Exploring the literature on artificial intelligence use in oncology.

Abstract

e13642 Background: Artificial Intelligence (AI) allows machines to analyze data in a way that mimics human intelligence. AI has been used in oncology diagnostics such as radiology and pathology. However, the specific processes and tasks that AI can support, especially in oncology, have not been fully explored. Methods: The electronic databases PubMed and MEDLINE were searched from September 1, 2018, to September 1, 2023, for publications related to the use of AI in oncology. Inclusion criteria consisted of English language, article type “study” and keywords Artificial Intelligence, Healthcare, Predictive/Prediction, Fever, Cytokine Release Syndrome, Nursing, Patient Care, Oncology, and Neutropenic Fever applied in a progressive manner to narrow the search. Results: The term Artificial Intelligence yielded over 124,385 studies. This number was reduced to 466 when adding “healthcare, predictive/prediction, and oncology.” Adding “patient care” further reduced the number to 102. Out of the 102, seven studies identified included the term “nursing” and focused on clinician decision support, readmissions, palliative care, and scheduling. One identified study discussed the potential benefits of AI-generated algorithms for risk stratification related to cytokine release syndrome. The keywords “neutropenic fever” did not generate any results. Six studies were identified by adding in the term “fever,” which focused on AI use in specialties outside of oncology. The literature reviewed shows that AI has had limited use in oncology outside of radiology and pathology. Only one study suggested the use of AI would be beneficial in predicting symptoms related to treatment. Small studies were identified showing the value of AI in increasing adherence to schedules, reducing readmissions, and providing precision medicine. Conclusions: AI is a promising tool for the future of oncology. The current availability of AI within oncology practices has been limited. Further studies should be undertaken to identify how AI can reduce provider burnout, care costs, and early identification of symptoms to prevent adverse events.