Over the past decade, we have seen a fast growth of data generated from medical and healthcare devices. A very important task in the computational area is to extract reliable surrogates of conventional physiological markers. For example, the heart rate variability has been shown indicative of the autonomic nervous system. Recently, promising surrogates started to emerge from the complicated transformation from a bundle of signals or medical images based on artificial intelligence and big data analytics. For example, researchers have tried to integrate the clinical, imaging, and genomic data to look for the genotype-phenotype association to provide new clues for clinical outcome evaluation.
The concept of digital biomarkers is related to the machine learning application to physiology and medicine, the latent features of a deep learning model could be regarded as a digital marker. Besides, influential adding values, for example, being more sensitive than a conventional marker for early intervention, or extracted from convenient devices for everyday are expected. Of note, considering the personal difference that exists in almost all kinds of bio-signals, a potential marker should be statistically significant.
There are two major questions on this Research Topic. The first one is how to use these diverse sources of signals/information to discover novel digital biomarkers, it includes but is not limited to the following sub-topics:
1. Physiological, bio-signals, and medical image data sharing and application pipeline bench-marking
2. Novel digital biomarker exploration with open dataset(s)
3. Novel digital biomarker based on private datasets and its preliminary results
The other one is how to enhance the reliability of the marker. As we know, the domain shift problem, for example, changes in imaging systems or ECG channels may significantly hamper the performance of the AI systems, which raises doubts about the validity of the markers. Therefore, we also welcome the manuscripts focusing on the following sub-topics:
4. Generalization enhancement of the AI system that generates the digital biomarkers
5. Interpretable AI for digital biomarkers
Review papers focusing on the pertinent areas are also welcome.
Over the past decade, we have seen a fast growth of data generated from medical and healthcare devices. A very important task in the computational area is to extract reliable surrogates of conventional physiological markers. For example, the heart rate variability has been shown indicative of the autonomic nervous system. Recently, promising surrogates started to emerge from the complicated transformation from a bundle of signals or medical images based on artificial intelligence and big data analytics. For example, researchers have tried to integrate the clinical, imaging, and genomic data to look for the genotype-phenotype association to provide new clues for clinical outcome evaluation.
The concept of digital biomarkers is related to the machine learning application to physiology and medicine, the latent features of a deep learning model could be regarded as a digital marker. Besides, influential adding values, for example, being more sensitive than a conventional marker for early intervention, or extracted from convenient devices for everyday are expected. Of note, considering the personal difference that exists in almost all kinds of bio-signals, a potential marker should be statistically significant.
There are two major questions on this Research Topic. The first one is how to use these diverse sources of signals/information to discover novel digital biomarkers, it includes but is not limited to the following sub-topics:
1. Physiological, bio-signals, and medical image data sharing and application pipeline bench-marking
2. Novel digital biomarker exploration with open dataset(s)
3. Novel digital biomarker based on private datasets and its preliminary results
The other one is how to enhance the reliability of the marker. As we know, the domain shift problem, for example, changes in imaging systems or ECG channels may significantly hamper the performance of the AI systems, which raises doubts about the validity of the markers. Therefore, we also welcome the manuscripts focusing on the following sub-topics:
4. Generalization enhancement of the AI system that generates the digital biomarkers
5. Interpretable AI for digital biomarkers
Review papers focusing on the pertinent areas are also welcome.