Page 119 - TD-4-3

Tumor Discovery Radiation oncologists in AI era

anatomical accuracy still remain key areas for further ensure informed consent for AI use, and respect patients’
research and standardization. The generative adversarial rights to refuse AI-based interventions. In addition, the
networks algorithm has shown promise in generating environmental impact of AI systems, including energy
clearer, more detailed synthetic images, offering an consumption and e-waste, must be considered. 16,17
advantage in image clarity compared to traditional In conclusion, AI is expected to play a central role in
methods. 8,9
shaping the future of radiation oncology. Personalized
AI-assisted adaptive radiotherapy offers significant treatment protocols, biomarker-driven dose adaptation,
advantages for patients with frequent anatomical changes, and even fully autonomous treatment planning may
such as those with head and neck, lung, or gynecological become the norm. However, strong ethical oversight, legal
cancers. Daily cone beam CT imaging allows real-time frameworks, and sustainable implementation models will
evaluation, while AI-based auto-contouring enables rapid be essential for this integration. Radiation oncologists
delineation of tumors and organs at risk. Treatment plans will need not only medical expertise but also a solid
can be updated instantly based on anatomical shifts. understanding of AI technologies. As the field evolves, the
However, limitations such as suboptimal image quality and focus will shift from technical tasks to clinical decision-
the need for expert validation of AI-generated contours making and patient-centered care, making it essential for
highlight the continued importance of human oversight in radiation oncologists to redefine their roles and actively
clinical decision-making. 10 integrate into multidisciplinary care teams to remain
Prognostic estimations in radiation oncology have indispensable in an increasingly automated landscape.
traditionally relied on clinical and anatomical data.
However, AI-based models now enable more precise Conflict of interest
predictions by integrating biological, clinical, dosimetric, The author declares that she has no conflict of interest.
treatment, and imaging data. 11,12 Toxicity prediction is
equally crucial for developing personalized treatment References
plans, minimizing both acute and late side effects, and 1. Wang L, Chen X, Zhang L, et al. Artificial intelligence in
ultimately improving patient quality of life. 13,14 Predictive clinical decision support systems for oncology. Int J Med Sci.
models, especially those using hybrid approaches, 2023;20(1):79-86.
combine radiomic/dosiomic features with clinical and
dosimetric parameters to improve both prognosis and doi: 10.7150/ijms.77205
toxicity prediction. Challenges include ensuring clinically 2. Nafees A, Khan M, Chow R, et al. Evaluation of clinical
acceptable model accuracy, ensuring data diversity, and decision support systems in oncology: An updated
navigating regulatory requirements, such as Food and systematic review. Crit Rev Oncol Hematol. 2023;192:104143.
Drug Administration approval. doi: 10.1016/j.critrevonc.2023.104143
As AI continues to reshape the landscape of radiation 3. Erden MB, Cansiz S, Caki O, et al. FourierLoss: Shape-
oncology, it is clear that the role of the radiation oncologist aware l function with Fourier descriptors. Neurocomputing.
will evolve. With AI automating routine tasks, clinicians 2025;638:130155.
15
are expected to focus more on patient monitoring and doi: 10.1016/j.neucom.2025.130155
complex decision-making. This shift necessitates new
competencies, including understanding AI algorithms, 4. Chen M, Wu S, Zhao W, Zhou Y, Zhou Y, Wang G.
interpreting data, and considering ethical implications. Application of deep learning to auto-delineation of
target volumes and organs at risk in radiotherapy. Cancer
Radiation oncologists will need to collaborate with Radiother. 2022;26(3):494-501.
multidisciplinary teams and develop skills in data literacy
and clinical validation of AI tools. doi: 10.1016/j.canrad.2021.08.020
As AI becomes increasingly integrated into radiation 5. Matoska T, Patel M, Liu H, Beriwal S. Review of deep
oncology, several ethical challenges must be addressed. learning based autosegmentation for clinical target volume:
Current status and future directions. Adv Radiat Oncol.
These include ensuring patient privacy through data 2024;9(5):101470.
anonymization and cybersecurity, defining accountability
in AI-assisted clinical decisions, and mitigating bias arising doi: 10.1016/j.adro.2024.101470
from non-representative training datasets. Equitable 6. Wang TW, Hong JS, Huang JW, Liao CY, Lu CF, Wu YT.
access to AI tools is a concern, as is the potential erosion Systematic review and meta-analysis of deep learning
of physicians’ decision-making autonomy. It is crucial applications in computed tomography lung cancer
to maintain a strong patient–physician relationship, segmentation. Radiother Oncol. 2024;197:110344.

Volume 4 Issue 3 (2025) 111 doi: 10.36922/TD025200039