DeepFake electrocardiograms using generative adversarial networks are the beginning of the end for privacy issues in medicine

Abstract

Recent global developments underscore the prominent role big data have in modern medical science. But privacy issues constitute a prevalent problem for collecting and sharing data between researchers. However, synthetic data generated to represent real data carrying similar information and distribution may alleviate the privacy issue. In this study, we present generative adversarial networks (GANs) capable of generating realistic synthetic DeepFake 10-s 12-lead electrocardiograms (ECGs). We have developed and compared two methods, named WaveGAN* and Pulse2Pulse. We trained the GANs with 7,233 real normal ECGs to produce 121,977 DeepFake normal ECGs. By verifying the ECGs using a commercial ECG interpretation program (MUSE 12SL, GE Healthcare), we demonstrate that the Pulse2Pulse GAN was superior to the WaveGAN* to produce realistic ECGs. ECG intervals and amplitudes were similar between the DeepFake and real ECGs. Although these synthetic ECGs mimic the dataset used for creation, the ECGs are not linked to any individuals and may thus be used freely. The synthetic dataset will be available as open access for researchers at OSF.io and the DeepFake generator available at the Python Package Index (PyPI) for generating synthetic ECGs. In conclusion, we were able to generate realistic synthetic ECGs using generative adversarial neural networks on normal ECGs from two population studies, thereby addressing the relevant privacy issues in medical datasets.

Golden Standard or Obsolete Method? Review of ECG Applications in Clinical and Experimental Context

“…Data samples can be generated artificially by specially trained ML or DL models (such as Gaussian mixture model (GMM), generative adversarial network (GAN), LSTM/biLSTM, CNN), as has been shown for time-series ECG (including dependent multichannel signals) and 2D spectrogram applications (e.g., Lima et al, 2019 ; Brophy, 2020 ; Hatamian et al, 2020 ; Hazra and Byun, 2020 ). Recently, a unique database of more than 120,000 artificial ECGs, generated by the GAN, has been introduced ( Thambawita et al, 2021 ). This model was trained on more than 7,000 real patients’ ECG records.…”

Using a Visual Turing Test to Evaluate the Realism of Generative Adversarial Network (GAN)-Based Synthesized Myocardial Perfusion Images

“…Another risk includes the malicious use of artificial intelligence (AI)-based deepfake technology [5]. Recently, Thambawita and colleagues have reported that a realistic electrocardiogram could be synthesized by deepfake technologies [6]. Although the authors positively interpret their results as the end of privacy issues in medicine, the same result can be seen as the beginning of the confusion unless the generated data are distinguishable from the real objects.…”

DeepFake knee osteoarthritis X-rays from generative adversarial neural networks deceive medical experts and offer augmentation potential to automatic classification

“…Currently, deep learning in medicine pertains mainly to clinical decision support and data analysis. By analyzing medical data for underlying patterns and relationships, deep learning systems have a broad range of applications, ranging from patient outcomes prediction 4-6 , diagnostics and classification [7][8][9][10] , and data segmentation 11,12 to the generation [13][14][15][16][17] and anonymization of datasets [18][19][20][21] with synthetic medical data.Policy and regulatory directives concerning medical data privacy and use continue to be updated globally. The US Health Insurance Portability and Accountability Act (HIPAA) 22 is similar to the General Data Protection Regulation (GDPR) 23 ; both were developed to restrict data flow and ascertain patient consent for health data dissemination.…”

“…The US Health Insurance Portability and Accountability Act (HIPAA) 22 is similar to the General Data Protection Regulation (GDPR) 23 ; both were developed to restrict data flow and ascertain patient consent for health data dissemination. GDPR is the strictest policy 16,24 concerning medical data and is implemented in addition to any EU national data policies. Such approaches further complicate implementation and downstream relevance to research groups.…”

“…Currently, deep learning in medicine pertains mainly to clinical decision support and data analysis. By analyzing medical data for underlying patterns and relationships, deep learning systems have a broad range of applications, ranging from patient outcomes prediction 4-6 , diagnostics and classification [7][8][9][10] , and data segmentation 11,12 to the generation [13][14][15][16][17] and anonymization of datasets [18][19][20][21] with synthetic medical data.…”