A Comprehensive Review: Artificial Intelligence in Oncology — From Diagnostics to Treatment Selection and Clinical Workflow

Artificial Intelligence (AI) has transformed the realm of oncology by improving cancer diagnosis, treatment planning, and management.1,2 By harnessing machine learning, deep learning, and natural language processing (NLP), AI has been seamlessly integrated into conventional oncology, leading to improved diagnostic accuracy, reduced variability, and the advancement of personalized medicine.3,4 Through the integration of AI with medical imaging, digital pathology, and genomic studies, AI algorithms have demonstrated high accuracy in cancer detection and classification.5,6 For example, deep learning models can analyse large medical imaging datasets and identify subtle patterns associated with malignancy that may be missed by human observers, thereby supporting early detection and timely intervention.11,12 AI algorithms also contribute to cancer risk stratification and severity assessment, enabling more accurate outcome prediction and prognosis estimation in cancer patients.13 In treatment planning, AI plays a crucial role in precision oncology by incorporating genomic and molecular profiles to tailor therapeutic decisions according to an individual patient’s cancer type and biological characteristics.7,31 Additionally, AI systems can process vast amounts of clinical data, medical literature, and prior patient outcomes to provide evidence-based recommendations that assist clinicians in optimizing treatment strategies.8,14 Beyond clinical decision-making, AI significantly impacts oncology workflows by automating repetitive and time-consuming tasks such as intelligent triaging and auto-documentation15 AI-assisted triage systems prioritize patients based on urgency, ensuring that those with critical conditions receive timely care.16 Automated documentation tools reduce administrative burden, allowing healthcare professionals to dedicate more time to direct patient care.15 Despite these advances, challenges remain related to data quality, algorithmic bias, model interpretability, and the “black-box” nature of certain AI systems, underscoring the need for further research and clinical validation before widespread implementation.9,10