tOLIet: Single-lead Thigh-based Electrocardiography Using Polimeric Dry Electrodes

Abstract

Our team previously introduced an innovative concept for an "invisible" Electrocardiography (ECG) system, incorporating electrodes and sensors into a toilet seat design to enable signal acquisition from the thighs. Building upon that work, we now present a novel dataset featuring real-world, single-lead ECG signals captured at the thighs, offering a valuable resource for advancing research on thigh-based ECG for cardiovascular disease assessment. To our knowledge, this is the first dataset of its kind.

The tOLIet dataset comprises 149 ECG recordings collected from 86 individuals (50 females, 36 males) with an average age of 31.73 ± 13.11 years, a mean weight of 66.89 ± 10.70 kg, and an average height of 166.82 ± 6.07 cm. Participants were recruited through direct contact with the Principal Investigator at Centro Hospitalar Universitário de Lisboa Central (CHULC) and via clinical consultations conducted at the same institution. Each recording includes four differential signals acquired from electrode pairs embedded in the toilet seat, with reference signals obtained from a standard 12-lead hospital ECG system.

Background

In the early 20th century, infectious diseases were the leading cause of death, responsible for nearly half of all fatalities. However, as societies advanced both socioeconomically and culturally, the primary causes of mortality have shifted toward chronic non-communicable diseases (NCDs) [1]. These conditions develop gradually over a lifetime, often progressing unnoticed until they pose serious health risks and impact quality of life. Major NCDs include cardiovascular diseases, chronic respiratory conditions (such as asthma, bronchitis, and rhinitis), hypertension, cancer, diabetes, and metabolic disorders like obesity and dyslipidemia [2, 3]. According to the World Health Organization (WHO), NCDs now account for 63% of global deaths [4]. Despite these concerning figures, most chronic illnesses can be prevented or effectively managed with early diagnosis and appropriate treatment, allowing individuals to maintain a higher quality of life. While genetic predisposition plays a role, preventive strategies are key in reducing the risks associated with NCDs. Many of these diseases can be delayed or prevented through proactive health measures, particularly in relation to cardiovascular diseases, cancer, diabetes, and chronic respiratory conditions. However, lifestyle modifications—central to prevention—can be difficult to implement without support. Public health initiatives and behavior-change programs at both the societal and individual levels play a crucial role in addressing this challenge [5].

Technological advancements have become deeply embedded in daily life, particularly in health monitoring. Consumers increasingly rely on smart devices—such as smartwatches, rings, and fitness trackers—to track their health in real-time. These wearables, which often incorporate advanced sensors, provide valuable physiological insights. Currently, approximately 20% of U.S. residents own a smart wearable device, and the global market is expected to grow at a compound annual rate of 25%, reaching $70 billion by 2025 [6]. While clinical adoption of wearable health technology remains in its early stages, it has seen rapid progress, as reflected in Gartner’s Hype Cycle for emerging technologies [7]. The COVID-19 pandemic further accelerated this trend, driving the expansion of telehealth solutions [8].

This study was motivated by the need to explore innovative methods for cardiovascular monitoring, specifically through the integration of health-tracking technology into everyday objects. We propose a novel, non-intrusive system embedded in a toilet seat, a household fixture used regularly and consistently, making it well-suited for passive health monitoring. Unlike wearable devices, which require continuous user engagement and often result in high dropout rates, a toilet seat-based system allows for seamless, automated data collection[9,10]. Continuous and long-term monitoring is particularly valuable for detecting intermittent cardiovascular events that conventional 12-lead ECGs often miss due to their limited recording duration [11,12].

Our single-center retrospective study aimed to assess the feasibility of using thigh-based ECG for cardiovascular event detection. Specifically, we evaluated the accuracy of the acquired signals, the reliability of this monitoring approach, and its potential for clinical application. By implementing this type of passive monitoring systems, we believe that it is possible to contribute to enhance early detection, reduce hospital visits, and improve long-term disease management. This innovative approach could significantly improve patient adherence to cardiovascular monitoring while contributing to better clinical decision-making and overall health outcomes [8, 13,14].

Methods

Methods

Participants

The study included 86 participants, all of Portuguese nationality, with a gender distribution of 50 females and 36 males. The average age was 31.73 ± 13.11 years, with a mean weight of 66.89 ± 10.70 kg and height of 166.82 ± 6.07 cm. Participants were recruited through direct contact with the principal investigator at CHULC and during clinical consultations at the same institution.

While some participants reported possible cardiac conditions, none of these cases were confirmed by a specialist. The study aimed to include a diverse range of ages and physical characteristics to enhance the dataset’s representativeness. To the best of our knowledge, this is the first publicly available dataset featuring thigh-based electrocardiography (ECG) signals.

Experimental Setup and Protocol

Data acquisition was performed using a smart toilet seat equipped with four sensors, each containing a pair of electrodes to capture single-lead ECG signals (ECG EXP). Different electrode surface textures were tested to evaluate their impact on signal quality—further details can be found in [12].

Each recording session lasted up to 5 minutes, during which participants ensured direct skin contact with the electrodes. The collected ECG data was transmitted via Bluetooth to a computer running OpenSignals (r)evolution software for recording and visualization. In selected cases, a 12-lead ECG was simultaneously recorded using a GE HealthCare MAC800 clinical-grade system (ECG REF) for comparative analysis.

The data acquisition system, from sensor output onwards, was based on the scientifically validated BITalino (PLUX, S.A., Lisbon, Portugal) development kit [13,15]. The ECG signal was processed with:

• Amplification: Analog front-end gain of 11,000
• Filtering: Band-pass filter with a frequency range of 0.5–40 Hz
• Sampling rate: 1 kHz, with 10-bit resolution for all four acquisition channels

For data analysis and reporting, we utilized Python 3.11.7, with the following libraries:

• BioSPPy (2.2.3) for digital filtering, segmentation, and feature extraction
• PyHRV (0.4.0) for Heart Rate Variability (HRV) analysis

Unless otherwise specified, mean values and standard deviations were computed for all participants.

Data Description

The dataset is organized to include raw ECG data, the code used for processing and analysis, and all graphs and tables generated throughout the research process, ensuring reproducibility and transparency.

Dataset Structure

1. ECG EXP (Experimental Data)

This folder contains TXT files extracted from the smart toilet seat sensors. Each file is structured as follows:

• Column 1: nSeq – Numerical sequence for detecting signal loss.
• Columns 2 & 3: I1 & I2 – BITalino digital inputs (not used).
• Columns 4 & 5: O1 & O2 – BITalino digital outputs (not used).
• Column 6: A1 – ECG signal from sensor A1 (flat electrode texture).
• Column 7: A2 – ECG signal from sensor A2 (sinusoidal electrode texture).
• Column 8: A3 – ECG signal from sensor A3 (pyramidal electrode texture).
• Column 9: A4 – ECG signal from sensor A4 (trapezoidal electrode texture).
• Columns 10 & 11: A5 & A6 – BITalino analog inputs (not used).

2. ECG REF (Reference System Data)

This folder contains files extracted from the clinical reference ECG system. Data characteristics can be extracted using the read_ref_data.py script, which returns:

• id – Subject identification code.
• acquisitionDateTime – Date and time of data collection.
• name_dev – Lead identification (e.g., I, II, III, AVR, AVL, AVF, V1–V6).
• meanTemplate – Average heartbeat model for each lead.
• dev – Time-series ECG signal for each lead over a 10-second period.
• unit – Signal unit (µV).

3. Scripts

The Scripts folder contains three main Python scripts used for signal processing and analysis:

• read_ecg_data.py – Reads raw physiological signals from ECG EXP, transforming them into structured dictionaries containing session matrix data and annotated segments.
• read_ref_data.py – Reads raw physiological signals from ECG REF, processing them into structured dictionaries similar to ECG EXP.
• data_analysis.py – Generates the characteristics extracted from the ECG complexes

4. Spreadsheet: dataSet.csv

This spreadsheet contains demographic and study-related metadata, including:

• ID
• Age
• Weight
• Height
• Gender
• Observations field

Usage Notes

Our dataset captures ECG signals recorded in an unconventional configuration, specifically from the thighs. The primary objective is to characterize these signals and assess the potential of this novel derivation. Since this type of ECG acquisition is largely unexplored in the literature, our dataset offers a unique opportunity to analyze previously unstudied data. Additionally, given the extensive number of collected samples, this dataset is well-suited for implementing deep learning models, which rely on large datasets for optimal performance.

This system builds upon our previous research, which has explored various aspects of thigh-based ECG acquisition—ranging from the surface textures on signal quality [12] and influence of electrode materials [16] to the feasibility of using single-lead ECG signals for biometric recognition [17].

However, the data collection process and the necessary hardware/software introduce certain challenges and constraints, which we detail below. Due to the intimate nature of this acquisition method, careful consideration was given to the location of the tests. Data collection took place in bathrooms or private hospital rooms, as participants needed to remove their pants or skirts to ensure direct skin contact with the electrodes. It is important to note that data was not recorded during actual toilet use—the smart toilet seat was installed on a standard toilet bowl, and participants sat for a maximum duration of five minutes per session.

To ensure real-world applicability, participants were given minimal instructions beyond ensuring skin-electrode contact, allowing for natural seating positions. However, several factors can influence signal quality, including bodily hair, skin care products (e.g., moisturizing creams), and individual physical characteristics. These variables may affect physiological responses, and future studies should consider incorporating such information to further refine the system and its applicability.

Ethics

This study received approval from the Ethical Committee of the Centro Hospitalar Universitário de Lisboa Central (CHULC), Portugal, under reference number INV496. The institutional Data Protection Officer conducted a risk assessment and determined that the potential risks to participants' rights and freedoms were minimal. This conclusion was based on several factors, including the voluntary nature of participation, informed decision-making, and robust data anonymization measures.

The dataset consists solely of physiological signal data, each linked to individual participants through unique study-specific identification codes that are not associated with personal identities, ensuring complete anonymity. Participants were fully briefed on the study’s objectives, data collection methods, and confidentiality safeguards before providing informed consent. No financial incentives were offered for participation.

Acknowledgements

This work is financed by National Funds through the Portuguese funding agency Fundação para a Ciência e a Tecnologia, I.P. (FCT, https://ror.org/00snfqn5816) within projects UID/50008: Instituto de Telecomunicações and LA/P/0063/2020, and grant 2022.10245.BDANA (DOI 10.54499/LA/P/0063/2020 | https://doi.org/10.54499/LA/P /0063/2020). The authors would also like to thank OLI - Sistemas Sanitários S.A. for providing all the resources and support that made this work possible.

Conflicts of Interest

The authors declare no competing interests.

References

1. Wagner KH, Brath H. A global view on the development of non communicable diseases. Prev Med. 2012;54 Suppl:S38–41.
2. Beaglehole R, Bonita R, Horton R, Adams C, Alleyne G, Asaria P, et al. Priority actions for the non-communicable disease crisis. Lancet. 2011;377(9775):1438–47.
3. World Health Organization. Top 10 causes of death fact sheet. World Health Organization; 2020. Available from: https://www.wareable.com/health-and-wellbeing/{ECG}-heart-rate-monitor-watch-guide-6508. Accessed January 2025.
4. Barold S. Willem Einthoven and the birth of clinical electrocardiography a hundred years ago. Card Electrophysiol Rev. 2003;7(1):99–104.
5. dos Santos Silva A, Correia MV, da Silva HP. Invisibles: A New Frontier in Vital Signs Monitoring. NATO Sci Peace Secur Ser D Inf Commun Secur. 2024.
6. Carreiras C, Lourenço A, Aidos H, da Silva HP, Fred AL. Morphological ECG analysis for attention detection. In: Proc Int Joint Conf Comput Intell (IJCCI); 2013. p. 381–90.
7. Batista D, Lourenço A, Silva HP, Fred AL, Martins R. Benchmarking of the bitalino biomedical toolkit against an established gold standard. Healthc Technol Lett. 2019;6(2):32–6.
8. Silva AS, Correia MV, de Melo F, da Silva HP. Identity recognition in sanitary facilities using invisible electrocardiography. Sensors (Basel). 2022;22(11):4201.
9. da Silva HP, Lourenço A, Fred A, Raposo N, de Sousa MA. Check Your Biosignals Here: A new dataset for off-the-person ECG biometrics. Comput Methods Programs Biomed. 2014;113(2):503–14.
10. Canento F, Fred A, Silva H, Gamboa H, Lourenço A. Multimodal biosignal sensor data handling for emotion recognition. In: Proc IEEE Sensors Conf; 2011. p. 647–50.
11. Pinto J, Fred A, da Silva HP. Biosignal-based multimodal emotion recognition in a valence-arousal affective framework applied to immersive video visualization. In: Proc Annu Int Conf IEEE Eng Med Biol Soc (EMBC); 2019. p. 3577–83.
12. dos Santos Silva A, Almeida H, da Silva H, Oliveira A. Design and evaluation of a novel approach to invisible electrocardiography (ECG) in sanitary facilities using polymeric electrodes. Sci Rep. 2021;11:6222.
13. da Silva HP, Guerreiro J, Lourenço A, Fred A, Martins R. Bitalino: A novel hardware framework for physiological computing. In: Proc Int Conf Physiol Comput Syst. 2014;2:246–53.
14. Lourenço A, Silva H, Leite P, Lourenço R, Fred AL. Real time electrocardiogram segmentation for finger based ECG biometrics. In: Proc Int Conf. 2012. p. 49–54.
15. Pereira Coutinho D, Plácido da Silva H, Gamboa H, Fred A, Figueiredo M. Novel fiducial and non-fiducial approaches to electrocardiogram-based biometric systems. IET Biom. 2013;2(2):64–75.
16. dos Santos Silva A, Correia MV, Costa A, da Silva HP. Towards industrially feasible invisible electrocardiography (ECG) in sanitary facilities. In: Proc IEEE Portuguese Meeting on Bioengineering (ENBENG); 2023. p. 1–4.
17. dos Santos Silva A, Correia MV, da Silva HP. Evaluation of biometric template permanence for electrocardiography (ECG) based user identification in sanitary facilities. In: Proc IEEE Mediterranean Electrotechnical Conf (MELECON); 2024. p. 288–93.