An Artificial Intelligence-Based Full-Process Solution for Radiotherapy: A Proof of Concept Study on Rectal Cancer

Abstract

Background and PurposeTo develop an artificial intelligence-based full-process solution for rectal cancer radiotherapy.Materials and MethodsA full-process solution that integrates autosegmentation and automatic treatment planning was developed under a single deep-learning framework. A convolutional neural network (CNN) was used to generate segmentations of the target and the organs at risk (OAR) as well as dose distribution. A script in Pinnacle that simulates the treatment planning process was used to execute plan optimization. A total of 172 rectal cancer patients were used for model training, and 18 patients were used for model validation. Another 40 rectal cancer patients were used for an end-to-end evaluation for both autosegmentation and treatment planning. The PTV and OAR segmentation was compared with manual segmentation. The planning results was evaluated by both objective and subjective assessment.ResultsThe total time for full-process planning without contour modification was 7 min, and an additional 15 min may require for contour modification and re-optimization. The PTV DICE similarity coefficient was greater than 0.85 for all 40 patients in the evaluation dataset while the DICE indices of the OARs also indicated good performance. There were no significant differences between the auto plans and manual plans. The physician accepted 80% of the auto plans without any further operation.ConclusionWe developed a deep learning-based automatic solution for rectal cancer treatment that can improve the efficiency of treatment planning.

Clinical evaluation of autonomous, unsupervised planning integrated in MR-guided radiotherapy for prostate cancer

“…Recently Xia et al [20] presented a proof-of-concept study for a similar AI-based planning approach for rectal cancer, where 80% of the automatically contoured and planned cases would have been accepted without manual fine-tuning for clinical treatment. Nevertheless, as demonstrated recently by McIntosh et al [15] acceptance of AI-based plans during a retrospective testing phase and a prospective phase may be different, where plans are intended to be used for clinical real-life treatments.…”

“…The full potential of automation in the RT treatment planning process would be exploited if the complete planning and data preparation workflow was automatized, including organ segmentation, CTV/PTV definition and treatment planning. Xia et al recently proposed full automation of the whole treatment preparation chain for rectal cancer [20]. We demonstrated clinical and technical feasibility of autonomous, un-supervised treatment plan-ning for MR-guided RT (MRgRT) of prostate cancer in a recent firstin-human application [21].…”

The dosimetric impact of deep learning-based auto-segmentation of organs at risk on nasopharyngeal and rectal cancer

“…This finding might have been caused by the algorithm difference between the two systems. Our system used a 2D U-Net network, which could have some outliers, as our previous study demonstrated [ 21 , 22 ]. UIH used a two-phase algorithm, which was more robust according to region location.…”

First experience of autonomous, un-supervised treatment planning integrated in adaptive MR-guided radiotherapy and delivered to a patient with prostate cancer

“…Parameters which only fulfill tolerance are highlighted (criteria for penile bulb and CTVs are only used for evaluation during online adaption, but not during PSO). practice, but a similar workflow was recently proposed for rectal cancer [17] and automation of single processes was also introduced earlier [7][8][9][13][14][15][16].…”

“…Even though automation of certain steps in the data preparation and treatment planning process have been proposed earlier, currently each workflow step still requires human interaction for validation and further processing. Recently, a full automation of the entire process with integration of data handling and RT planning without any human interaction has been presented for rectal cancer [17]. Such autonomous, unsupervised data preparation and RT planning approach including OAR segmentation, target volume definition and plan optimization may have the potential to harmonize treatments and reduce the time from simulation to RT start.…”