BiomedBench

General Information

BiomedBench is an open-source benchmark suite of TinyML biomedical applications targeting low-power wearbles [1].

This repository contains a set of biomedical applications designed to run in low-power wearable platforms for patient monitoring.

All applications are coded in C/C++. For each application, we include a Desktop version and the ported versions to different commercial platforms.

Applications

• Heartbeat Classifier (HeartBeatClass)

  The HeartBeatClass [2] detects abnormal beating patterns in real time for common heart diseases using the ECG signal. For further details check [1].

• Seizure Detector SVM (SeizureDetSVM)

  The SeizureDetSVM [3] works on ECG input and recognizes real-time epileptic episodes. For further details check [1].

• Seizure Detector CNN (SeizureDetCNN)

  The SeizureDetCNN [4] is based on EEG data and detects real-time epileptic seizure episodes. For further details check [1].

• Cognitive Workload Monitor (CognWorkMon)

  The CognWorkMon [5] is designed for real-time monitoring of the cognitive workload state of a subject and is based on EEG input. For further details check [1].

• Gesture Classifier (GestureClass)

  The GestureClass [6] aims to classify hand gestures by inspecting signals captured by sEMG of the forearm. For further details check [1].

• Cough Detector (CoughDet)

  The CoughDet [7] is a novel application using non-invasive chest-worn biosensors to count the number of cough episodes people experience per day, thus providing a quantifiable means of evaluating the efficacy of chronic cough treatment. For further details check [1].

• Emotion Classifier (EmotionClass)

  The EmotionClass [8] classifies patients’ fear status to prevent gender-based violence based on three physiological signals: Galvanic skin response (GSR), PPG, and skin temperature (ST). For further details check [1].

• Biological backpropagation-free (BioBPfree)

  BioBPfree [9] is the only benchmark that performs on-device training. For further details check [1].

Considered Platforms

Currently, the considered boards and their MCUs are:

• Raspberry Pi Pico featuring RP2040
• Nucleo-L4R5ZI featuring STM32L4R5ZI
• Ambiq Apollo3 Blue AMA3BEVB featuring the Apollo 3 Blue
• Gapuino featuring GAP8
• GAP9_EVK featuring GAP9
• X-Heep on the FPGA board PYNQ-Z2

Repository Structure

The folder structure look like that:

├── Applications/
|   ├── <App_name>/
|   |   ├── <App_version>/

|   |   |   ├── single_core/
|   |   |   |   ├── <platform_name>
|   |   |   |   |   ├── Inc/
|   |   |   |   |   ├── Src/
|   |   |   |   |   ├── Makefile
|   |   |   |   |   ├── Readme.md

|   |   |   ├── multicore/
|   |   |   |   ├── <platform_name>
|   |   |   |   |   ├── Inc/
|   |   |   |   |   ├── Src/
|   |   |   |   |   ├── Makefile
|   |   |   |   |   ├── Readme.md

Look at the Readme of each platform folder (./Applications/.../<platform_name>/Readme.md) for more information on how to run the application on each platform.

Issues and Troubleshooting

If you find any problems or issues with the applications, please check out the issue tracker and create a new issue if your problem is not yet tracked.

Useful links

For an overview of BiomedBench, check out: BiomedBench website

For more details, check out the paper preprint: BiomedBench article

References

1. Samakovlis, Dimitrios, et al. "BiomedBench: A benchmark suite of TinyML biomedical applications for low-power wearables", 2024.
2. Rubén Braojos, Giovanni Ansaloni, and David Atienza. 2013. A methodology for embedded classification of heartbeats using random projections. In 2013 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, Grenoble, France, 899-904.
3. Farnaz Forooghifar, Amir Aminifar, and David Atienza Alonso. 2018. Self-Aware Wearable Systems in Epileptic Seizure Detection. In DSD 2018. IEEE, Prague, Czech Republic, 426-432.
4. Catalina Gomez, Pablo Arbelaez, Miguel Navarrete, Catalina Alvarado-Rojas, Michel Le Van Quyen, and Mario Valderrama. 2020. Automatic seizure detection based on imaged-EEG signals through fully convolutional networks. Scientific Reports 10 (12 2020).
5. Renato Zanetti, Adriana Arza, Amir Aminifar, and David Atienza. 2022. Real-Time EEG-Based Cognitive Workload Monitoring on Wearable Devices.IEEE Trans. Biomed. Eng. 69, 1 (2022), 265-277.
6. Mattia Orlandi, Marcello Zanghieri, Victor Javier Kartsch Morinigo, Francesco Conti, Davide Schiavone, Luca Benini, and Simone Benatti. 2022. sEMG Neural Spikes Reconstruction for Gesture Recognition on a Low-Power Multicore Processor. In 2022, IEEE Biomedical Circuits and SystemsConference (BioCAS). IEEE, Taipei, Taiwan, 704-708.
7. Orlandic L, Thevenot J, Teijeiro T, Atienza D. A Multimodal Dataset for Automatic Edge-AI Cough Detection. Annu Int Conf IEEE Eng Med Biol Soc. 2023 Jul:2023:1-7.
8. Jose Angel Miranda Calero, Rodrigo Marino, Jose M. Lanza-Gutierrez, Teresa Riesgo, Mario Garcia-Valderas, and Celia Lopez-Ongil. 2018. Embedded Emotion Recognition within Cyber-Physical Systems using Physiological Signals. In 2018 Conference on Design of Circuits and Integrated Systems (DCIS). IEEE, Lyon, France, 1-6.
9. Saleh Baghersalimi, Alireza Amirshahi, Tomas Teijeiro, Amir Aminifar, and David Atienza. 2023. Layer-Wise Learning Framework for Efficient DNN Deployment in Biomedical Wearable Systems. In 2023 IEEE 19th International Conference on Body Sensor Networks (BSN). IEEE, Boston, MA, USA, 1-4.