Precision Radiotherapy for Refractory Brain Metastases: A Multicenter Study

Conventional image-guided techniques (such as cone-beam CT) have poor soft tissue contrast and unsatisfactory treatment outcomes. MRgART offers high-resolution imaging of brain tissue and, by acquiring daily MR images, successfully achieves real-time image monitoring, displaying tumor position and volume changes during treatment. This may improve local control rates for brain metastases, reduce toxicities, and translate into survival benefits. Our research team has already conducted a phase II prospective study and achieved favorable results, with a 1-year local control rate of 100% for intracranial brain metastases, significantly superior to historical controls of conventional radiotherapy, no severe late toxicities, and real-time individualized precision treatment. Based on the phase II study results, this phase III single-arm prospective study was designed. This study aims to conduct a multicenter prospective study to establish an MR-guided adaptive radiotherapy (MRgART) technical platform, enabling real-time monitoring of tumor position and volume changes and individualized plan adaptation. Through technological innovation, we aim to significantly improve local control rates for large and complexly located brain metastases while reducing severe adverse effects such as radiation brain necrosis, thereby laying the technical foundation for precise and safe treatment. This study will establish a new technical system for adaptive radiotherapy (ART) for brain metastases based on per-fraction MR images acquired during radiotherapy, using adaptive radiotherapy to improve target dose coverage and reduce radiation doses to organs at risk such as normal brain tissue. It is anticipated to validate the phase II findings of high local control, low toxicity, and significantly prolonged survival. Based on the obtained results, combined with multicenter clinical practice and application experience, domestic and international expert symposiums and special sessions will be conducted, and a multicenter consensus on MRI-linac guided adaptive radiotherapy for brain metastases from lung cancer will be developed.

Furthermore, this study recognizes that MRI generates high noise levels and that without appropriate hearing protection, repeated treatments may cause hearing damage to patients. Therefore, the study will simultaneously collect hearing-related scales from patients before, during, and after treatment (through pure-tone audiometry tests performed in the otolaryngology department or self-administered pure-tone screening via a mini-program) to explore the impact of performing or not performing strict hearing protection on changes in patients' hearing scales, as well as its effect on treatment interruption or repeated setup.

Compared to conventional cone-beam CT-guided radiotherapy, MRI-guided adaptive radiotherapy still has some shortcomings. MR image acquisition is slow, and after each fraction's images are acquired, they require manual registration with CT images, followed by manual re-contouring and manual plan calculation, and finally, execution requires triple review by physicians, physicists, and therapists. Currently, each patient requires approximately 40 minutes to 1 hour from lying on the treatment couch to the end of treatment. If there are setup errors or significant target volume changes requiring adaptive re-contouring, even more time may be needed to complete the treatment, posing significant challenges to patients' physical strength and mental state, and limiting its potential application value in patients with altered consciousness, the elderly and frail, or those unable to cooperate. Additionally, it consumes substantial medical resources. To address this situation, our research group plans to develop an artificial intelligence-assisted system for target contouring, treatment planning, and treatment decision-making. Based on our center's previously published results, we aim to establish a full-process AI-assisted model for MRgART in the treatment of lung cancer brain metastases, and compare its efficacy and treatment time differences with manual-only radiotherapy delivery. On this basis, we will explore the feasibility of exempting a small subset of patients (approximately 10-20 patients) in the phase III study from conventional CT and MR simulation, contouring targets and generating plans on simulation CT and MRI images, and then directly using ATS (adaptive target shaping based on interfractional target deformation) to generate radiotherapy plans within the MRgART workflow and proceed with subsequent treatment.