployments can remain on premises, and how the garment behaves during connectivity loss.

Cost is absent from every published surface. A dedicated pass located no pricing page, no unit of charge, no range, no tiering, no implementation fee position and no minimum commitment.

The unit question is genuinely open and unusually consequential because the product is a garment. Clothing worn against skin is consumable in a way a rigid device is not, so whether the undergarment is purchased, leased, replaced per patient, laundered and reused, or supplied per episode determines the economics entirely, and nothing indicates which. A hospital at home programme cycling patients through garments faces a very different cost structure from a cardiology practice fitting a patient for a diagnostic period.

Sizing compounds it. A size adjustable garment still implies inventory across a range, and stocking, fitting and reprocessing are real operational costs that no published material addresses.

The reimbursement position is also unstated, which matters because the cleared indications sit in territory where remote monitoring and diagnostic codes plausibly apply, and other vendors in this segment publish their billing route as the primary commercial disclosure. This one does not.

No return proxy is published either. The replacement argument, that one garment substitutes for several separate instruments and their associated appointments, is directly calculable and no figure supports it.

Ask whether the garment is sold, leased or consumable, the cost per patient episode, sizing inventory requirements, and which reimbursement codes apply.

Cleared scope is narrow and precisely stated; marketed scope is much wider and largely prospective.

What is cleared and published covers use at home or in a healthcare facility under the direction of a licensed medical professional, which is a genuinely dual setting indication and includes clinical research as a stated context. Cardiology and cardiopulmonary assessment are the established domain, with post surgical recovery addressed through a distribution partnership.

The wider range is pipeline rather than product. Hypertension, chronic obstructive pulmonary disease, sleep apnoea, worsening heart failure and post surgical recovery have been described as target verticals since 2021, and of those only hypertension has since reached clearance. A reader encountering the marketed breadth should understand that most of it describes intent.

Adjacent markets are named that sit outside healthcare entirely, including sports medicine and performance, industrial safety and government, which broadens the addressable market and says nothing about clinical depth.

What is not evidenced anywhere is installed base. No health system deployment, patient volume or programme count was located, and the named institutional relationships are research alliances rather than operational deployments.

One physical constraint deserves noting and is unaddressed. A sensing undergarment must fit, and fit varies with body habitus, mastectomy, ostomy, wound sites and mobility, none of which is discussed.

Ask which conditions are cleared today, current deployment scale, and how fit is handled across body types and post surgical patients.