A machine learning-based approach supported by HJ-biplot statistical analysis reveals novel avenues of artificial intelligence in oncological research

Artificial intelligence (AI) is being used in oncology at an exponential rate. However, unless the intricacies of its research landscape are clarified, AI's potential in cancer research may remain unrealized. To circumvent the bias associated with human-assisted review, we adopted a combination of Latent Dirichlet Allocation (LDA) with HJ-Biplot to examine 11149 documents published in Web of Science from 2003 to 2023. The probabilistic topic LDA algorithm automatically found up to 32 research thematic trends with the highest semantic quality. The HJ-Biplot further confirmed that automated imaging and computer-assisted diagnosis of tumors were emerging scientific fields showing great promise in guiding precision oncology. Furthermore, the analysis highlighted the AI's capacity to assess a wide range of data types, which is necessary for cohort studies to develop new predictive models. Other AI applications included evaluating treatment response, testing novel molecular therapies in clinical trials, laparoscopy, and robotics in surgery. The socioeconomic and cultural impact of cancer were neglected topics that require additional AI research. Consequently, we proposed an enhanced topic analytical framework more accurate than simple probabilistic topic modeling and bibliometrics in examining novel research paths, because the LDA model gained an extra analytical boost using the HJ-Biplot tool.