By 2028, a lot of cardiac remote-monitoring clinics will probably have some quiet little browser add-on doing the clerical grunt work in the background while humans handle the parts that actually require judgment, nerves, and maybe a functioning coffee machine.

That is why PASTEC is interesting. Not because it unveils a flashy new heart-detecting super-brain, but because it tackles the far less glamorous problem that keeps plenty of medical AI stuck in PowerPoint: getting real clinical data, inside real workflows, without turning the staff into unpaid data janitors.

The heart monitor has entered the group chat

Patients with cardiac implantable electronic devices, or CIEDs, like pacemakers and defibrillators, generate a steady stream of remote data. That is good for safety and follow-up, but it also means clinicians must sift through piles of transmissions on vendor-specific platforms. The tech promised efficiency. The tech also brought homework.

That workload problem is not hypothetical. A 2023 real-world study from a large tertiary center logged 7,087 remote CIED transmissions plus 1,212 implantable loop recorder transmissions over the study period, with nearly half of CIED transmissions arriving as automatic alerts [3]. In other words, the inbox is not calm. The inbox is doing cardio.

PASTEC, developed by Sacristan and colleagues, tries to solve the plumbing problem. Their platform sits on top of existing remote-monitoring systems through a browser extension. It performs client-side pseudonymization, lets staff add structured annotations, and stores the cleaned data in a centralized academic database for research and future AI validation [1]. Translation: instead of building a whole new hospital IT universe, they slipped a useful layer into the one clinicians already inhabit.

Not a new AI model. A runway for future ones.

This is the key point. PASTEC is not mainly saying, "look at our amazing algorithm." It is saying, "here is a way to collect, label, and safely evaluate data so algorithms can someday earn their keep."

That matters because cardiology AI has a validation problem. A 2024 scoping review of prospective human validation studies in cardiology found that most such studies were published only after 2020, and that randomized trials were still relatively uncommon [6]. Another 2024 scoping review concluded there is still a gap between impressive AI papers and full-scale clinical integration [4]. Which is academia's elegant way of saying, "cool demo, now make it survive Tuesday morning clinic."

PASTEC goes after exactly that gap.

What the study actually found

The authors tested the platform during routine activity in a high-volume remote-monitoring center and compared one week using PASTEC annotations with one week of standard review. They reviewed 1,276 recordings, including 697 with structured annotation [1].

The practical question was simple: does this system slow clinicians down so much that everyone quietly hates it?

A little, yes. Too much, no.

Median processing time was 5.31 seconds per recording with annotation versus 2.88 seconds without annotation, an absolute difference of 2.43 seconds per recording [1]. When extrapolated across the week, that came to about 59 extra minutes total, spread across two operators [1]. In hospital terms, that is not nothing. But it is also not the kind of workflow disaster that gets a tool deleted by Friday.

Just as important, the pseudonymization held up in their review: no identifiable patient information was found in sampled pseudonymized recordings [1]. That is a big deal, because healthcare AI is not just about model accuracy. It is also about whether your legal, ethical, and IT people stop making the same face you make when Reviewer 2 asks for "just one more ablation study."

Why this matters outside one French center

Remote monitoring of CIEDs already has strong clinical support, and reviews keep pointing to better follow-up, earlier detection, and fewer unnecessary in-person visits when used well [2]. But "used well" is carrying a lot of emotional weight there. The systems still create operational burden, and AI cannot help much if the data are scattered across proprietary silos or never labeled in a usable way.

That is where PASTEC feels more substantial than it first appears. It is building the boring middle layer between device dashboards and prospective AI evaluation. Boring middle layers, to be clear, are how civilization functions. Sewers are not sexy either.

The broader AI-cardiology field is moving fast. Reviews published in 2025 describe growing use of AI for telemetry, arrhythmia detection, and predictive monitoring, but they also keep repeating the same warning labels: workflow fit, trust, governance, local validation, and human oversight still matter a lot [5]. Industry is moving too, with FDA-cleared or updated AI-enabled cardiac monitoring products from companies such as AliveCor and iRhythm in 2024 [7,8]. The message is consistent: the models are arriving, but deployment lives or dies on infrastructure.

The catch, because there is always a catch

This was a single-center usability study over a short period, not proof that PASTEC improves outcomes or that any attached AI model will work across hospitals, vendors, or patient populations [1]. It shows feasibility, not final victory. Which is fair. If medical AI papers stopped at "we built a secure, workflow-compatible data layer and did not make the nurses revolt," that would honestly be more useful than half the hype cycle.

PASTEC's real contribution is modest in the best way. It does not promise a robot cardiologist. It offers a practical scaffold for testing whether future AI tools deserve a seat in clinic at all.

And in healthcare AI, that kind of humility is not a bug. It is the part that might actually ship.

References

1. Sacristan B, Strik M, Ploux S, Hocini M, Bordachar P, Dubois R, Duchateau J. PASTEC: an open clinical infrastructure for data aggregation and prospective validation of AI in cardiac remote monitoring. npj Digital Medicine. 2026. DOI: 10.1038/s41746-026-02702-6. PubMed: 42120570

2. Reinhardt A, Ventura R. Remote Monitoring of Cardiac Implantable Electronic Devices: What is the Evidence? Current Heart Failure Reports. 2023;20(1):12-23. DOI: 10.1007/s11897-023-00586-1. PMCID: PMC9877501

3. Sane M, Annukka M, Toni J, Elina P, Charlotte A, Eeva T, Leena K, Pekka R, Jarkko K. Real-life data on the workload of cardiac implantable electronic device remote monitoring in a large tertiary center. Pacing and Clinical Electrophysiology. 2023;46(9):1109-1115. DOI: 10.1111/pace.14792. PubMed: 37486912

4. Makimoto H, Kohro T. Adopting artificial intelligence in cardiovascular medicine: a scoping review. Hypertension Research. 2024;47(3):685-699. DOI: 10.1038/s41440-023-01469-7. PubMed: 37907600

5. Lu J, Xiao R, Hu X, Do DH. Artificial intelligence in cardiac telemetry. Heart. 2025;111(19):897-903. DOI: 10.1136/heartjnl-2024-323947. PubMed: 40122590

6. Prospective Human Validation of Artificial Intelligence Interventions in Cardiology: A Scoping Review. JACC: Advances. 2024;3(9 Part 2):101202. DOI: 10.1016/j.jacadv.2024.101202

7. AliveCor Announces Dual FDA Clearance of AI Technology That Delivers 35 Cardiac Determinations and First-of-its-Kind Kardia 12L ECG System. Diagnostic and Interventional Cardiology. June 25, 2024. Link: dicardiology.com

8. iRhythm Technologies Receives FDA 510(k) Clearance for Design Modifications to Its Zio AT Device. August 30, 2024. Link: investors.irhythmtech.com

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.