Model

Compact decodable ECG text encoding for LLMs (Dorukbaşı & Türker)

PreprintCatalog paper date: Nov 17, 2026Added to atlas: Nov 18, 2026Source

Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.

Reusable signal-to-text method: ≤4,080 tokens per PTB-XL record (mean 2,866), exact round trip, decoded-signal correlation 0.982.

Signals
ECG
Year
2026
Architecture

Deterministic key-point selection (cardiac-cycle tolerance) + 26-letter log amplitude codebook + exact time gaps; 8 leads encoded, 4 limb leads recovered algebraically; 10-step tolerance schedule for a fixed token budget (abstract)

At a glance
Type
Task model
Modalities
ECG
Form factor
Not reported
Parameters
Not reported
Architecture
Deterministic key-point selection (cardiac-cycle tolerance) + 26-letter log amplitude codebook + exact time gaps; 8 leads encoded, 4 limb leads recovered algebraically; 10-step tolerance schedule for a fixed token budget (abstract)
Venue
Not reported
Year
2026
DOI
Not reported
Code
Not released
Weights
Not released
Licence
Not reported
Caveats
Not reported
Datasets
Added
Nov 18, 2026
Updated
Nov 18, 2026

Abstract

When a twelve-lead electrocardiogram is written as text for a large language model, the number of tokens has to be balanced against the information lost through compression. This paper presents a deterministic encoding that writes eight independent leads as amplitude letters and exact time gaps and recovers the remaining four limb leads algebraically. The key points are selected with a cardiac-cycle-dependent tolerance, the amplitudes are parameterized by a 26-letter logarithmic codebook learned from a training set, and a 10-step tolerance schedule keeps all records within a fixed token budget. The encoding was applied to 21,388 records from the public PTB-XL database and evaluated using an exact decode-and-re-encode round trip, waveform reconstruction error, and a paired comparison of convolutional classifiers trained with 5 different seeds on signals from the original and decoded databases. All records were contained in 4,080 tokens (average: 2,866 tokens, 64.3% fewer than a per-sample transcription), and all stored lead strings were exactly reproduced. The average correlation between the decoded and reference signals was 0.982. The macro-averaged area under the receiver operating characteristic curve was 0.9167 for the reference signals and 0.9157 for the decoded signals (paired difference -0.0010; 95% patient-level bootstrap interval, -0.0028 to 0.0007), within the pre-specified non-inferiority margin of -0.010; this is not a confirmatory result, because the test fold had been used during encoder development. The diagnostic content of a compact text can therefore be preserved, while the language model's ability to read the text is yet another matter.

Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)

Compact decodable ECG text encoding for LLMs (Dorukbaşı & Türker): ChatPPG research summary card

How it works

How Compact decodable ECG text encoding for LLMs (Dorukbaşı & Türker) works: Input signals → Model → Downstream tasksInput signals: ECG (Electrocardiogram). Dataset: PTB-XL. Model: Compact decodable ECG text encoding for LLMs (Dorukbaşı & Türker) (Convolutional network · Language-aligned).Input signalsECG: ElectrocardiogramECGElectrocardiogramDataset: PTB-XLDatasetPTB-XLModelCompact decodable ECG text…Convolutional network ·…

Datasets linked in the hub; pretraining vs evaluation roles follow the paper.

Architecture
Deterministic key-point selection (cardiac-cycle tolerance) + 26-letter log amplitude codebook + exact time gaps; 8 leads encoded, 4 limb leads recovered algebraically; 10-step tolerance schedule for a fixed token budget (abstract)
Inputs
ECG
Approach
tokenizer · signal to text · ecg llm · method

Datasets used· 1

Papers· 1

Cite this model

Cite the publication recorded for this model. Check it against the publisher before submitting.

Dorukbaşı, E., & Türker, İ. (2026). Compact and Decodable Electrocardiogram Text for Large Language Models. arXiv preprint. https://arxiv.org/abs/2611.22827
Dorukbaşı, Elif and İlker Türker. 2026. "Compact and Decodable Electrocardiogram Text for Large Language Models." arXiv preprint. https://arxiv.org/abs/2611.22827.
Dorukbaşı, E. and Türker, İ. (2026) 'Compact and Decodable Electrocardiogram Text for Large Language Models', arXiv preprint. Available at: https://arxiv.org/abs/2611.22827
Dorukbaşı, Elif, and İlker Türker "Compact and Decodable Electrocardiogram Text for Large Language Models." arXiv preprint, 2026. https://arxiv.org/abs/2611.22827.
E. Dorukbaşı and İ. Türker, "Compact and Decodable Electrocardiogram Text for Large Language Models," arXiv preprint, 2026, [Online]. Available: https://arxiv.org/abs/2611.22827.
@misc{dorukbas2026compact,
  title         = {{Compact and Decodable Electrocardiogram Text for Large Language Models}},
  author        = {Dorukbaşı, Elif and Türker, İlker},
  year          = {2026},
  howpublished  = {arXiv preprint},
  eprint        = {2611.22827},
  archivePrefix = {arXiv}
}

Source metadata: 2026

Record updated Nov 18, 2026

History

  1. Added to the ChatPPG research hub