DOI: 10.66106/syzyal.20250203 ISSN: 3105-6857

衰老与再生医学中人工智能应用的现状与挑战（Current situation and challenges of artiffcial intelligence application in aging and regenerative medicine）

张明昭 Mingzhao Zhang
Show PDF Cite
This paper focuses on the ffeld of aging and regenerative medicine. The core work is to analyze the mechanism of de-generative changes in the body and restore tissue function. The research covers multiple levels of molecules, cells and tissues. The data is large and heterogeneous. Traditional analysis methods can not effectively integrate this kind of information. AI is good at complex pattern recognition and high-dimensional data mining. It has been preliminarily applied in the construction of aging clock, the discovery of regeneration targets and the optimization of tissue engineering parameters, which has brought a new technical path to related research. At present, there are still big limitations in the application: there is no uniffed standard for multi-source data, the generalization ability of the model in cross species and cross organization scenarios is insufficient, and the results of in-depth learning cannot be explained from the biological mechanism level, making many models stay in the proof of concept stage, which is far from the clinical and transformation needs. In view of these shortcomings, this paper focuses on the key methodological issues in the cross ffeld of artiffcial intelligence and aging regenerative medicine, proposes to build an interpretable analysis framework integrating mul-timodal data, and also discusses how to establish more stringent speciffcations in data governance, model design and validation. The main cause of the problem is that the aging and regeneration process itself has dynamic and networked characteristics. The existing algorithms are mostly based on static correlation learning, and lack of description of causal chain; Domain knowledge is not fully em-bedded in the model structure, making it difficult for the output to obtain biological support. This paper proposes to introduce causal inference, prior knowledge constraints and standardized benchmark data sets to improve.

More from our Archive