Development of a Personalized Voxel-wise Prediction of Brain Metastases using Multi-Parametric MR Imaging to Reduce Treatment Toxicity

Abstract: Brain metastases (BM) present a significant therapeutic dilemma, particularly in patients diagnosed with non- small cell lung cancer (NSCLC) adenocarcinoma who are at high risk of development of BM but where advances in systemic and immunotherapies have significantly improved the 3-year survival rates with a resultant evolution in therapeutic goals. For those with ≤4 BM, focal stereotactic radiosurgery (SRS) is standard but those with greater burden of BM are indicated for whole-brain RT (WBRT). SRS has the advantage of minimizing dose to macroscopically BM-uninvolved regions of the brain. However, the latent effectiveness of SRS remains unclear as it may miss micro-metastatic disease undetectable by current MR imaging, evident in post-SRS progression in untreated areas. WBRT achieves comprehensive BM control but may represent over-treatment of healthy brain tissue, resulting in 65-80% cognitive failure rate. A key observation is the uneven BM risks across brain regions, particularly at the white-grey matter junctions, thought to be in high perfusion areas near the blood- brain-barrier. Recognizing this, our research's main goal is to devise and validate a tailored BM radiotherapy that addresses at-risk areas while minimizing dose to unaffected regions of low risk. The initial goal seeks to integrate four unique components: population-based, anatomical, vascular, and patient BM history, to formulate a time-based patient-specific voxel-wise risk model. Drawing from a multi-centric database of 1,139 NSCLC adenocarcinoma patients and 4,627 lesions, this model predicts areas with a higher propensity for BM. The subsequent goal is to assess the potential dosimetric implications of these BM risk maps on individualized WBRT. We anticipate that incorporating our BM risk predictions will achieve dose reductions ranging from 20- 100% for critical functional sub-structures with no more than 5% global risk of BM misses, a promising step in addressing the pressing issue of WBRT-associated neurotoxicity. Through radiation simulations on retrospective patients, our innovative "WBRT-PROTECT" (WBRT: Personalized Radiation Optimization To Eliminate Collateral Toxicity) method will be juxtaposed against standard treatments to elucidate potential dose optimization. The final goal focuses on clinical validation. Preliminary data suggests that our BM risk maps could inform significant mitigation of brain radiation doses compared to conventional methods, through a modified WBRT approach, enhancing neurocognitive and survival outcomes. To this end, our dual-pronged approach comprises an in silico non-intervention trial across three institutions, matched with a single-center phase I WBRT-PROTECT clinical trial. Notably, the design and endpoints of this clinical trial follow the neurocognitive evaluations and intricate analyses exemplified in NRG CC001, the recently published landmark trial supporting de-intensification of WBRT. Through these evaluations, our research not only pioneers a revolutionary treatment strategy for patients with BM but also deepens our comprehension of the intricate mechanisms behind BM, paving the way for future innovations in BM imaging, detection, segmentation, and therapeutic strategies.