What we do
About our project
Brachytherapy is a crucial curative treatment for locally advanced cervical and vaginal cancer. While highly effective, it is invasive, typically requiring 3–4 sessions of nearly 9 hours each, causing significant physical and psychological strain, and even sometimes leading to post-traumatic stress. Long preparation times increase the risk of changes in implant position and patient anatomy, affecting treatment precision. Much of this time is spent on tumor and organ contouring, implant reconstruction, and dose optimization by experienced staff.
Treatment plan preparation—including imaging, contouring, implant reconstruction, and dose optimization—can take 3–5 hours manually, during which anatomy may change, reducing treatment plan accuracy. Automating these steps with AI could reduce preparation time, improve accuracy, enhance patient comfort, decrease staff workload, and increase clinical capacity.
Our aim is to develop a push-button AI-driven workflow for fast plan preparation, achieving a 75% reduction in preparation time while maintaining physician evaluation and correction, when desired. We will develop, test, and integrate three AI-based solutions for cervical and vaginal brachytherapy.
Treatment plan preparation—including imaging, contouring, implant reconstruction, and dose optimization—can take 3–5 hours manually, during which anatomy may change, reducing treatment plan accuracy. Automating these steps with AI could reduce preparation time, improve accuracy, enhance patient comfort, decrease staff workload, and increase clinical capacity.
Our aim is to develop a push-button AI-driven workflow for fast plan preparation, achieving a 75% reduction in preparation time while maintaining physician evaluation and correction, when desired. We will develop, test, and integrate three AI-based solutions for cervical and vaginal brachytherapy.
Our research focus
To achieve our aims, we will develop, test, evaluate, and integrate three AI-based solutions for cervical and vaginal brachytherapy:
Automated Contouring of Organs at Risk
We will create and validate a deep learning system for automatic contouring of organs at risk near the tumor on both MRI and CT scans, with the implant (applicator) in place. This system will work for both cervical and vaginal treatments, delivering contours in seconds. Validation will include measuring any necessary physician editing time, and we will explore transferring patient-specific anatomical knowledge from previous fractions to improve contouring accuracy.Automated Implant Reconstruction
We will develop AI-based reconstruction for brachytherapy applicators and interstitial/free needles based on imaging. Existing applicator reconstruction systems for cervical cancer will guide the development of similar tools for vaginal treatments. The same datasets will also be used for automated needle reconstruction for both cervical and vaginal cases.Automated Treatment Planning (Dose Optimization)
Fast, AI-driven treatment planning will be developed in two stages. First, our existing rule-based BiCycle system, proven for cervical cancer, will be adapted for vaginal tumors, ensuring superior plan quality compared to manual planning. Second, BiCycle-generated high-quality plans will train deep learning models for rapid automated planning in both cervical and vaginal brachytherapy, following approaches successfully applied in external beam radiotherapy.Funds & Grants
Granted by Hanarth Foundation; a foundation that aims to promote and enhance the use of artificial intelligence to improve cancer treatments.
Our team
Miranda Christianen
Lorne Luthart
