Radiology & Imaging AI
D

Deep Breathe

Deep Breathe builds AI software that reads lung and trauma ultrasound. BlineSlide is its cloud reading aid for credentialed physicians reviewing adult lung ultrasound clips after acquisition: the B line module flags artifacts, the pleural line motion module flags absent lung sliding or lung pulse, and the software can mark a clip as unable to evaluate. FDA cleared the B line module in October 2025 and the pleural line motion module in October 2026, the latter tested on 818 clips with 94% sensitivity and 87% specificity against expert consensus for absent pleural line motion.

ATLAS, the company's AI Tactical Assessment System, runs on standard iOS and Android devices with no connection and works with ultrasound hardware from any maker. The trauma platform is sold for trauma diagnosis at the point of care, and ATLAS Pneumothorax for real time pneumothorax detection in hospital workflows, including post procedure checks that could cut chest X-rays, with a result about three seconds after a scan. ATLAS carries no clearance, and its diagnostic products are pursuing authorization. ATLASim is a training product, and the US Marine Corps has used ATLAS in a simulated trauma exercise.

Critical care physician Robert Arntfield founded the London, Ontario company, incorporated in 2020, and its team has 16 peer reviewed publications. In October 2026, FedDev Ontario announced more than C$1 million to help commercialize its ultrasound software.

AI Health Index verifiedNovember 20, 2026
Compare Deep Breathe with other vendors
Founded
2020
Headquarters
London, Ontario, Canada
Website
deepbreathe.ai/
Categories
radiology-and-imaging-ai
Indexed Products
BlineSlide, ATLAS Trauma, ATLAS Pneumothorax, ATLASim
Buyer Segments
Hospital emergency and critical care departments, Clinicians using lung ultrasound, Prehospital emergency medical services, Defense medical and education teams
Assessment

Capability Axes

An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read

AI Capability
AA on AI CentralityThe artificial intelligence is the product. Remove the model and there is nothing left to sell.
Vendor Published

BlineSlide interprets lung ultrasound clips for artifacts or movement along the pleural line. The B line detector marks whether it identified its target artifact. The pleural line motion module detects the absence of lung sliding or lung pulse and can return Cannot Evaluate when the clip falls outside its input criteria. The displayed result is tied to the source image the clinician is interpreting.

The company is also developing ATLAS for ultrasound diagnosis in trauma and postprocedure workflows. Its focus is software for image interpretation and guidance that works with ultrasound hardware obtained separately.

BB on Autonomy and Oversight ModelThe oversight structure is described and one part is missing, commonly the threshold at which the system stops or what happens after it is wrong.
Regulatory Filing

A Cannot Evaluate result comes back from BlineSlide when a clip falls outside its input requirements or when it cannot give a reliable reading. The 2026 indication limits use to concurrent reading of already acquired clips by a credentialed physician, calls for reading without the software when connectivity prevents a timely upload, and leaves treatment decisions with the physician. The earlier B line function lets the user review and edit the device output in the report workflow. ATLAS, which returns a diagnosis on the device in about three seconds, has no comparable rule for an indeterminate result, for who sees it or for how an override is recorded.

BB on Model and Technology TransparencyThe approach or the suppliers are named without the version and update discipline behind them. Naming a supplier is the entry to this band both here and on Model Supply Chain Disclosure, which ask different questions of the same disclosure: who receives the data, and what produces the output.
Vendor Published

The pleural line motion model uses neural networks trained on ultrasound to recognize subtle movement, including movement that is absent, and 172 million ultrasound frames went into training across the platform. The BlineSlide clearance filing sets out clip eligibility and compares the model with expert consensus on test data from sites geographically separate from development, and it keeps motion analysis apart from the earlier artifact module. The architecture is named only as neural networks, and the production model revision, the breakdown of training subjects by site and the way customers learn of a model change are not given.

BB on Model Supply Chain DisclosureSubstantial partial disclosure, or a chain that is structurally short: an in house build, a cleared model that cannot be quietly swapped, or a deployment where the transfer does not occur at all. Naming the model provider exits the band below into this one; the axis rises from here on the completeness of the party list and on the terms that govern data once it arrives.
Vendor Published

Deep Breathe trained the neural networks behind its lung sliding detection itself, and the BlineSlide clearances cover separate artifact and pleural motion functions from the same manufacturer. ATLAS runs locally on the mobile device, which removes the transfer for an on device reading. The cloud version receives previously acquired clips, and its filing names no cloud operator and no party handling hosting, crash telemetry or storage. No processor list is offered, and no terms say how a processor may keep or use the clips it receives.

BB on Clinical and Operational EvidenceNamed deployments with dated outcome figures and enough method to test them, or published research short of independent validation.
Regulatory Filing

The cleared pleural line motion model was assessed on 818 clips from sites geographically separate from its development data, reaching 94% sensitivity and 87% specificity against expert consensus with Cannot Evaluate results scored as errors. A retrospective comparison had 25 physicians read 400 clips with and without the device, and the earlier B line clearance reports 91% sensitivity and 84% specificity on 1,005 clips.

The 2024 Chest study evaluated a development system prospectively at the bedside. The homepage claims ATLAS can detect 70% of preventable trauma deaths, with no study attached, and the clearance evidence covers the uploaded clip workflow rather than the on device ATLAS products.

CC on AI Safety and PHI StewardshipGeneral assurances of privacy and security that do not answer the questions artificial intelligence raises: what is retained, what reaches a model, and what happens to it there.
Vendor Published

Clips uploaded to BlineSlide can be declined for evaluation, including some that otherwise look technically adequate, and the 2026 filing covers risk management, validation, security testing and limits on relying on an output during patient care. ATLAS works offline without sending a clip to a server.

The privacy policy covers visitors to the company website, and nothing sets how long the cloud service keeps images and model outputs, whether submitted clips can train models, or how an error by the deployed model is investigated.

Regulatory and Compliance
DD on HIPAA and BAA PostureNo statement of status and no privacy document that reaches the product. Both halves are absent at this grade: a service level privacy document on its own lifts a vendor to the band above even where no status is stated. A website privacy notice, including one that says it does not cover customer data, does not reach the product, so a vendor whose only document is that notice and who states no status grades here.
Vendor Published

Previously acquired lung ultrasound clips reach BlineSlide through a cloud service, while ATLAS can run the model on the device. The website privacy policy covers information given while using the site, refers to Canadian privacy law for that scope, and sets no terms for identifiable clinical images in the cloud application.

No US business associate position, agreement or explanation of how an equivalent regime applies to the clinical service is offered, including for any data exported from the device. ATLAS is sold as free from cloud dependence, which keeps images local for the on device product but leaves the cloud reading aid without a stated privacy position.

CC on Security Certifications and Trust CenterControls are described with an outside check behind them, such as independent penetration testing on a stated cadence, but no attestation against a recognized framework. A certification that tests the software security features rather than the vendor operations, such as federal health IT certification, is an outside check of this kind.
Regulatory Filing

Penetration testing, vulnerability scanning and a review of known vulnerabilities were run on BlineSlide for its FDA submissions, alongside a device risk analysis and software verification and validation. That testing covers the cleared software rather than the cloud hosting or mobile distribution operations, and the device summaries name neither the testing firm nor how often the work repeats after release. No SOC 2 report, ISO 27001 certificate or other audit of Deep Breathe's own operations is claimed for any of its clinical products.

DD on FDA and Regulatory StatusThe positioning is unclear or unsupported: clinical claims with no regulatory position stated, or language implying a status the databases do not carry. An open enforcement action that names the product also grades here, whatever the positioning: an FDA warning letter with no close out letter posted, or a consent decree in force. The note says which of the two applies, and the grade returns to the positioning bands when FDA closes the action.
Regulatory Filing

BlineSlide received FDA clearance in October 2025 for B line artifact assistance under K252007 and in October 2026 for pleural line motion assistance under K261679, which covers a physician's review of uploaded clips after acquisition. The clearance announcement describes the AI working during the scan, although the cleared device takes no data during the bedside examination and does not interface with an ultrasound system.

The homepage places its clearance line beside ATLAS Pneumothorax without saying which product is cleared, and that product's page sells real time diagnostic detection of pneumothorax on the device with no regulatory status stated. The ATLAS products are pursuing authorization, and their marketing runs past both clearances.

AA on AI Governance and Bias DisclosureA bias or fairness evaluation with a stated method, subgroup performance, or an independent audit of model behavior. Subgroup performance means published results compared across the groups the product serves, clinical and physiological as well as demographic, so a buyer can see where it underperforms.
Regulatory Filing

The BlineSlide motion module study reports sensitivity and specificity by scanner manufacturer and transducer type, as well as by clip characteristics. The aided reader study also breaks out performance by patient characteristics and reader specialty: aided specificity was 0.71 against 0.78 without assistance for patients with a BMI over 30.

The summary gives sample sizes and withholds estimates where a group holds too few clips, and missing age, sex and BMI data limit what the comparison can show. No such breakdown exists for the on device ATLAS trauma and pneumothorax tools.

BB on AI Liability and RecourseA published falsifiable commitment, or a real correction route the affected person can exercise against the vendor. A published error rate with its method and denominator grades here. A statutory right that runs to the covered entity rather than to the vendor does not reach this band on its own: every other route here asks something of the vendor, and being located in a particular jurisdiction is not conduct.
Regulatory Filing

The motion detection filing gives sensitivity and specificity against expert consensus for 818 clips and defines how an uninterpretable clip counts in that comparison, so a hospital can test the stated error rate against its own reads. A clinician sees the source clip beside the result and decides whether to use it, and a Cannot Evaluate response sends the reader back to unassisted interpretation.

The website terms disclaim warranties for the site, and no contract term gives a customer a correction or reimbursement route for a clinical reading or an accuracy commitment against a missed finding.

Integration and Deployment
DD on EHR and Interoperability DepthNo integration evidence. A connector described as available on request grades here until one exists.
Regulatory Filing

Clinicians upload ultrasound clips to BlineSlide after acquiring them; the 2026 clearance says the application does not connect directly to ultrasound systems or take a bedside feed. The earlier module takes MP4 files and lets users review the output in a report workflow.

ATLAS works with footage from any ultrasound hardware and runs its analysis on a mobile device, and the pneumothorax page says it integrates with existing ultrasound systems and workflows without naming one, a claim about image sources rather than records. No connection to a chart, a picture archive or a clinical worklist is offered for either product.

BB on Deployment Model and Data ResidencyOptions and residency are stated with isolation or the processing path left open.
Vendor Published

ATLAS runs on standard iOS and Android devices and analyzes a clip with no internet connection, so its inference stays on the user's device. BlineSlide is a separate cloud service that takes clips after acquisition and needs reliable connectivity to return the result to the reviewing physician. Where the BlineSlide cloud stores and processes clips is not stated, and neither is whether images stay on the device after ATLAS finishes or what may leave it for support. Tenant separation, backups and cross border handling for the cloud service are not described.

Commercial
CC on Commercial TransparencyNo price is published and the posture is discoverable: a buyer can establish how the product is sold and what drives the cost before contacting the vendor. Most of the index sits here.
Vendor Published

Organizations arrange evaluations and pilot deployments directly with Deep Breathe, with training and phased implementation agreed in the same conversation, and ATLASim is sold for education and training. The clinical software runs on ultrasound hardware bought separately, and BlineSlide has its own upload workflow. The pneumothorax page models about a 50% cut in hospital cost from ultrasound based rule out, without the model behind the figure.

No starting license, pilot fee or price per clip is shown, and whether a hospital buys access by site, physician or volume, or pays for a mobile rollout and its training together or apart, is not stated.

BB on Setting and Specialty CoverageCoverage is named with validation behind part of it.
Regulatory Filing

BlineSlide's cleared motion detection is for credentialed physicians reading adult lung ultrasound clips, with emergency departments, intensive care units and outpatient facilities named in its filing and validation readers drawn from emergency medicine, internal medicine and critical care. ATLAS Pneumothorax is designed for hospital use, including post procedure checks.

The trauma platform is sold for hospital and austere environments, the Marine Corps has used it in a simulated exercise, and testimonials come from a Canadian thoracic surgery chair, a Canadian air ambulance service and a US Army sergeant major. The retrospective reading results cover the cloud clips, with no field deployment results and none across a wider range of trauma findings.

Alternatives

Alternatives to Deep Breathe

Other products that do the same kind of work as Deep Breathe, and what each one does.

Sonio makes prenatal ultrasound software for sonographers, maternal fetal medicine specialists and the obstetricians who read the scans, built as one platform with AI reading aids and a reporting and image management layer.

UK company whose red dot platform sits at the autonomous end of imaging AI: it reports chest X-rays it classifies as High Confidence Normal autonomously, removing them from the radiologist worklist (about 15 percent of studies, according to the company), triages suspected lung cancer for…

French imaging AI company based in Paris whose Milvue Suite reads plain radiographs.

Paris based imaging AI company whose Rayvolve suite reads plain radiographs: AZtrauma detects fractures, dislocations and joint effusions, AZchest covers chest findings including pneumothorax, pleural effusion and lung nodules, and AZmeasure and AZboneage add measurements and bone age.

Qure.ai, founded in 2016 in Mumbai with a US entity in New York, sells radiology AI in more than 105 countries through more than 5,500 sites by its own count.

Ohio based company whose ClearRead software applies AI to chest CT and chest radiographs.

Commercial

Pricing

Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.

Entry Price Pricing Basis BAA Tier Implementation Source
Quote based
Organizations arrange a demonstration, evaluation or pilot directly with Deep Breathe. The company offers training and phased implementation for the ultrasound software. No business associate agreement is offered for the cloud service. Evaluations, training and phased implementation are offered, with no fee shown. Vendor Published

No cost per site, clinician or clip is shown. ATLASim is sold for education, and no price appears for it or for clinical deployments.