Artificial intelligence in oncology

Recent advancements in AI (artificial intelligence) associated with computer vision, natural language processing, and multimodal data analysis are poised to help optimize and improve patient oncologic care. For medical image analysis, the workflows require preprocessing of biomedical images prior to data analysis with either traditional machine learning classifiers (e.g., SVM) or, more recently, deep learning neural networks, to achieve the downstream task, which include segmentation, feature extraction, classification, and risk prediction. Additionally, advances in AI methods to analyze genomic sequencing data have enabled unprecedented discoveries. These methods rely heavily on data quantity and quality. Thus, it is critical to acquire large and generalizable datasets for model development. However, data heterogeneity, access, and the lack of annotations remain challenging problems for training robust models ready for clinical implementation. Through collaborative efforts, data data-sharing initiatives, and advances in AI methodologies, AI applications for oncology may find clinical adoption and improve patient outcomes.