Harnessing multi-source data for AI-driven oncology insights: Productivity, trend, and sentiment analysis

Abstract

This study aims to provide an overall view of the current status of AI publications in the entire field of oncology, encompassing productivity, emerging trends, and researchers’ sentiments. A total of 1,296 papers published between January 2019 and January 2024, were selected using the PRISMA framework. Citespace software and the R package “Biblioshiny” were utilized for bibliographic analysis. China has been the leading contributor to global production with over 2,596 publications, followed by Europe. Among 8339 authors, Kather JN was the third most prolific author and held a central position in the co-authorship network. The most prominent article emphasized the Explainability of AI methods (XAI) with a profound discussion of their potential implications and privacy in data fusion contexts. Current trends involve the utilization of supervised learning methods such as CNN, Bayesian networks, and extreme learning machines for various cancers, particularly breast, lung, brain, and skin cancer. Late image-omics fusion was the focus of various studies during 2023. Recent advancements include the use of "conductive hydrogels" and "carbon nanotubes" for flexible electronic sensors. Ninety and a half percent of the researchers viewed these advancements positively. To our knowledge, this study is the first in the field to utilize merged databases from WoS, Scopus, and PubMed. Supervised ML methods, Multimodal DL, chatbots, and intelligent wearable devices have garnered significant interest from the scientific community. However, issues related to data-sharing and the generalizability of AI algorithms are still prevalent.

Ähnliche Arbeiten