CSPM EDP Sciences logo
Access to submission and peer review system
Search
Menu
All issues Volume 5 (2024) Vis Cancer Med, 5 (2024) 8 Full HTML
Open Access
Issue
Vis Cancer Med
Volume 5, 2024
Article Number 8
Number of page(s) 4
DOI https://doi.org/10.1051/vcm/2024011
Published online 26 November 2024

© The Authors, published by EDP Sciences, 2024

Licence Creative CommonsThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Artificial intelligence (AI) has been increasingly applied in cancer prevention [1], cancer diagnosis and patient prognosis prediction [2], radiotherapy treatment planning [3], and radiotherapy implication [4]. In response to the call from Visualized Cancer Medicine for collecting the most important questions in cancer research and clinical oncology [5], the following three key questions are proposed in this article related to integrating AI into radiation therapy for cancer patients.

Q105: How can we develop individualized radiation therapy based on biological variations combined with artificial intelligence for better treatment outcomes and less treatment toxicity?

Most radiation therapy plans are optimized and reviewed based on dose-volume constraints recommended by guidelines, resulting in standardized plans. However, the sensitivity of different patients to radiotherapy is different. Even when treated for the same tumor type and with similar dose-volume parameters, the radiotherapy toxicity after treatment can differ significantly among individuals. Therefore, the “one-size-fits-all” approach is no longer the best treatment choice and individualized precision radiotherapy has gradually attracted attention [6]. The studies of low-dose radiation therapy (< 1Gy/fraction, LDRT) illustrate the discontinuity of the dose-response curve [7]. Preclinical studies and clinical radiotherapy have suggested that LDRT might have favorable immunological effects such as macrophage differentiation and an increased infiltration of effector immune cell [8, 9]. These biological factors, which are often ignored in the design of radiotherapy plans, can explain a more accurate dose-response relationship. Personalized treatment plans based on the individual characteristics of the patient and tumor have the potential to minimize toxicity and maximize benefit.

Recent research has incorporated biological response models, such as normal tissue complication probability (NTCP) and generalized equivalent uniform dose (gEUD), into the design of radiation therapy plans, to minimizing predicted radiation risks [10]. These studies offer a new perspective on personalized radiation therapy plan design and aim to improve treatment outcomes. Meanwhile, molecular profiling of the primary tumor, which may be associated with tumor radiosensitivity [11], is achievable via the application of minimally invasive biopsy coupled with high throughput sequencing, e.g., fine needle aspiration followed by various sequencing technologies [12, 13]. It is therefore reasonable to expect that molecular profiling of adjacent normal tissues to the tumor could be also achievable via similar technologies.

There are some challenges in integrating biological markers into the process of personalized radiation therapy: (1) Extract and select the biological markers related to tumor radiosensitivity or biological response. The large scale of basic information collected from one single patient could not be timely handled by human beings to keep the treatment plan executed within an acceptable time window. Artificial intelligence possesses efficient and powerful capabilities for large-scale data mining and analysis, and it has been used for disease diagnosis and prognosis prediction. We believe that, in the face of high-throughput biomarkers, the application of artificial intelligence is helpful to identify key relevant biomarkers. (2) Develop and validate algorithms for predicting the sensitivity of tumors and adjacent normal tissues to radiotherapy by combining the information from radiotherapy parameters and the molecular profiles of both primary tumors and normal tissues. The influence of biological markers on the dose-response relationship has not been clearly defined. Artificial intelligence, particularly neural networks, can fit nonlinear complex relationships to obtain prediction models with enhanced performance. However, during the validation and implementation of prognostic prediction models, it is essential to carefully consider that the variations in data sets might affect the model’s generalization ability. If necessary, it is advisable to collect the local center’s data set in advance for calibration. (3) Make the decision and design the radiotherapy plan based on models for predicting radiotherapy sensitivity. In the process of radiotherapy planning, prescription dose, dose fractionation, and limits for organs at risk are all related to radiotherapy sensitivity [14]. Artificial intelligence algorithms can effectively integrate multi-dimensional information and use the established accurate response model to provide more effective guidance, therefore obtaining plans that minimize toxicity and maximize the benefits of radiotherapy.

In summary, assistance from artificial intelligence is therefore expected to accelerate the whole procedure. Artificial intelligence algorithms can facilitate the completion of complex and time-consuming tasks, such as the processing of large-scale molecular profiling data and the rapidly delineating regions of interest. Additionally, they can aid in modeling and decision-making, thereby enhancing the accuracy of radiotherapy and improving potential treatment outcomes. Combining molecular profiling and artificial intelligence should pave the road for future personalized radiotherapy treatment planning based on biological variation with better treatment outcomes and less treatment toxicity.

Q106: Can artificial intelligence improve real-time dose monitoring and adjustments in radiotherapy?

An optimal radiation therapy plan can achieve excellent outcomes only when executed accurately. However, during the implementation of the plan, various factors – such as organ movements (e.g., respiratory motion, gastrointestinal peristalsis), mechanical errors, and positioning discrepancies – can lead to deviations of the target area from its intended location, and the irradiation of normal tissues with high doses. Consequently, monitoring these changes throughout the therapy process and ensuring the precise execution of the plan is crucial for achieving effective treatment outcomes. Current strategies commonly involve controlling and tracking target movements to maintain anatomical positioning as closely as possible to the planned. Nevertheless, the actual dose delivered to the patient remains unknown with these methods. Additionally, many tracking techniques are invasive, which restricts their application.

With advancements in the field of computer vision, artificial intelligence is expected to improve real-time, non-invasive target tracking, dose reconstruction, and treatment plan adjustment [15]. Markerless tracking, which utilizes deep learning utilizing various types of real-time medical images such as ultrasound, 4D CT, and X-ray, has been reported in numerous studies [16–18]. Additionally, accurate and rapid motion prediction algorithms can be used to deal with the delays caused by system positioning and action time. Compared to traditional dose calculation algorithms, dose reconstruction based on artificial intelligence is faster and can satisfy the time constraints required for real-time plan adjustments [19]. The radiotherapy plan optimization and adjustment technology based on artificial intelligence exhibits similar dosimetric parameters while requiring less time compared to conventional techniques [20, 21].

In addition, anatomical changes, such as tumor shrinkage and weight loss, during the radiotherapy treatment process could result in normal tissues and organs at risk of being involved in the high-dose irradiation area, thereby increasing radiation toxicity. The application of artificial intelligence technology to predict or monitor anatomical changes can assist physicians in quickly determining the actual dose distribution based on the predicted anatomical changes. This information is crucial for assessing whether the radiotherapy plan needs to be adjusted according to the dose received by normal tissues. Lee et al. [22] developed a ConvLSTM model to predict the deformation vector fields (DVFs) in the subsequent CT images based on the planning CT and weekly CBCTs, and then designed and adjusted the adaptive plan based on these DVFs.

The radiosensitivity of tumor tissues for the same patient changes dynamically during radiotherapy. With advancements in biological imaging techniques, including functional imaging, molecular imaging, and metabolic imaging, we are expected to visualize the physiological and metabolic changes in tissues and cells during radiotherapy. These imaging techniques help workers identify regions that require high doses more accurately [23], while artificial intelligence algorithms can help to explore the critical connections within the workflow. For instance, artificial intelligence tools can efficiently and accurately perform automatic delineation of target volumes combined with PET/CT imaging to determine precise target volumes [24]. These technologies are anticipated to provide an effective technical foundation for real-time dose detection and plan adjustment, however, their stability and clinical applicability still require further investigation and validation.

Q107: Can molecular profiling plus artificial intelligence be helpful for predicting the benefits of adjusting the plan in adaptive radiotherapy?

During radiotherapy treatment, anatomical changes such as weight loss, shrinkage of tumor volume, and physiological movement of organs at risk may occur. Adaptive radiotherapy has been proposed and applied in clinical to adapt to these changes and the main goal is to ensure target coverage and reduce treatment toxicity. While adaptive radiation therapy (ART) improves patient outcomes [25], it requires re-delineation of target areas and the re-optimization of treatment plans, resulting in additional workload for practitioners as well as higher time and economic costs [26, 27]. This complexity poses challenges for the broader adoption of this technology.

Currently, the decision to implement ART is contingent upon the degree of target area shrinkage and anatomical changes observed in current versus planned imaging, relying heavily on experience and subjective judgment. Some patients may experience minimal changes in treatment outcomes, such as tumor control or normal tissue toxicity, before and after adjustments, which can lead to wasted treatment time and resources [28, 29]. Therefore, predicting the patient benefits at the time of adaptive planning adjustments is a critical issue that requires further research.

For this purpose, more accurate dose-response prediction models for the treatment targets and surrounding normal tissue are needed [30]. Meanwhile, biological and molecular responses often precede the anatomical changes suggestive of ART and molecular profiling may provide clinically relevant prognostic information [31]. With the advances in tumor analysis next-generation sequencing and other profiling technologies [32], we believe that molecular profiling of the treatment targets as well as the adjacent normal tissue might be helpful for generating more accurate predicting algorism. In the interest of efficiency, artificial intelligence is again expected to play a critical role in conducting in-time dose-effect analyses prior to plan adjustment, which can promote adaptive planning to a more accurate level.

Funding

This research did not receive any specific funding.

Conflicts of interest

The authors declare that they have no conflicts of interest in relation to this article.

Data availability statement

Data sharing is not applicable to this article.

Author contribution statement

Ting Song is responsible for conceptualization, supervision, and manuscript writing, Huali Li is responsible for manuscript editing and proofreading.

Ethics approval

Ethical approval was not required.

References

  1. Leatherdale ST, Lee J. Artificial intelligence (AI) and cancer prevention: the potential application of AI in cancer control programming needs to be explored in population laboratories such as COMPASS. Cancer Causes Control. 2019;30:671–675. [CrossRef] [PubMed] [Google Scholar]
  2. Huang S, Yang J, Fong S, Zhao Q. Artificial intelligence in cancer diagnosis and prognosis: Opportunities and challenges. Cancer Lett. 2020;471:61–71. [CrossRef] [PubMed] [Google Scholar]
  3. Jones S, Thompson K, Porter B, Shepherd M, Sapkaroski D, Grimshaw A, et al. Automation and artificial intelligence in radiation therapy treatment planning. J Med Radiat Sci. 2024;71(2):290–298. [CrossRef] [PubMed] [Google Scholar]
  4. El Naqa I. AI applications in radiation therapy and medical physics. In: Valdes G, Xing L, editors. Artificial intelligence in radiation oncology and biomedical physics, Boca Raton: CRC Press; 2023. p.1–23. [Google Scholar]
  5. Qian CN, Pezzella F, Lu Z. Revisiting and proposing the most important questions in cancer research and clinical oncology. Vis Cancer Med. 2024;5:E1. [CrossRef] [EDP Sciences] [Google Scholar]
  6. Overgaard J, Aznar MC, Bacchus C, Coppes RP, Deutsch E, Georg D, et al. Personalised radiation therapy taking both the tumour and patient into consideration. Radiother Oncol. 2022;166:A1–A5. [CrossRef] [PubMed] [Google Scholar]
  7. Le Reun E, Foray N. Low-dose radiation therapy (LDRT) against cancer and inflammatory or degenerative diseases: Three parallel stories with a common molecular mechanism involving the nucleoshuttling of the ATM protein? Cancers. 2023;15(5):1482. [CrossRef] [PubMed] [Google Scholar]
  8. Herrera FG, Ronet C, Ochoa de Olza M, Barras D, Crespo I, et al. Low-dose radiotherapy reverses tumor immune desertification and resistance to immunotherapy. Cancer Discov. 2022;12(1):108–133. [CrossRef] [PubMed] [Google Scholar]
  9. Monjazeb AM, Giobbie-Hurder A, Lako A, Thrash EM, Brennick RC, et al. A randomized trial of combined PD-L1 and CTLA-4 inhibition with targeted low-dose or hypofractionated radiation for patients with metastatic colorectal cancer. Clin Cancer Res. 2021;27(9):2470–2480. [CrossRef] [PubMed] [Google Scholar]
  10. Kierkels RG, Korevaar EW, Steenbakkers RJ, Janssen T, van’t Veld A, Langendijk JA, et al. Direct use of multivariable normal tissue complication probability models in treatment plan optimisation for individualised head and neck cancer radiotherapy produces clinically acceptable treatment plans. Radiother Oncol. 2014;112(3):430–436. [CrossRef] [PubMed] [Google Scholar]
  11. Gkikoudi A, Kalospyros SA, Triantopoulou S, Logotheti S, Softa V, Kappas C, et al. Molecular biomarkers for predicting cancer patient radiosensitivity and radiotoxicity in clinical practice. Appl Sci. 2023;13(23):12564. [CrossRef] [Google Scholar]
  12. Genshaft AS, Subudhi S, Keo A, Vasquez JDS, Conceição-Neto N, Mahamed D, et al. Single-cell RNA sequencing of liver fine-needle aspirates captures immune diversity in the blood and liver in chronic hepatitis B patients. Hepatology. 2023;78(5):1525–1541. [CrossRef] [PubMed] [Google Scholar]
  13. Kanagal-Shamanna R, Portier BP, Singh RR, Routbort MJ, Aldape KD, Handal BA, et al. Next-generation sequencing-based multi-gene mutation profiling of solid tumors using fine needle aspiration samples: promises and challenges for routine clinical diagnostics. Mod Pathol. 2014;27(2):314–327. [CrossRef] [PubMed] [Google Scholar]
  14. Price JM, Prabhakaran A, West CML. Predicting tumour radiosensitivity to deliver precision radiotherapy. Nat Rev Clin Oncol. 2023;20(2):83–98. [CrossRef] [PubMed] [Google Scholar]
  15. Landry G, Kurz C, Traverso A. The role of artificial intelligence in radiotherapy clinical practice. BJR Open. 2023;5(1):20230030. [PubMed] [Google Scholar]
  16. Bengs M, Sprenger J, Gerlach S, Neidhardt M, Schlaefer A. Real-time motion analysis with 4D deep learning for ultrasound-guided radiotherapy. IEEE Trans Biomed Eng. 2023;70(9):2690–2699. [CrossRef] [PubMed] [Google Scholar]
  17. Shao H-C, Wang J, Bai T, Chun J, Park JC, Jiang S, et al. Real-time liver tumor localization via a single X-ray projection using deep graph neural network-assisted biomechanical modeling. Phys Med Biol. 2022;67(11):115009. [CrossRef] [Google Scholar]
  18. Zhou D, Nakamura M, Mukumoto N, Yoshimura M, Mizowaki T. Development of a deep learning‐based patient‐specific target contour prediction model for markerless tumor positioning. Med Phys. 2022;49(3):1382–1390. [CrossRef] [PubMed] [Google Scholar]
  19. Muurholm CG, Ravkilde T, De Roover R, Skouboe S, Hansen R, Crijns W, et al. Experimental investigation of dynamic real‐time rotation‐including dose reconstruction during prostate tracking radiotherapy. Med Phys. 2022;49(6):3574–3584. [CrossRef] [PubMed] [Google Scholar]
  20. Qiu Z, Olberg S, den Hertog D, Ajdari A, Bortfeld T, Pursley J. Online adaptive planning methods for intensity-modulated radiotherapy. Phys Med Biol. 2023;68(10):10TR01. [CrossRef] [Google Scholar]
  21. Sun Z, Xia X, Fan J, Zhao J, Zhang K, Wang J, et al. A hybrid optimization strategy for deliverable intensity‐modulated radiotherapy plan generation using deep learning‐based dose prediction. Med Phys. 2022;49(3):1344–1356. [CrossRef] [PubMed] [Google Scholar]
  22. Lee D, Alam S, Jiang J, Cervino L, Hu YC, Zhang P. Seq2Morph: a deep learning deformable image registration algorithm for longitudinal imaging studies and adaptive radiotherapy. Med Phys. 2023;50(2):970–979. [CrossRef] [PubMed] [Google Scholar]
  23. Shirvani SM, Huntzinger CJ, Melcher T, Olcott PD, Voronenko Y, Bartlett-Roberto J, et al. Biology-guided radiotherapy: redefining the role of radiotherapy in metastatic cancer. Br J Radiol. 2021;94(1117):20200873. [CrossRef] [PubMed] [Google Scholar]
  24. Wang S, Mahon R, Weiss E, Jan N, Taylor RJ, McDonagh PR, Quinn B, Yuan L. Automated lung cancer segmentation using a PET and CT dual-modality deep learning neural network. Int J Radiat Oncol Biol Phys 2023;115(2):529–539. [CrossRef] [PubMed] [Google Scholar]
  25. Bruynzeel AM, Tetar SU, Oei SS, Senan S, Haasbeek CJ, Spoelstra FO, et al. A prospective single-arm phase 2 study of stereotactic magnetic resonance guided adaptive radiation therapy for prostate cancer: early toxicity results. Int J Radiat Oncol Biol Phys. 2019;105(5):1086–1094. [CrossRef] [PubMed] [Google Scholar]
  26. Glide-Hurst CK, Lee P, Yock AD, Olsen JR, Cao M, Siddiqui F, et al. Adaptive radiation therapy (ART) strategies and technical considerations: a state of the ART review from NRG oncology. Int J Radiat Oncol Biol Phys. 2021;109(4):1054–1075. [CrossRef] [PubMed] [Google Scholar]
  27. Krishnatry R, Bhatia J, Murthy V, Agarwal J. Survey on adaptive radiotherapy practice. Clin Oncol. 2018;30(12):819. [CrossRef] [Google Scholar]
  28. Brouwer CL, Steenbakkers RJ, Langendijk JA, Sijtsema NM. Identifying patients who may benefit from adaptive radiotherapy: Does the literature on anatomic and dosimetric changes in head and neck organs at risk during radiotherapy provide information to help? Radiother Oncol. 2015;115(3):285–294. [CrossRef] [PubMed] [Google Scholar]
  29. Castadot P, Geets X, Lee JA, Grégoire V. Adaptive functional image-guided IMRT in pharyngo-laryngeal squamous cell carcinoma: is the gain in dose distribution worth the effort? Radiother Oncol. 2011;101(3):343–350. [CrossRef] [PubMed] [Google Scholar]
  30. Enderling H, Alfonso JCL, Moros E, Caudell JJ, Harrison LB. Integrating mathematical modeling into the roadmap for personalized adaptive radiation therapy. Trends Cancer. 2019;5(8):467–474. [CrossRef] [PubMed] [Google Scholar]
  31. Matuszak MM, Kashani R, Green M, Lee C, Cao Y, Owen D, et al. Functional adaptation in radiation therapy. Semin Radiat Oncol. 2019;29(3):236–244. [CrossRef] [PubMed] [Google Scholar]
  32. Akhoundova D, Rubin MA. Clinical application of advanced multi-omics tumor profiling: Shaping precision oncology of the future. Cancer Cell. 2022;40(9):920–938. [CrossRef] [PubMed] [Google Scholar]

Cite this article as: Song T & Li H. Artificial intelligence plus molecular profiling for personalized radiotherapy: Questions 105–107 in the 150 most important questions in cancer research and clinical oncology series. Visualized Cancer Medicine. 2024; 5, 8. https://doi.org/10.1051/vcm/2024011.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.