According to the first systematic review and meta-analysis of its kind, artificial intelligence (AI) is just as good as healthcare professionals at diagnosing a disease based on a medical image. More high-quality studies, however, are required. A new study looks at the existing evidence to see if AI can diagnose illnesses as well as healthcare professionals.
Medicine and healthcare will be fundamentally altered by artificial intelligence (AI): Diagnostic patient data, such as ECG, EEG, or X-ray images, can be analyzed using machine learning, allowing diseases to be detected at an early stage based on subtle changes. However, implanting AI within the human body remains a significant technological challenge. Scientists have now developed a bio-compatible implantable AI platform that classifies healthy and pathological patterns in biological signals such as heartbeats in real-time. It detects pathological changes even when no medical supervision is present.
The researchers sought studies that compared the diagnostic effectiveness of deep learning algorithms to that of healthcare professionals when making a diagnosis based on medical imaging. They looked at the quality of the reporting in the studies, their clinical value, and the design of the studies. In addition, when comparing the AI’s diagnostic performance to that of healthcare professionals, the researchers looked at two outcomes: specificity and sensitivity.
AI will fundamentally change medicine and healthcare: Diagnostic patient data, e.g. from ECG, EEG, or X-ray images, can be analyzed with the help of machine learning so that diseases can be detected at a very early stage based on subtle changes.
Medicine and healthcare will be fundamentally altered by artificial intelligence (AI): Diagnostic patient data, such as ECG, EEG, or X-ray images, can be analyzed using machine learning, allowing diseases to be detected at an early stage based on subtle changes. However, implanting AI within the human body remains a significant technological challenge.
Scientists at the TU Dresden Chair of Optoelectronics have succeeded for the first time in developing a biocompatible implantable AI platform that classifies healthy and pathological patterns in biological signals such as heartbeats in real-time. It detects pathological changes even when no medical supervision is present. The findings of the study have now been published in the journal Science Advances.
The study discovered that AI can correctly diagnose disease in 87 percent of cases, while detection by healthcare professionals yielded an 86 percent accuracy rate. Deep learning algorithms had a specificity of 93 percent, while humans had a specificity of 91 percent.
The research team led by Prof. Karl Leo, Dr. Hans Kleemann, and Matteo Cucchi demonstrates a method for real-time classification of healthy and diseased bio-signals using a biocompatible AI chip in this paper. They used polymer-based fiber networks that are structurally similar to the human brain and enable reservoir computing, a neuromorphic AI principle. The random arrangement of polymer fibers creates a “recurrent network,” which allows it to process data in a manner similar to the human brain.
The nonlinearity of these networks allows them to amplify even the smallest signal changes, which are often difficult for doctors to evaluate in the case of the heartbeat, for example. The nonlinear transformation using the polymer network, on the other hand, makes this possible without any problems.
The AI was able to distinguish between healthy heartbeats and three common arrhythmias with an 88 percent accuracy rate in trials. The polymer network used less energy than a pacemaker during the process. The possibilities for implantable AI systems are numerous: They could, for example, be used to monitor cardiac arrhythmias or complications after surgery and report them to both doctors and patients via smartphone, allowing for immediate medical assistance.
Using computers to communicate is not a novel concept, but developing direct interfaces between technology and the human mind without the use of keyboards, mice, and monitors is a cutting-edge area of research with significant implications for some patients.
“With the development of so-called organic mixed conductors in recent years, the vision of combining modern electronics with biology has come a long way,” explains Matteo Cucchi, Ph.D. student and first author of the paper. “However, thus far, success has been limited to simple electronic components such as individual synapses or sensors.” So far, complex tasks have been impossible to solve. In our research, we have now taken a crucial step toward realizing this vision.
We have succeeded in solving complex classification tasks in real-time by harnessing the power of neuromorphic computing, such as reservoir computing used here, and we may be able to do so within the human body in the future. This approach will allow for the future development of intelligent systems that can aid in the saving of human lives.
