Artificial Intelligence for Prognostic Modelling and Adaptive Treatment Monitoring in Radiation Oncology.
Authors
Affiliations (6)
Affiliations (6)
- Department of Radiology, Durga Diagnostic Centre, Maharashtra University of Health Sciences, Nashik, IND.
- Department of Radiodiagnosis, Punjab Institute of Liver and Biliary Sciences, Mohali, IND.
- Department of Radiodiagnosis, Government Medical College Jammu, Jammu, IND.
- Department of Radiology, School of Medical Sciences, Sri Satya Sai University of Technology and Medical Sciences, Sehore, IND.
- Department of Pharmacy Practice, Teerthanker Mahaveer College of Pharmacy, Teerthanker Mahaveer University, Moradabad, IND.
- Department of Agadatantra and Vidhi Vaidyaka, Rajiv Gandhi University of Health Sciences, Bengaluru, IND.
Abstract
Artificial intelligence (AI) is increasingly being used in radiation oncology to help doctors predict patient outcomes and monitor treatment response. However, its routine use in clinical practice is still limited because available studies are not always consistent, methods are not fully standardised, and many AI tools have not been tested widely in real-world settings. This narrative review examines how AI is used for prognosis, adaptive radiotherapy, treatment response assessment, and toxicity prediction in radiation oncology. This review also considers AI applications in cancer screening, radiological diagnosis, and radiology-histopathology correlation, as these areas directly support prognostic modelling and adaptive treatment decisions. A structured literature search from 2015 to 2025 was performed across major biomedical databases, with attention to radiomics, machine learning, deep learning, response modelling, and adaptive treatment planning. Studies were reviewed for their design, validation methods, and clinical outcomes. Current evidence suggests that AI can improve risk prediction, support automatic tumour and organ segmentation, track changes during treatment, and identify early signs of toxicity better than some conventional approaches. However, many studies still lack external validation and multicentre data. Challenges also remain in making AI models easy to understand and compatible with existing clinical systems. Combining imaging data with genomic information and radiation dose parameters may further improve prediction. In clinical practice, AI may help personalise radiation dose, support timely treatment plan adjustment, and improve resource use. Wider adoption will require stronger validation, standardised workflows, and clear model governance. Overall, AI should be used as a decision-support tool to assist clinicians rather than replace clinical expertise.