BFU scientists have determined that medical professionals prefer to use simple machine learning algorithms over deep learning AI. This is due to the fact that artificial intelligence does not give experts the opportunity to interpret the results, but the final opinion must be given by the doctor to avoid mistakes. We identified which AI algorithms are being used in different medical fields. As a result, this article could serve as his reference manual to help medical staff choose the best machine learning method for a given task. The results of this research have been published in the International Journal of Environmental Research and Public Health.
AI technology is now widely used in medicine. They help doctors detect photo abnormalities such as tumors, select individual treatment options for each patient, and monitor patient conditions in real time. Beyond that, AI is being used to develop robotic medical devices such as wearable sensors that take various measurements and send data to doctors.
Scientists from the Immanuel Kant-Baltic Federal University (Kaliningrad), together with colleagues from the National Medical-Surgical Center named after NI Pirogov (Moscow) and the Russian Institute of Health (Moscow) analyzed their mention. to determine which AI algorithms are most popular in medicine. His 10,000 research articles from the PubMed database.
The authors proposed an approach that can recover different structures of datasets by creating networks of co-occurrences. Using a program called VOSviewer, the scientist visualized a network that reflected the co-occurrence of medical terms and concepts about his AI algorithm. As key concepts, scientists used words and collocations such as “artificial intelligence,” “machine learning,” “deep learning,” “brain,” “prediction,” and “tomography.” It turns out that all the terms he can classify into five topic clusters.
The signal processing cluster included concepts related to electroencephalography (EEG) and electrocardiography (ECG), where AI is commonly used. Machine clusters and deep learning clusters combined the working principles and approaches of various AI systems. At the same time, deep learning can be viewed as a form of machine learning, the main feature of which is its use in AI. It is not used in regular machine learning and the corresponding systems work with simpler mathematical algorithms.
An image processing cluster is a combination of algorithms used to analyze medical images and related areas such as magnetic resonance imaging. A final cluster of retrospective studies was devoted to concepts related to reality assessment and reproducibility of results.
The largest machine learning clusters had a huge amount of connections with sets of other terms. As a result, the most frequently connected clusters were deep learning, image processing, and retrospective studies. On the one hand, these connections indicate that such algorithms are commonly used for the analysis of medical images, especially neuroimaging. On the one hand, it has been shown to be used to predict diseases such as tumor growth.
Research shows that deep learning, or AI, is rarely used in medicine, and AI can be used to help analyze images. This is due to the fact that such medical AI decisions are difficult to articulate. In other words, looking at the “black box” analogy, medical AI technology should be transparent. Physicians, while using an algorithm, must be able to recover why the algorithm reached such conclusions. Deep learning technologies, on the other hand, often leave little room for doctors to interpret the results obtained. The same authors determined that deep learning-based methods are popular in cardiovascular disease research. Simpler, but more easily interpreted, machine learning algorithms are most commonly used for signal analysis, such as EEG and ECG, and for predicting disease onset.
In addition, the authors determined that “visualization” and “health parameter” were the most common words combined with AI terms in medical articles. This indicates that these techniques are mainly used for image processing and analysis of human condition based on various body characteristics.
“Our findings may help medical professionals to choose the best AI algorithms for their research activities. Thus, our article is a useful tool for physicians and IT professionals working in the medical AI field. It may serve as a reference guide for both , and the proposed method may be useful for complex data analysis in other scientific fields. We plan to use the data we receive to expand the range of machine learning algorithms we use to diagnose brain diseases.”— says Alexander Framov, PhD, Professor and Chief Scientific Officer of Physics and Mathematics at the Baltic Center for Neurotechnology and Artificial Intelligence at Immanuel Kant-Baltic University.
For more information:
Oleg E. Karpov et al., Analysis of Publication Activity and Research Trends in the Field of AI Medical Applications: A Network Approach, International Journal of Environmental Research and Public Health (2023). DOI: 10.3390/ijerph20075335
Courtesy of Immanuel Kant-Baltic Federal University
