ESTRO 2024 Congress report

At ESTRO 2024, as in previous years, artificial intelligence (AI) was highly prevalent. A wide range of innovative and exciting research was showcased, which ranged from applications such as automatic segmentation to discussions during the closing debate regarding the proposition that AI would take over the radiation therapy care pathway by 2040. In the brachytherapy track, several advances in the realm of AI were highlighted.

Many industry partners showcased how their products improved automatic segmentation. Regarding brachytherapy, contributions from AI were evident in liver tumour brachytherapy (Haddad, Hermani, & Pinkawa, 2024). There, the process was explained by which organs-at-risk (OAR) contouring was automated through the use of AI. These researchers’ work showed substantially reduced contouring time compared with that required by radiation oncologists to produce contours manually.

The BiCycle method was another interesting AI development that was discussed. BiCycle has recently been extended to use in cervical cancer brachytherapy. This automatic treatment planning system has demonstrated significantly reduced planning times (Rossi, et al., 2024). It is noteworthy that the automatic plans, after manual adjustments, were highly favoured over full manual planning.

AI's impact on cervical cancer brachytherapy was also highlighted through the clinical evaluation of auto-segmentation of OARs (Keek, et al., 2024). Their work showed that AI auto-segmentation performed well in segmentation that was conducted according to the EMBRACE II guidelines. However, a general warning was issued: overreliance on automatic segmentation could result in a slight human bias towards AI-generated segmentations. Overall their findings reported that all results were within clinically accepted variability whilst the time required to complete delineation was cut in half.

The research programme on brachytherapy via artificially intelligent GOMEA heuristic-based treatment planning (BRIGHT) stood out as a prime example of AI innovation. This treatment planning method was mentioned by Dr Bradley Pieters during his lecture after he had received the Iridium 192 award. Originally developed for prostate high-dose-rate brachytherapy, BRIGHT has been extended to enable the sparing of specific organs. This use could benefit patients with comorbidities or in cases of suboptimal implant geometry. Intuitive categorisation of treatment plans that spare specific OARs was shown (Scholman, et al., 2024). BRIGHT has also been validated internationally in a centre in Aarhus, Denmark. Plans optimised through the use of BRIGHT significantly reduce doses to OARs whilst achieving similar target coverage compared with clinical planning (Hansen, et al., 2024). Lastly, BRIGHT has been extended to cervical cancer and has been shown to be intuitively customisable with clinical objectives specific to institutional demands (Dickhoff, et al., 2024). These various uses of BRIGHT not only highlight its adaptability but also its potential to streamline treatment planning across various cancer types.

Ir. Renzo J. Scholman

PhD candidate at Centrum Wiskunde & Informatica and Delft University of Technology

Delft, The Netherlands

renzo.scholman@cwi.nl

Bibliography

L. R. Dickhoff, E. M. Kerkhof, H. H. Deuzeman, L. A. Velema, D. L. J. Barten, B. R. Pieters, C. L. Creutzberg,  P. A. N. Bosman, T. Alderliesten, Tanja. (2024). Intuitively customizable AI-based cervix brachytherapy treatment planning. ESTRO 2024, (pp. 357-361). Glasgow.

H. Haddad, H. Hermani, M. Pinkawa. (2024). Integration of Artificial Intelligence in Interstitial HDR Brachytherapy for Liver Tumors. ESTRO 2024, (pp. 169-170). Glasgow.

A. T. Hansen, A. Bouter, P. A. N. Bosman, T. Alderliesten, S. Buus, J. G. Johansen. (2024). External validation of automated treatment planning for HDR prostate brachytherapy using BRIGHT. ESTRO 2024, (pp. 341-344). Glasgow.

S. A. Keek, A. Mans, M. E. Nowee, E. E. Rijkmans, E. C. Schaake, R. Simões, T. M. Janssen. (2024). Clinical evaluation of organs at risk deep learning auto-segmentation for cervix brachytherapy. ESTRO 2024, (pp. 334-337). Glasgow.

L. Rossi, R. Bijman, H. Westerveld, M. Christianen, L. Luthart, M. Huge, I. Kolkman-Deurloo, J. W. Mens, H. Abusaris, R. De Boer, S. Breedveld, B. Heijmen, R. Nout, (2024). Clinician preferred, fast autoplanning in cervical cancer brachytherapy using BiCycle. ESTRO 2024, (pp. 215-217). Glasgow.

R. J. Scholman, D. L. J.  Barten, B. R. Pieters, P. A. N. Bosman, T. Alderliesten. (2024). Finding the best options to spare organs at risk with AI-based prostate HDR brachytherapy planning. ESTRO 2024, (pp. 344-347). Glasgow.