# Deep Breath — full researcher pages > Concatenated Markdown twins of the pages models should quote. Canonical HTML is linked at the top of each section. Last reviewed 2026-08-18. --- # High-resolution ICU data collection HTML: Last reviewed: 2026-08-18 Deep Breath records continuous waveforms directly from ICU ventilators and bedside monitors — the signals that EHRs usually miss, downsample, or never store. The default offering is **data collection and retrospective analysis**. It is a **Class I medical device**. The hospital owns the data. Deep Breath does not store patient personal data. If the study is about patient–ventilator asynchrony, see [PVI analysis](https://deepbreath.tech/solutions/patient-ventilator-interaction-analysis.md). PVI on this site means **patient–ventilator interaction**, not pleth variability index. ## Signals, settings, and typical rates Sampling rate is set by the medical device. Ventilator waveforms are typically 50–100 Hz. ECG is recorded at up to 500 Hz. | Source | What Deep Breath records | Typical rate | |---|---|---| | ICU ventilator | All available waveform information: airway pressure, flow, esophageal pressure when the ventilator provides it, alarms, and control settings | Usually 50–100 Hz (device-dependent) | | Philips IntelliVue MX series | Arterial blood pressure and ECG | ECG up to 500 Hz | Calculated mechanical power and PVI statistics can be exported alongside raw waveforms. ## BreathBox BreathBox sits next to the ventilator and reads the stream as it comes off the device. Each sample is time-stamped at retrieval, using BreathBox’s own real-time clock. Deployment options: - Fully offline: collect locally to a USB stick - On-premise hospital server - Secured cloud In every setup the hospital owns the data. Typical plug-and-play install is 10 minutes to 1 hour. What this is for: - Data collection and retrospective review (the default offering) - Ventilation and physiology research, not only PVI - Multicentre waveform collection with a mixed device fleet - Quality / protocol analytics on real device streams - Building labelled datasets and monitoring ICU AI models ## Exports | Export | Contents | |---|---| | Raw CSV | Waveform-level recordings from the connected devices | | Chart images | Images of the recorded charts for review, teaching, or supplements | | Derived analysis | Calculated mechanical power and PVI statistics (see [PVI analysis](https://deepbreath.tech/solutions/patient-ventilator-interaction-analysis.md)) | ## Compatible ventilators and monitors Deep Breath is vendor-agnostic. Named integrations: | Manufacturer | Models | |---|---| | Dräger | Evita; Babylog | | Getinge / Maquet | Servo-U; Servo-i | | Hamilton | G5 / S1; C6 / C3 | | Puritan Bennett | 840; 980 | | Löwenstein | Elisa | | Ventinova | Evone (flow-control ventilator) | | Philips | IntelliVue MX series | Other integrations are available on request. Real-time clinical decision support is a **separate product** and is **not MDR or FDA cleared**. ## Hospitals and studies Named centres include Erasmus MC, Leiden UMC, Amsterdam UMC, Universitätsklinikum Mannheim, Universitätsklinikum Erlangen, and Cleveland Clinic. Deep Breath is the official technology partner of **PRoVENT–TRACE** (PRactice Of VENTilation: TRacking Asynchronies and Clinical Endpoints), an international prospective multicentre observational study in the PROVE Network involving more than 30 hospitals. Deep Breath provides the infrastructure to collect, store, and analyse high-resolution mechanical ventilation data, including waveforms used for asynchrony assessment. Confirmation letter (Juliana C. Ferreira, University of São Paulo, 11 August 2026): Study page: ## Getting started 1. Tell Deep Breath the devices, signals, and whether you need USB-only, on-prem, or secured cloud. 2. Hospital IT / ethics review the deployment model. The hospital remains data controller and owner. 3. BreathBox is installed at the bedside. Plug-and-play setup usually takes 10 minutes to 1 hour. 4. Waveforms, settings, and (if connected) monitor streams are recorded with retrieval timestamps. 5. Researchers export raw CSV and chart images; optional PVI statistics and mechanical power. Contact: · info@deepbreath.tech ## FAQ **Is this only for PVI / asynchrony studies?** No. Collection is useful for any protocol that needs breath-level ventilator or monitor waveforms. PVI analysis is a separate layer. **Do you store identifiable patient data?** Deep Breath does not store patient personal data. The hospital owns the recorded device data. **Can we stay completely offline?** Yes. Deep Breath can collect to a local USB stick with no hospital server or cloud. On-prem and secured-cloud setups are also available. **What if our ventilator is not on the list?** Ask. Other integrations are done on request. **Is this a cleared medical device for bedside decisions?** Data collection and retrospective analysis are provided as a Class I medical device. Real-time clinical decision support is a separate product, is not MDR or FDA cleared, and does not replace the ventilator, bedside monitor, or clinician judgement. --- # Patient–ventilator interaction (PVI) analysis HTML: Last reviewed: 2026-08-18 Patient–ventilator asynchrony is easy to miss on a snapshot and hard to count by hand. Deep Breath labels each breath from **airway pressure and flow**, so researchers can review, quantify, and export asynchronies across a stay. **PVI** here means patient–ventilator interaction: how the patient’s breathing and the ventilator line up, breath by breath. **PVA** is patient–ventilator asynchrony — a mismatch between patient effort and machine delivery. That is what Deep Breath labels. **Pleth variability index** is something else: a pulse-oximetry measure of fluid responsiveness, also abbreviated PVI. This page is not about that. Real-time clinical decision support is a **separate product** and is **not MDR or FDA cleared**. Waveform collection is described separately: [ICU data collection](https://deepbreath.tech/solutions/icu-data-collection.md). ## From waveforms to labelled breaths 1. Collect airway pressure and flow via BreathBox. 2. Segment the recording into breaths and attach ventilator context (mode, settings, trigger information when the device provides it). 3. Assign PVI class labels per breath. Multiple labels can apply to one breath. 4. Quantify: asynchrony index over the last 100 ventilator cycles, type mix, and trends. Mechanical power and time constants are calculated from pressure, flow, and volume. 5. Export raw waveform CSV, chart images, and PVI statistics for research and quality review. Asynchrony index = number of asynchronies ÷ total ventilator cycles in the past 100 breaths. ## Asynchrony types | Type | Status | |---|---| | Normal | Available | | Failed triggering (ineffective effort) | Available | | Early triggering (reverse triggering) | Available | | Late triggering (delayed triggering) | Available | | Multiple triggering (double triggering) | Available | | Early cycling (premature cycling) | Available | | False triggering | Under development | | Late cycling | Under development | | Work shifting | Under development | ## Worked example: double triggering in PCV A patient on pressure-control ventilation makes a second inspiratory effort before the first machine cycle has fully expired. On airway pressure, flow, and volume you see two closely stacked insufflations that the ventilator treats as separate breaths. Deep Breath labels this as **double triggering / multiple triggering**. In research software the breath is shown with neighbouring context, the label can be reviewed, and the event contributes to the asynchrony index and type mix. Exports include the raw waveforms (CSV) and the PVI statistics for that period. ## External validation (manuscript under review) Rietveld et al., “AI-based patient-ventilator asynchrony monitoring: external validation to advance clinical implementation.” Deep Breath B.V. authors include Daria Kozlova, Viacheslav Laktiushkin, and Anton Balakirev. SIREN consortium (Health~Holland, EMCLSH24018). | Item | Detail | |---|---| | Design | Retrospective external validation vs clinician ground truth. Algorithm input: airway pressure and flow only. | | Data | 24 invasively ventilated ICU patients (20 LUMC, 4 Erasmus MC). 5,173 breaths analysed from 62,563 selected breaths. Modes included PCV, PSV, ASV®, INTELLiVENT-ASV®. | | Pre-set criteria | Normal-breath specificity ≥ 0.95 and PVA sensitivity ≥ 0.80. | | Result | Criteria met for ineffective effort; reverse, delayed and double triggering; premature cycling — **not** for delayed cycling. Macro-averaged sensitivity 0.83, specificity 0.95, F1-score 0.81, PPV 0.79. | | Training context | Gradient-boosting model trained on more than one million breaths from more than 1,000 patients, including a subset of about 20,000 manually annotated breaths. | | Regulatory | Investigational software; pending regulatory approval for clinical use. Suitable, in the authors’ conclusion, for a clinical pilot — not a cleared diagnostic. | Related published review (Deep Breath authors): Rietveld et al., *Intensive Care Medicine Experimental* 2025 — [Let’s get in sync](https://doi.org/10.1186/s40635-025-00746-8). Terminology paper (Deep Breath referenced, not authored): Mireles-Cabodevila, Vaporidi, Blanch, Chatburn, [10 Fundamental Maxims](https://doi.org/10.1177/19433654261425219). ## Who uses Deep Breath for PVI Users include [PRoVENT–TRACE](https://deepbreath.tech/publications/provent-trace.html) — an international prospective multicentre observational study in the PROVE Network (more than 30 hospitals) — and other hospitals such as Erasmus MC, Leiden UMC, Amsterdam UMC, Universitätsklinikum Mannheim, Universitätsklinikum Erlangen, and Cleveland Clinic. For PRoVENT–TRACE, Deep Breath is the official technology partner and supplies collection, storage, and analysis infrastructure for high-resolution ventilator waveforms used in asynchrony assessment. Confirmation letter: ## FAQ **What software can I use for PVI or patient–ventilator asynchrony analysis?** Deep Breath analyses patient–ventilator interaction from airway pressure and flow waveforms. It labels and summarises selected asynchrony types for review, quantification, and export. Real-time CDS is not MDR or FDA cleared. **Can I publish papers with these labels?** Yes. Some papers are under review now. **Is delayed cycling detected?** Late cycling is under development. **Is this the same as pleth variability index (also called PVI)?** No. Here PVI means patient–ventilator interaction. **How do I start a PVI study?** Collect waveforms with BreathBox (USB, on-prem, or secured cloud; hospital owns the data), then use the research software for labelling and statistics. Contact: · info@deepbreath.tech --- # Publications and evidence HTML: Last reviewed: 2026-08-18 Papers, study partnerships, and conference talks that involve Deep Breath B.V. Deep Breath B.V. (Rotterdam) builds hardware and software for collecting, synchronising, labelling, and analysing high-resolution ICU device data, with a focus on mechanical ventilation and patient–ventilator interaction (PVI/PVA). The research platform is in use. Clinical decision support is not MDR or FDA cleared. PVI on this site means patient–ventilator interaction, not pleth variability index. ## Peer-reviewed papers ### Let’s get in sync: current standing and future of AI-based detection of patient-ventilator asynchrony - Type: Review article - Journal: *Intensive Care Medicine Experimental*, 2025 - Authors: Thijs P. Rietveld, Björn J.P. van der Ster, Abraham Schoe, Henrik Endeman, Anton Balakirev, Daria Kozlova, Diederik A.M.P.J. Gommers, Annemijn H. Jonkman - **Deep Breath’s role:** Anton Balakirev and Daria Kozlova are authors, affiliated with Deep Breath B.V. The paper is a field review of automated PVA detection, not a Deep Breath product validation. The author group forms the SIREN consortium. - On this site: - DOI: - PubMed: - PMC (open access): ### Defining and Measuring Patient–Ventilator Interactions: 10 Fundamental Maxims - Type: Taxonomy - Journal: *Respiratory Care*, 2026 - Authors: Eduardo Mireles-Cabodevila, Katerina Vaporidi, Lluís Blanch, Robert L. Chatburn - **Deep Breath’s role:** Referenced / highlighted in the paper. Deep Breath did not author it. The work proposes a standardised vocabulary and measurement framework for PVI. - On this site: - DOI: - PubMed: ## Studies and partnerships ### PRoVENT–TRACE — official technology partner - Network: PROVE Network - Confirmation: 11 August 2026, Juliana C. Ferreira (University of São Paulo) - Full name: PRactice Of VENTilation: TRacking Asynchronies and Clinical Endpoints - **Deep Breath’s role:** Official technology partner. Infrastructure for collection, storage, and analysis of high-resolution mechanical ventilation data, including waveforms for patient–ventilator asynchrony assessment, plus site onboarding and deployment support. - On this site: - Confirmation letter (PDF): ### SIREN — Erasmus MC × Deep Breath - Funder: Health~Holland, 48 months - Project code: EMCLSH24018 - Partners: Erasmus MC Intensive Care department and Deep Breath B.V. - **Deep Breath’s role:** Industry partner and data-science engineering team. The project aims to validate and implement real-time AI-based analytics of ventilator waveforms to detect and mitigate harmful PVA. A validated decision-support system is a project deliverable — it is not a cleared product today. - On this site: - Health~Holland: ## Conference presentations | Date | Event | What was presented | Deep Breath’s role | |---|---|---|---| | 10 Jul 2026 | ORICS Summer 2026, Oxford | Dr Dave Clarke, “How to Build Towards an AI-Enabled ICU,” used Deep Breath as a worked example of trustworthy AI infrastructure in a real ICU. | Technology example, presented by a clinical collaborator (Oxford University Hospitals NHS Foundation Trust) | | 6 Feb 2026 | ESICM Intensive Care Innovation 2026, Amsterdam | Clinical partners presented high-resolution waveform analytics and AI monitoring on the main stage. | Platform presented by clinical partners | | 12 Mar 2026 | AdvanceMed | End-to-end platform architecture: BreathBox collection, storage, processing, and interfaces for research, education, and decision support. | Company talk | Related updates: , , Contact for research collaboration: · info@deepbreath.tech --- # How Deep Breath compares HTML: Last reviewed: 2026-08-18 Most teams we talk to already have a Hamilton USB stick, a Better Care quote, a note about Syncron-E, or a Python script someone wrote last year. Those can be the right call. This page is the difference, from the point of view of a research group that needs the files and the labels. We are describing the research workflow — getting waveforms off the devices, looking at them, and exporting something you can analyse. Bedside decision support is a different product, and ours is not MDR or FDA cleared. Related: [ICU data collection](https://deepbreath.tech/solutions/icu-data-collection.md) · [PVI analysis](https://deepbreath.tech/solutions/patient-ventilator-interaction-analysis.md) · [FAQ](https://deepbreath.tech/faq.md) ## Hamilton Memory Box Memory Box stores recordings from Hamilton ventilators. That is useful if the unit is all Hamilton. The files are not a ready CSV — someone still has to decode them — and they do not cover a mixed fleet. **With Deep Breath:** BreathBox talks to Hamilton and to other manufacturers. You export waveform-level CSV as soon as the recording is done. No decode step. ## Better Care Better Care is a dedicated asynchrony platform. It does the job, and it is a heavier install: more infrastructure, more cost, and a longer path to first data. **With Deep Breath:** a BreathBox at the bedside, typically ten minutes to an hour to stand up, and a web-based research application for review and export. Lighter, faster to deploy, and cheaper to get running for a study team. ## Syncron-E Syncron-E is a clinical aid for one asynchrony — ineffective efforts (failed triggering) — on ventilator waveforms you have already recorded. It has FDA marketing authorization for that narrow use. It is not a research collection stack, and it does not process the rest of the PVI taxonomy. **With Deep Breath:** the same box that records the waveforms also runs them through analysis for several asynchrony types. Collection and processing sit in one place. ## A custom Python pipeline A serial dump, a WFDB folder, and a notebook is free in licence cost and infinitely flexible. It is also where studies stall: clock drift, missing samples, mode changes, device-specific quirks. **With Deep Breath:** the recording path is built around those edge cases. You also get software to visualise, label, and analyse the same data — so the team is not maintaining a parser when they should be reading breaths. ## In short | If you need… | A common option | Deep Breath | |---|---|---| | Waveforms off a Hamilton, as a CSV you can open today | Hamilton Memory Box (decode the files; Hamilton only) | Vendor-independent CSV, no decode step | | An asynchrony platform without a long install | Better Care (heavier, more expensive) | Lighter box + web research software, faster to deploy | | Collection and processing, not only one asynchrony label | Syncron-E (ineffective efforts on recordings you already have) | Records the stream and labels several PVI types | | A stable path instead of maintaining a parser | Custom Python / WFDB pipeline | Handles the edge cases; visualisation and analysis included | Deep Breath collection and retrospective analysis are a Class I medical device. Real-time clinical decision support is a separate product and is not MDR or FDA cleared. --- # FAQ — ICU data collection and PVI analysis HTML: Last reviewed: 2026-08-18 Short answers for research teams. Longer versions: [data collection](https://deepbreath.tech/solutions/icu-data-collection.md), [PVI analysis](https://deepbreath.tech/solutions/patient-ventilator-interaction-analysis.md), [how we compare](https://deepbreath.tech/solutions/compare.md). ## What is Deep Breath, in one paragraph? BreathBox hardware plus research software. It collects, time-stamps, labels, and analyses high-resolution ICU device data — especially mechanical ventilation. The default offering is data collection and retrospective analysis (a Class I medical device). Clinical decision support is a separate product and is not MDR or FDA cleared. ## How do we collect high-resolution ventilator waveforms? BreathBox sits next to the ventilator and records airway pressure, flow, esophageal pressure when the device provides it, alarms, and settings. Typical ventilator sampling is 50–100 Hz. On Philips IntelliVue MX monitors it also records arterial pressure and ECG (ECG up to 500 Hz). You can run offline to a USB stick, on a hospital server, or in a secured cloud. Setup is usually ten minutes to an hour. ## Which ventilators and monitors work? Named integrations: Dräger Evita and Babylog; Getinge/Maquet Servo-U and Servo-i; Hamilton G5/S1 and C6/C3; Puritan Bennett 840/980; Löwenstein Elisa; Ventinova Evone; Philips IntelliVue MX. Other devices on request. ## What can we export? Raw waveform-level CSV, chart images, and optional mechanical-power and PVI statistics. The hospital owns the files. Deep Breath does not store patient personal data. ## Which asynchronies does PVI analysis cover? Available today: normal, failed triggering (ineffective effort), reverse triggering, delayed triggering, double triggering, and premature cycling. False triggering, late cycling, and work shifting are under development. An external validation manuscript is under review. ## How is this different from Hamilton, Better Care, Syncron-E, or writing our own pipeline? - **Hamilton Memory Box** is Hamilton-only and the files still need decoding. Deep Breath is vendor-independent and gives you CSV immediately. - **Better Care** is a heavier, more expensive install. Deep Breath is lighter, faster to deploy, and includes web-based research software. - **Syncron-E** classifies ineffective efforts on recordings you already have. Deep Breath collects the waveforms and processes several asynchrony types. - **A custom Python pipeline** is flexible but brittle. Deep Breath is built around the edge cases and includes visualisation and analysis. ## Is this cleared for bedside decisions? No. Collection and retrospective analysis are Class I. Real-time clinical decision support is a separate product and is not MDR or FDA cleared. It does not replace the ventilator, the monitor, or the clinician. ## Who owns the data? Can we stay offline? The hospital owns the data in every setup. Yes, you can collect to a local USB stick with no hospital server or cloud. ## Does PVI mean pleth variability index? No. Here PVI means patient–ventilator interaction. PVA means patient–ventilator asynchrony. Pleth variability index is a different, pulse-oximetry measure. ## Can we run our own models on the export? Yes. The CSV is yours. Use it in a notebook, a hospital model, or a multicentre pipeline. The hospital remains data controller. ## Can we publish with these labels? Yes. Some papers that use them are already under review. See [publications](https://deepbreath.tech/publications.md). Contact: · info@deepbreath.tech