Current Advances in AI for Cancer Care: A Review of Predictive Models for Patient Readmission

AI and ML have been successfully applied to cancer care and have shown possibilities to enhance the existing approach to patient management by increasing effectiveness of prediction, allowing correct diagnosis of early-stage cancer, as well as customising treatment plans. This review is therefore based on recent studies capturing different aspects of AI in oncology including 30-day readmission prediction, the role of various factors on the cancer patients’ survival and novel non-invasive diagnosis. Research showed that AI models such as CNNs, RNNs, random forest, and ensemble learning approaches have significantly outperformed traditional statistic-based models in terms of prediction and decision-making. Through the use of multi-modal data inpatient data models such as EHRs, SDOH, and genomics data, the models offer an encompassing insight of patient characteristics and healthcare outputs. Next, based on the analysis of experts’ insights, the following innovations are defined as disruptive solutions enhancing early detection and promoting engagement: machine learning breath analyzers and healthcare chatbots. Many of these interventions may alleviate inequalities in healthcare, and / or maximize the utilization of health resources, as well as decrease the stress that overwhelms clinical personnel. Nevertheless, many problems persist still ranging from data privacy to algorithmic explainability, ethical use and utilization of AI in real clinical environments. However, these issues can only be solved if there is inter-disciplinary work cut across with good ethical standards in matters concerning technology. Owning to this review, there herein lies a call to support ongoing AI and ML research based on their effectiveness in oncology and the need for improvement of these technologies for efficiency and increased access to healthcare for cancer patients.