ESTRO 2024 Congress report
Artificial intelligence (AI) in radiotherapy was a dominating theme at the ESTRO 2024 conference. The potential for AI to transform how we give radiotherapy was clear, as we heard talks about the roles of prediction models, radiomics, treatment decision-making and patient support tools. The majority of AI-based talks were focused, however, on automatic delineation systems and auto-planning.
In the realm of gynaecological cancer, AI’s role in auto-segmentation and auto-planning of intra-uterine brachytherapy and AI-enhanced adaptive radiotherapy were key topics of discussion.
Linda Rossi from The Netherlands presented work on BiCycle, which is a piece of automated planning software for prostate and cervical brachytherapy that was developed in accordance with EMBRACEII adaptive cervix brachytherapy guidelines. This software underwent internal validation at Dr Rossi’s centre, Erasmus University Rotterdam, and external validation at TATA Memorial Hospital in India. In a prospective study, autoplans that were generated by BiCycle were compared with manually generated plans, all of which had been adjusted by treating physicians. The physicians scored each plan on four domains: overall quality, clinical target volume (CTV), its handling of organs-at-risk (OARs), and its loading pattern. Each autoplan consistently outperformed the corresponding manual plan in these evaluations. BiCycle’s autoplans achieved similar target doses with significantly lower doses to OARs compared with the manual plans. The planning time was significantly reduced from 45 minutes to 9.4 minutes. Based on these results, BiCycle has been in clinical use to generate autoplans for brachytherapy since August 2023 at Dr Rossi’s centre.
Matteo Maspero, also from The Netherlands, delivered a lecture on AI for auto-contouring in brachytherapy, highlighting work by Kraus et al.1 In this study, the accuracy and efficiency of auto-contouring OARs for CT-based brachytherapy planning in gynaecological cancers was evaluated. Using 50 training cases, the tool auto-contoured five OARs with high accuracy with mean scores of 3.92 or above (on a Likert scale in which a score of three out of five meant ‘minor edits are necessary’ and a score of four meant ‘minor edits are not necessary’) and a high mean dice similarity coefficient (>0.86). The auto-contoured and edited contours showed no significant differences in dose-volume histogram metrics, and the planning time was reduced from 144.9 minutes to 117.0 minutes with auto-contouring. This underscores the potential of AI in the improvement of planning efficiency with a relatively small training dataset.
AI has also shown promise in the enhancement of daily adaptive radiotherapy. A poster presented by Nuengsigkapian et al. from Thailand reported the use of EthosTM Therapy for daily adaptive radiotherapy with daily CTV auto-contours. This approach resulted in a low incidence of acute grade 2 gastrointestinal toxicity (8.3%) and a slight average reduction in planned PTV to adapted PTV of 2.54% +/- 0.02% (45Gy in 25 fractions of treatment).
In summary, AI-enhanced radiotherapy is proving increasingly valuable, with intrauterine brachytherapy moving towards a fully automated pathway from auto-contouring to auto-planning. Given the limited availability of commercially available auto-contouring tools and the effectiveness that has been demonstrated with just 50 training cases, further research and development in this area could significantly advance patient care.
Reference:
- Kraus AC, Iqbal Z, Cardan RA, Popple RA, Stanley DN, Shen S, Pogue JA, Wu X, Lee K, Marcrom S, Cardenas CE. Prospective Evaluation of Automated Contouring for CT-Based Brachytherapy for Gynecologic Malignancies. Adv Radiat Oncol. 2023 Dec 10;9(4):101417. doi: 10.1016/j.adro.2023.101417. PMID: 38435965; PMCID: PMC10906166.
Dr Sooha Kim
The Institute of Cancer Research/ The Royal Marsden Hospital
London, UK
Sooha.kim@rmh.nhs.uk
@sooha88
Dr Katherine Mackay
The Royal Marsden Hospital/ The Institute of Cancer Research
London, UK
Katherine.mackay@rmh.nhs.uk
@mackaykatherine