An international team of researchers have proposed a standardized registry for I in biomedicine. This is to improve the reproducibility of results and create trust to use AI-based algorithms in biomedical research, and in clinical practice in the future.
Earlier, in the previous decades, the development of new technologies has enabled to create a wide spectrum of systems that can generate massive amounts of biomedical data, for example in cancer research. The development of completely new possibilities at the same time being created for examining and evaluating the data using artificial intelligence methods.
If AI algorithms are used in intensive care units, it can anticipate circulatory failure at an early stage on the basis of massive amounts of data from several monitoring systems. This data is obtained by processing lot of complex information from different sources concurrently, which is way beyond human capabilities.
Owing to the great potential of AI systems, it leads to an excessively large number of biomedical AI applications. Meanwhile, the corresponding reports and publications do not always stick to best practices or offer only incomplete information about the algorithms used or the source of the data. Thus, the assessment and comprehensive comparisons of AI modes become difficult.
This is because the conclusions of AI are not always comprehensible to humans and results are seldom fully reproducible. The situation is untenable, in particular for clinical research, where transparency and trust in AI models and research reports are critical to increase how AI algorithms are accepted and to develop improved AI methods for simple biomedical research.
To present a solution to this, a team of researchers have proposed AIMe registry for artificial intelligence in biomedical research.
