Artificial intelligence is accelerating cancer research by pairing distinct computational techniques with distinct problems: computer vision reads medical images, machine learning and deep learning find patterns in genomic and clinical data, and natural language processing mines the unstructured text buried in health records and scientific literature. A review published in npj Precision Oncology, authored by Jun Li and colleagues, describes AI as a rapidly growing tool in modern oncology with promising uses in early diagnosis and drug discovery, spanning machine learning, deep learning, reinforcement learning, natural language processing, and generative models. The National Cancer Institute frames the same trajectory in research terms: AI methods are being developed to predict patient outcomes, analyze epidemiological and behavioral data, and ultimately accelerate the effort to end cancer as we know it.
The AI Revolution in Oncology: An Overview
The clearest way to understand AI's growing footprint is to look at the breadth of functions it now touches. A review of AI applications across cancer care, summarized in a historical analysis of the convergence between cancer research and AI, identified five major areas of application: imaging and radiomics, genomics and biomarker discovery, drug discovery and repurposing, clinical decision support, and patient monitoring. That spread is deliberate rather than incidental. As a broader synthesis of AI applications in cancer explains, the integration of AI in research and clinical practice encompasses advancing screening and diagnostic accuracy, improving precision treatment, enhancing cancer surveillance, accelerating drug discovery, optimizing healthcare delivery, and elucidating cancer mechanisms—all with the collective aim of improving patient outcomes and streamlining clinical workflows.
What unites these functions is AI's defining advantage: its ability to integrate and find patterns in massive, diverse datasets—CT and MRI scans, stained tissue slides, genomic sequences, and free-text clinical notes—whose volume and complexity exceed what human analysts can process unaided. Different data types demand different techniques. Convolutional neural networks, a deep learning architecture, dominate imaging tasks; graph neural networks handle structured molecular and genomic relationships; language models extract meaning from text. The result is not one monolithic AI tool but a family of specialized methods, each matched to a specific stage of the cancer research pipeline.
