Purpose : Modern management of brain metastases requires close surveillance with serial magnetic resonance imaging (MRI), creating interest in techniques for automated detection. The convolutional neural networks (CNN) is currently the dominant approach to automated metastasis detection, and multiple variations of the CNN have been investigated to improve performace. In this work, we looked beyond the impact of network architecture and assessed the impact on performance of providing anatomical contextual information to a CNN in the training period. Materials and Methods : The nn-Unet, a widely-adopted CNN, was selected for this study. The nn-Unet was trained on an institutional dataset consisting of 301 high-resolution, T1-weighted, contrast-enhanced MRIs and associated structure labels consisting of both target brain metastases (1,111 total) and several organs at risk (OARs) that were labeled in the process of radiosurgery planning. In the training period, the model was presented either with labeled metastases alone (BM) or with the complete structure set of metastases and OARs (BM+OAR). The test set contained 100 cases and 421 total metastases. Results : We found that by multiple standard performance metrics, the BM+OAR model outperformed the BM model. Median case-wise Dice Coefficient for BM was 0.75 (IQR 0.51-0.86) and for BM+OAR was 0.79 (IQR 0.58-0.86) (p<0.001). Median sensitivity for BM was 1.0 (IQR 0.71-1.0) and for BM+OAR was 1.0 (IQR 0.75-1.0) (p<0.01). Median positive predictive value for BM was 0.5 (IQR 0.25-0.74) and for BM+OAR was 0.77 (IQR 0.50-1.0) (p<0.001). Importantly, amongst lesions scored as “false-positives”, 47/151 detected by BM+OAR could be retrospectively verified to be real tumors while only 22/231 detected by BM could be verified to be real (p<0.001). Conclusions : These findings suggest that providing anatomical contextual information can improve accuracy of CNN-based automated brain metastasis detection algorithms and opens new opportunites for research into optimization of automated detection approaches for brain metastases.

Abstract Craniospinal irradiation (CSI) is an effective treatment for hematologic leptomeningeal disease (LMD). Comparative outcome data for proton CSI (“pCSI”) vs volumetric modulated arc therapy (VMAT, “xCSI”) are not well described in hematologic malignancies, a vulnerable population with often-limited bone marrow reserve. We compared outcomes between xCSI and pCSI, hypothesizing higher toxicity with xCSI but similar survival. Hematologic patients who received CSI between 2013–2025 were identified. Toxicities were graded with CTCAE v.5.0 at baseline, during treatment, and 1- and 3-months post-CSI. Overall survival (OS) from CSI start was analyzed with Cox proportional hazards models adjusted for ECOG. Progression-free survival (PFS) and CNS-PFS were also evaluated. 34 patients received CSI: 26 (76%) xCSI, 8 (24%) pCSI. Dose ranged from 7.2 Gy-30.6 Gy in 4-17 fractions. Baseline characteristics were similar: ALL (59%) and NHL (18%) were most common (p=.51), and most ECOG 1 (p=.80). Non-hematologic toxicities were largely grade 1-2 (fatigue, nausea, pain) with no differences between modalities. Grade ≥3 lymphopenia was more common during xCSI vs pCSI (87.5% vs 50.0%; p=.047) but not at 1 month (36.4% vs 66.7%; p=.35). Median OS was 28.3 months with no difference between xCSI vs pCSI (24.1m vs NR; p=.23), including after ECOG adjustment (HR 0.16; 95% CI 0.02–1.45; p=.10). PFS (8.6m vs 27.9m; p=.30) and CNS-PFS (18.6m vs NR; p=.20) trended higher after pCSI without significant differences. CNS-PFS differed by treatment indication (bridging vs definitive vs consolidation; 18.6m vs 8.6m vs NR, respectively; p=.02). pCSI and xCSI demonstrated comparable toxicity in hematologic patients, supporting the safety of VMAT xCSI. However, higher rates of severe lymphopenia during xCSI warrant attention and may support pCSI in high-risk patients. Outcomes also differed by treatment indication, with poorer CNS control in patients treated definitively. This is an important step toward optimizing neuraxial radiotherapy in hematologic LMD.

Abstract Background Leptomeningeal metastasis (LM) is a severe complication of advanced metastatic cancer, with poor prognosis and limited therapeutic options. The optimal treatment strategy is unclear. This study evaluated overall survival (OS) associated with different combinations of treatment modalities in patients with LM. Methods We conducted a retrospective cohort study of 94 patients with LM treated at our tertiary cancer center (breast n = 57, lung n = 25, other n = 12). OS was defined as time from LM diagnosis to death. Survival was estimated using Kaplan–Meier methods. Cox proportional hazards models evaluated associations between treatment exposures and OS, adjusting for primary cancer type, age, sex, and baseline Karnofsky Performance Status (KPS). Interaction between intrathecal (IT) therapy and KPS was assessed. Results The median OS was 9.3 months. 79% received systemic therapy, 68% radiotherapy (13% craniospinal), and 84% IT therapy. IT therapy and systemic therapy were associated with improved survival in unadjusted analysis (p < 0.001). CSI demonstrated a trend towards improved OS (p = 0.068), whereas other RT did not. Patients receiving IT as part of combination therapy with other modalities had improved OS compared with those that did not (p = 0.012). In multivariable analysis adjusting for baseline characteristics, IT therapy remained strongly associated with improved survival (HR 0.26, 95% CI 0.13–0.51, p < 0.001). A significant interaction between IT therapy and KPS was observed (p = 0.019), with a greater relative benefit among patients with KPS < 80. Higher KPS was associated with improved survival. Conclusions IT therapy was strongly associated with improved survival, especially as part of combined modality regimen, with the greatest relative benefit observed among patients with poorer functional status. These findings also suggest that IT therapy may be underutilized in patients with lower KPS.

We present the University of California San Diego post-treatment glioblastoma (UCSD-PTGBM) annotated multimodal MRI dataset. The UCSD-PTGBM dataset includes 243 timepoints on 178 subjects with histopathologically-proven glioblastoma who were imaged with an advanced brain tumor protocol on 3 Tesla MRI scanners. Sequences include standard 3D imaging, as well as multishell diffusion (Restricted Spectrum Imaging, RSI) and perfusion imaging techniques (Arterial Spin Labelling, ASL and Dynamic Susceptibility Contrast, DSC), and neuroradiologist approved voxelwise tumor segmentations for both traditional segmentation masks and cellular tumor segmentations. The dataset also includes isocitrate dehydrogenase (IDH) mutation status and O6-methylguianine-DNA methyl-transferase (MGMT) promotor methylation status, as well as overall survival and progression free survival information for a subset of cases. We hope that researchers around the world will use these data to continue to improve analysis of post-operative MRI on glioblastoma patients, translate their findings into clinical practice and improve the management and outcome for these patients.

BACKGROUND:Proton beam radiotherapy (PBRT) is a mainstay in treatment of pediatric brain tumors and may reduce risk of neurotoxicity. We hypothesized that dose-dependent effects on resting-state networks (RSNs) would manifest as impaired neurocognition and quality of life (QoL) in pediatric patients undergoing PBRT. METHODS:Patients and age- and sex-matched healthy controls (HCs) were prospectively enrolled and underwent resting-state functional connectivity (FC) MRI and neurocognitive tests (NIH Toolbox, PROMIS QoL Questionnaire) at 3-5 months post-PBRT. Functional connectivity within and between RSNs (default mode network [DMN], dorsal attention network [DAN], executive central network [ECN], salience network [SN]) was compared. One-sample t-tests assessed average RSN-specific FC. Group-level analysis used paired t-tests (q = 0.05). RESULTS:Twenty-eight participants (14 patients, 14 HCs; median age 14 years) were analyzed. Compared to HCs, patients displayed worse attention/executive function (P = .018), working memory (P = .011), and processing speed (P = .047). Patients had lower within-network FC in DMN (P = .013) and SN (P = .026). Patients had disrupted anticorrelation between DMN and SN, and less local efficiency of RSNs (corrected P < .05). Higher brain volume receiving ≥20 Gy, left hippocampus mean dose, and left temporal lobe mean dose were associated with RSN disruption in DMN and SN (|r|≥0.70). Lower SN FC was associated with poorer processing speed (P = .049). On multivariable analysis, ECN FC was associated with working memory (P = .016) and patients had worse working memory vs HCs (P = .009). CONCLUSIONS:Compared to HCs, pediatric patients after PBRT displayed significant disruption in several RSNs, notably between SN and DMN critical in neurocognitive control. Dose-dependent neurocognition and QoL impairment warrant further investigation for mitigation.

PURPOSE:To introduce a novel cognition-sparing stereotactic radiosurgery (SRS) planning framework that limits dose to the cognitive connectome (10 bilateral white matter tracts, the corpus callosum, and the hippocampi) and to demonstrate its capability to preferentially spare the cognitive domains of memory, language, executive function, attention, and fine motor control. METHODS AND MATERIALS:Three cases representing the spectrum of intracranial disease treated with SRS (intact metastases, postoperative cavity, and arteriovenous malformation) and presenting competing domain-specific priorities based on lesion location were selected from the phase 2 cognition-sparing SRS (COG-SRS) trial. The cognitive connectome was autosegmented using AtlasTrack (white matter tracts) and FreeSurfer/FastSurfer (hippocampi) and registered to the computed tomography simulation scan. For each case, a standard plan, 4-6 domain-optimized plans, and an all-domain composite plan were generated using noncoplanar volumetric modulated arc therapy. All plans were required to meet target coverage and conventional organ-at-risk constraints. Dose to domain-specific structures and plan quality metrics were compared to quantify achievable dose sparing and dosimetric trade-offs. RESULTS:Domain-optimized plans consistently reduced dose to structures supporting the prioritized domain, with expected cross-domain trade-offs. The all-domain composite plan produced the most balanced reductions across eloquent structures. When lesions abutted cognition-critical structures, Dmax reductions were limited. However, substantial Dmean reductions were still achieved. All plans maintained standard target coverage, with generally small differences in plan quality indices relative to standard planning. Plan complexity varied across domain-specific optimizations, reflected by differences in monitor units. CONCLUSIONS:Cognition-sparing SRS that limits dose to the widely distributed cognitive connectome while maintaining target coverage is feasible and allows for a domain-tunable approach applicable across common SRS indications. This flexibility supports a patient-centered paradigm, in which cognitive priorities are incorporated through shared decision-making. Using standard magnetic resonance imaging sequences and automated segmentation, the technique is readily translatable, and an open-access website (www.cogsrs.com) has been created to facilitate dissemination.

Intracranial metastases (ICM), specifically parenchymal brain metastases, remain a major clinical challenge in solid tumor oncology, despite recent advances in cancer therapies which have led to improvements in survival for these patients. Improving outcomes even further in this patient population will require a multi-disciplinary approach, including pre-clinical and translational studies, clinical trials, and studies of patient reported outcomes and quality of life. At the 2023 and 2024 joint Society for Neuro-Oncology (SNO) and American Society of Clinical Oncology (ASCO) CNS Metastases Conferences, two ICM collaborative group think tanks convened, composed of diverse, multi-disciplinary stakeholders, including basic and translational researchers, clinical trialists, and clinicians from academia and the community setting. Here we summarize the key knowledge gaps and consensus recommendations put forth by these two think tanks. Advances in ICM research and improvements in patient outcomes will require close inter-specialty and inter-institutional collaboration between stakeholders, including pre-clinical and translational researchers, clinical investigators, industry, and regulatory bodies.

Leptomeningeal Metastases (LM) is a devastating complication of advanced cancer with historically poor outcomes. Median overall survival (OS) has typically ranged from 2–4 months, although recent advances in craniospinal irradiation (CSI) and newer systemic chemotherapies are improving outcomes. We evaluated the role of intrathecal chemotherapy (IT) on overall survival, for the treatment of LM, as part of multimodal approaches with systemic chemotherapy and radiation. We retrospectively evaluated treatment patterns and survival in 60 patients with metastatic carcinoma who received IT for LM, from 2016-2024. All patients underwent Ommaya reservoir placement and received IT. The median age at LM diagnosis was 53.5 years. 50/60 were female. Histologies were as follows: Breast cancer (n=36), Lung cancer (n=18), Other (n=6). 40/60 also received involved-field RT to brain or spine, 7/60 received CSI. 46/60 had a change in their systemic chemotherapy regimen. The median OS for all patients was 13 months (breast 12 months, lung 10.5 months, other 5 months). Kaplan-Meier analysis demonstrated that patients who received CSI at any recurrence (n=8) had a trend for better survival, with a hazard ratio (HR) of 0.21 (95% CI: 0.03–1.55, p=0.127, median OS not reached). Patients who had a change in systemic chemotherapy had a median OS of 10.5 months, vs 8 months when prior systemic treatment was continued. Additionally, we evaluated patients who did not receive antibody-drug conjugates after LM diagnosis or CSI, which can improve OS, and the median OS of this group was still 10.5 months (n=38). Initial Karnofsky Performance Status (KPS) was borderline significant in predicting improved OS, (HR = 0.98, 95% CI: 0.96–1.00, p=0.059). These findings suggest that the inclusion of IT chemotherapy in multimodal treatment strategies may meaningfully improve outcomes in LM, especially when accompanied by CSI and changes in systemic chemotherapy.

Craniospinal irradiation (CSI) is an effective treatment for solid tumor leptomeningeal carcinomatosis (LMC). Although proton CSI (pCSI) improves survival vs involved-field photon therapy, differences in outcomes when treating the entire neuraxis with volumetric modulated arc therapy (VMAT, or “xCSI”) remain unclear. We compared toxicity and outcomes for solid tumor LMC treated with pCSI vs xCSI, hypothesizing greater hematologic toxicity with xCSI but similar performance status-adjusted survival. Adults diagnosed with LMC and planned for CSI to 30Gy/10fx from 2020–2025 were included. Acute and late (1, 3m post-CSI) hematologic toxicities were graded per CTCAE v5.0. Systemic and intrathecal (IT) therapy pre- and post-CSI were recorded. CNS progression-free (PFS) and overall (OS) survivals were calculated from CSI start with Cox proportional hazards models adjusting for ECOG status. Of 36 patients, 33 completed CSI: 20 (61%) pCSI, 13 (39%) xCSI. Median age was 53yo, most were ECOG 0-1 (p=0.4), and most histologies were breast (42%) or lung (36%; p=0.3). After CSI, systemic therapy use was higher in pCSI (92% vs 46%, p=0.03), while IT use was similar. Acute anemia (G1+) was more common with xCSI (61% vs 20%, p=0.04), trending towards higher acute lymphopenia (88% vs 67% G3+, p=0.09). At 1m, leukopenia (G1+) trended higher with xCSI (46% vs 10%, p=0.05). At 3m, toxicity differences resolved. Median OS was 10.7m (95% CI 4.6-NR: 10.9m pCSI, 4.6m xCSI; p=0.03). Median CNS PFS was 5.5m (95% CI 1-NR: 8.5m pCSI, 3.0 xCSI; p=0.02). There was no significant difference in ECOG-adjusted OS (HRxCSI 1.3 [95% CI 0.9-14.9]; p=0.07) or CNS PFS (HRxCSI 1.4 [95% CI 1.0-15.4]; p=0.05). This is the first comparison of pCSI vs xCSI for solid tumor LMC, suggesting VMAT xCSI safety and that survival differences reflect patient selection. Given its wider availability, further work is needed to clarify VMAT’s role in LMC management.

BACKGROUND:Limbic white matter (WM) of the Papez circuit, including the fornix, dorsal cingulum, and parahippocampal cingulum (PHC), interplay with the hippocampus as key components of the memory network. We analyzed biomarkers of injury to these pathways to understand their impact on post-radiation therapy (RT) memory performance. METHODS:Primary brain tumor patients on a prospective trial receiving fractionated brain RT (n = 57) underwent volumetric MRI, diffusion tensor imaging, and memory assessments (Hopkins Verbal Learning Test-Revised [HVLT-R] Total and Delayed Recall and Brief Visuospatial Memory Test -Revised [BVMT-R] Total and Delayed Recall) at baseline and 3, 6, and 12 months post-RT. MRI biomarkers included volume, fractional anisotropy (FA), and mean diffusivity (MD). Linear mixed-effects models assessed associations between biomarkers and memory performance over time. RESULTS:Smaller volumes in the right fornix were associated with lower BVMT-R-Total scores (P = 0.019) and left PHC volume loss was associated with worse performance on BVMT-R-Delayed (P = 0.039). Lower FA in the left (P = 0.010) and right (P = 0.019) fornix was associated with lower BVMT-R-Total performance. Lower FA in the left dorsal cingulum (P = 0.038) and right PHC (P = 0.039) were associated with lower HVLT-R-Total and HVLT-R-Delayed scores, respectively. Higher MD in bilateral fornix (P = 0.01) and right PHC (P = 0.011) correlated with lower BVMT-R-Total scores; higher MD in the right PHC (P = 0.046) also correlated with lower HVLT-R-Total scores. Hippocampal volume was not associated with memory scores. CONCLUSION:Poorer microstructural integrity in limbic WM tracts of the Papez circuit predicted worse memory performance, while hippocampal injury did not. Dose avoidance in these tracts may preserve memory outcomes.

Radiotherapy (RT) is an established consolidation strategy for primary CNS lymphoma (PCNSL). However, the role of RT in secondary CNSL (SCNSL) is unclear. Here, we describe RT doses, disease failure patterns, and survival outcomes in an RT-treated CNSL cohort. Patients treated with CNS-directed RT for CNSL at UCSD were identified by ICD-10 codes, RT course description, or radiation oncologist. Site of CNS disease failure was local (involving initial site) or distant (outside of initial site). Overall survival (OS) and CNS progression-free survival (CNS PFS) were evaluated from time of RT start using Kaplan-Meier. Outcomes were stratified by PCNSL vs SCNSL. 30 patients (13 [43%] PCNSL, 17 [57%] SCNSL) and 39 total CNS-RT courses were included. Most received RT for relapsed/refractory disease (23, 77%) as opposed to definitive (6, 20%) or consolidative (1, 3%) intent. At the time of RT, 1 (3%) patient had CSF, 13 (43%) had ocular, and 12 (40%) had extracranial disease (6 progressive, 6 stable). Median RT dose was higher for PCNSL (40.5 Gy vs 30 Gy for SCNSL, p=.031). First course RT targets were whole-brain RT (WBRT; 17, 57%; 6 included orbits [3 electively]), orbits alone (10, 33%), focal (2, 7%), and spine (1, 3%). 17 patients relapsed, most only distantly (9 [53%] vs 5 [29%] locally). Only SCNSL patients had synchronous local and distant failures (3, 18%). 2 patients received focal RT to dura and did not progress. Of the 4 (36%) who relapsed after WBRT, 2 were local-only SCNSL failures, and 2 were distant-only PCNSL failures. Of the 4 (67%) who relapsed after WBRT + orbits, 3 failed in the parenchyma. Most of the 8 who relapsed after orbits RT alone had distant (5, 63%) failures, most (3, 60%) parenchymal. Those with ocular involvement received WBRT + orbits or orbital RT alone, most relapsing in the parenchyma (8, 62%). No ocular failures occurred in those without ocular disease at RT start. Median OS was longer for PCNSL vs SCNSL (18.6 months [95%CI 0.0-42.4] vs 6 months [95%CI 3.0-8.9]), respectively; log-rank p=.058). Median CNS PFS was similar between PCNSL vs SCNSL (2.83 months [95%CI 0.0-19.8] vs 4.41 months [95%CI 1.4-7.4], respectively; log-rank p=.256). 3 patients had systemic relapse preceding CNS relapse. Different disease failure patterns were seen in mostly relapsed/refractory SCNSL vs PCNSL receiving CNS-RT. Only SCNSL patients failed both distantly and locally, suggesting the inclusion of more comprehensive treatments alongside RT (e.g., systemic or intrathecal). As SCNSL received a lower RT dose and most patients suffering local failure were SCNSL, further dose optimization should be considered. Future studies will examine MRI predictors of disease failure patterns. Brianna Hostler, Jona A. Hattangadi-Gluth, Parag Sanghvi, Kathryn R. Tringale. Radiation therapy utilization and outcomes in primary and secondary central nervous system lymphoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4675.

e13043 Background: Leptomeningeal carcinomatosis (LMC) is a rare consequence of metastatic breast cancer (MBC) with a dismal prognosis, with a historical 15-week median overall survival (OS). There is no standard of care treatment (tx), and a paucity of data on MBC subtype specific incidence and prognosis. The impact of newer CNS penetrating systemic (sys) tx like antibody drug conjugates (ADCs) and multimodal CNS tx on BC related LMC prognosis is underreported in contemporary literature. Here, we present a real-world analysis of a contemporary cohort of patients (pts) with LMC. Methods: This is a single-institution retrospective study of 48 pts diagnosed with BC related LMC between 2016-2024, based on radiographic features and/or positive cerebrospinal fluid (CSF) cytology or CSF circulating tumor cells (CTC). Pts consented to share information with a research biorepository (IRB 181755). We conducted a retrospective chart review of clinical features, tx, and OS. Results: 25 pts (52%) had invasive ductal carcinoma, 14 (29%) had invasive lobular carcinoma, 4 (8%) had metaplastic carcinoma, 2 (4%) had inflammatory carcinoma, and 3 (6%) had mixed histology. 20 pts (42%) had estrogen receptor positive (ER+)/human epidermal growth factor receptor 2 negative (HER2-) BC, 4 (8%) had HER2+ BC, and 24 (50%) had triple negative BC (TNBC). Median age at LMC diagnosis (dx) was 50 years. Median time from MBC to LMC dx varied significantly by subtype: 24.5 months (mos) in ER+/HER2-, 7 mos in HER2+, and 4.5 mos in TNBC (p = 0.01). Prior tx with ADCs (n = 25) was associated with a significantly longer time from MBC to LMC dx (21 vs. 8 mos, p = 0.037). 32 pts (67%) had confirmed LMC by CSF cytology or CTC. 28 pts (58%) received intrathecal chemotherapy (ITC), 9 (19%) received craniospinal irradiation (CSI), 26 (54%) received whole brain radiation (WBRT), and 18 (38%) received ITC+ RT (CSI or WBRT). Median OS varied by subtype: 7.0 mos for ER+/HER2-, 22.4 mos for HER2+, and 5.8 mos for TNBC. Significantly longer OS was seen with ITC+/- RT vs RT alone (9.1 vs. 1.3 mos, p < 0.0001). 20 pts received ADCs post LMC dx, all of whom also had CNS tx (ITC and/or RT), with a significantly longer OS than pts with CNS tx and no ADC (13 vs. 5 mos, p = 0.002). Eastern Cooperative Group performance status (median = 1) did not vary by ADC use. In pts naïve to ADCs pre-LMC dx, a longer OS trend was seen with ADC tx post LMC dx vs other sys tx (32 vs 7 mos, p = 0.14). In pts with LMC and HER2 low or ultralow MBC (n = 25) who received CNS tx, the addition of trastuzumab deruxtecan (n = 6) significantly improved OS (23 vs 5.5 mos, p = 0.02) compared to CNS tx +/- other sys tx. Conclusions: LMC in BC historically has a grim prognosis. Our institutional experience supports that a multimodal, multidisciplinary approach with sys tx and CNS tx improved OS in LMC in BC. Moreover, ADCs, when added to CNS tx, significantly improved OS. Larger prospective studies are needed to inform tx paradigms.

Multimodality therapy, including surgery, radiotherapy, and systemic therapy, has significantly improved overall survival for patients with brain metastases. However, treatment-related neurocognitive sequelae remain a major challenge in survivorship. Although advances in radiotherapy delivery techniques have reduced toxicity, the potential interaction with chemotherapy, targeted therapy, and immunotherapy, and the consequent effect on neurocognitive outcomes is poorly characterised. We conducted a systematic review of clinical trials reporting neurocognitive endpoints in patients with brain metastases receiving radiotherapy with or without other concurrent systemic therapies. Neurocognitive outcomes were manually extracted from published reports. 39 studies from 1997 to 2024 involving 6617 patients met inclusion criteria (n=27 whole-brain radiotherapy; n=12 radiosurgery), including six studies evaluating combined-modality therapy. Baseline neurocognitive disability was frequently observed, and the majority of randomised trials evaluating advanced radiotherapy delivery techniques (hippocampal avoidance and radiosurgery) compared with whole-brain radiotherapy reported reduced cognitive decline and improved quality of life. There was no signal for increased toxicity with combined-modality therapy, including radiotherapy with concurrent systemic therapy, although evaluable trials were few in number. Given improvements in survival for patients with brain metastases, characterisation of long-term neurocognitive outcomes is growing in importance. There is an urgent need for targeted research to resolve evidence gaps around modality-specific neurocognitive toxicity and optimal sequencing of therapies. Systemic issues, such as integration of routine neuropsychological screening or assessment and incorporation of rehabilitation strategies into neuro-oncology care pathways, warrant evaluation. Exploration of emerging strategies, ranging from neuroprotectants to dose-sparing radiotherapy techniques, could further mitigate long-term adverse effects.

Importance:Leptomeningeal metastasis (LM) is associated with limited survival and few treatment options. Photon involved-field radiotherapy (IFRT) is the most common radiotherapy treatment for patients with LM from solid tumors. Objective:To assess whether proton craniospinal irradiation (pCSI) would result in superior central nervous system progression-free survival (CNS-PFS) compared with IFRT. Design, Setting, and Participants:A randomized, phase 2 trial of pCSI vs IFRT was conducted between April 16, 2020, and October 11, 2021, and included patients with non-small cell lung cancer and breast cancer with LM. Patients with other solid tumors were also enrolled in an exploratory pCSI cohort. Intervention:For the randomized groups, after stratifying by histology and systemic disease status, patients were assigned (2:1) to pCSI or IFRT. Main Outcomes and Measures:The primary end point was CNS-PFS. Secondary end points included overall survival (OS). Results:Of 98 total patients, 72 individuals (73.5%) were female, and the median (IQR) age was 59 (50-65) years. A total of 42 and 21 patients were randomly assigned to pCSI and IFRT, respectively. At planned interim analysis, a significant benefit in CNS-PFS was observed with pCSI compared with IFRT, leading to the early discontinuation of the trial. In this final analysis, a significant benefit was continually observed in CNS-PFS with pCSI (median, 8.2 months; 95% CI, 6.6-15.3) vs IFRT (median, 2.3 months; 95% CI, 1.2-4.0; P < .001). A statistically significant and clinically meaningful OS benefit with pCSI (median, 11.3 months; 95% CI, 7.5-18.3) vs IFRT (median, 4.9 months; 95% CI, 3.9-15.0; P = .04) was also observed. For the exploratory pCSI cohort (n = 35), the median CNS-PFS was 5.8 months (95% CI, 4.4-9.1) and OS was 7.0 months (95% CI, 5.4-10.6). Conclusions and Relevance:This randomized clinical trial that assessed the optimal radiotherapy treatment for LM found improved CNS-PFS and OS with pCSI compared with IFRT. The results suggest that pCSI should be considered when available. Trial Registration:ClinicalTrials.gov Identifier: NCT04343573.

Craniospinal irradiation (CSI) is an effective treatment for solid tumor leptomeningeal carcinomatosis (LMC). While proton CSI (pCSI) has demonstrated improved survival compared to involved-field photon therapy, outcomes using volumetric modulated arc therapy (VMAT, or “xCSI”) for full neuraxis coverage remain unclear. This study compared toxicity and outcomes among patients with solid tumor LMC treated with either pCSI or xCSI, with the hypothesis that xCSI would be associated with increased toxicity but similar survival when adjusted for performance status. Adults diagnosed with LMC and treated with CSI to 30 Gy in 10 fractions between 2020 and 2025 were included. Acute and late toxicities (hematologic, fatigue, nausea, headache) were graded using CTCAE v5.0 during treatment and at 1- and 3-months post-CSI. Overall survival (OS) was calculated from CSI start and analyzed using Cox proportional hazards models adjusted for ECOG performance status and metastatic burden, defined as low (<2 organs) vs high (≥3 organs) involvement. Of 37 patients (median age 54), 21 (57%) received pCSI and 16 (43%) xCSI. Most patients had ECOG 1 (p=0.262), with breast (41%) and lung (38%) as the most common histologies (p=0.323). Grade ≥1 thrombocytopenia was significantly higher during xCSI (1.54 vs 0.36, p=0.005) but resolved by 1 month. No other significant differences in hematologic or non-hematologic toxicities were observed. Median OS was not significantly different: 11.6 months for pCSI vs 4.6 months for xCSI (p=0.238) and remained nonsignificant after adjusting for ECOG (HR 1.3; 95% CI, 0.17–1.34) and metastatic burden (HR 0.52; 95% CI, 0.19–1.47). However, low-volume metastatic disease was significantly associated with improved OS (HR 0.35; 95% CI, 0.14–0.91; p=0.032), independent of treatment modality. These findings support the safety of VMAT xCSI and suggest observed survival differences may reflect patient selection. Further study is needed to define xCSI’s role in LMC management.

2516 Background: ICANS is a complication of CAR T-cell therapy, yet risk factors and quantitative diagnostic criteria, particularly neuroimaging criteria, remain incompletely characterized. We implemented a novel application of a deep learning (DL)-based MRI approach alongside clinical and liquid biomarkers to better characterize neurotoxicity after CAR T-cell therapy. Methods: We analyzed all patients with non-Hodgkin lymphoma (NHL) or acute lymphoblastic leukemia (ALL) who underwent CAR T-cell therapy at UCSD with a commercial product from 2018-2024. ICANS was graded as per American Society for Transplantation and Cellular Therapy (gr1-4). Variables included stage, performance status, and prior receipt of high-dose methotrexate (HD MTX), intrathecal (IT) chemotherapy, central nervous system (CNS) involvement, CNS-directed radiotherapy (CNS RT), and extracranial RT. Labs obtained pre-infusion, 3 days post-infusion, and during ICANS (or 7 days post-infusion for those without ICANS) were evaluated. Available post-infusion brain MRIs were processed with a 3D U-Net convolutional neural network to quantify T2 FLAIR hyperintensity volumetrics. Linear mixed regression models accounting for zero inflation assessed longitudinal DL-derived FLAIR. Multivariable regression models assessed factors associated with ICANS. Results: Of 163 patients (89% NHL, 11% ALL), 52 had IT chemotherapy, 27 had HD MTX, 24 had prior CNS disease, and 22 had prior CNS RT. Most (106) received axicabtagene ciloleucel (34 tisagenlecleucel, 23 brexucabtagene autoleucel) and most had CRS (133, 82%). ICANS occurred in 73 (45%) at a median of 7 days post-infusion (39 gr1-2, 34 gr3-4). Post-infusion, 21 patients had ³1 brain MRI (93 MRIs total). Baseline factors associated with ICANS were lactate dehydrogenase (LDH; odds ratio [OR] 1.03 p = 0.002) and prior IT chemotherapy (OR 2.5 p = 0.01). There was a trend toward association of gr3-4 ICANS with HD MTX (OR 2.8 p = 0.07). Post-infusion, CRS grade was associated with ICANS (OR 2.8 p < 0.001). LDH (1.02 p = 0.004) and C-reactive protein (OR 1.2 p < 0.001) were elevated during ICANS. Patients with ICANS had significantly greater FLAIR (intercept 23.8 cm³ p < 0.001) and there was increased FLAIR over time across all patients (b = 3.3 cm³ p = 0.05). There was a trend toward association between higher ICANS grade and DL-derived FLAIR (p = 0.09). Conclusions: Here, we demonstrate a novel application of DL-based MRI quantification of ICANS post-CAR T-cell therapy. This metric, along with clinical features, emerged as potential quantitative biomarkers of ICANS. These findings warrant further investigation and have informed a prospective study, including standardized brain MRI pre- and post-infusion, to develop a comprehensive phenotype of neurotoxicity following CAR T-cell therapy.

Background:Differentiating recurrent tumor from post-treatment changes remains a major challenge in glioblastoma (GBM) patients. In this work, we compared the performance of 2 different MR perfusion techniques, dynamic susceptibility contrast (DSC), and arterial spin labeling (ASL) to differentiate recurrent tumor and post-treatment changes from the volume of cellular tumor segmented from combined Deep Learning and multimodal MRI measurements, including multishell diffusion and perfusion. Methods:In this retrospective study, 137 MRIs from 107 patients with GBM were analyzed. Cellular tumor maps were segmented by 2 radiologists based on imaging, clinical history, and pathology. Multimodal MRI with perfusion and multishell diffusion were inputted into 5 nnU-Net Deep Learning models using either DSC or ASL with combination of multishell diffusion and standard MRI sequences to segment cellular tumor. Models with DSC and ASL were compared using segmentation performance (Dice score) and accuracy to detect recurrent tumor from post-treatment changes (area under the curve [AUC] under the receiver operating characteristic curve). Results:Segmentation performances were similar in both cases, with a median Dice score of 0.75 (IQR: 0.53-0.84) for ASL and 0.76 (IQR: 0.57-0.84). AUC was 0.88 (CI 0.82-0.94) for ASL and 0.86 (CI, 0.80-0.92) for DSC, and this difference was statistically significant (P < .05, n = 10 000 permutation test). In 11 individual cases, recurring disease was detected with ASL but missed with cerebral blood volume, including recurring tumor in the vicinity of a surgical cavity (n = 5), close to the skull base (n = 1), and adjacent to an Ommaya reservoir (n = 2). Conclusions:Our results demonstrate the utility of ASL in regions where susceptibility artifacts decrease the quality of DSC images.

Purpose This study investigated patterns of cortical atrophy longitudinally in a prospective cohort of well-characterized patients with primary brain tumors before radiation therapy (RT) and at discrete time points up to 1-year post-RT. Patients with more impaired profiles were hypothesized to show greater and more widespread atrophy post-RT than patients with minimally impaired profiles. Methods and Materials Adults (aged 20-72 years) with primary brain tumors were enrolled in a prospective, observational study examining the effects of fractionated, partial brain RT on brain structure and cognition (n = 85). Cortical atrophy rates were compared using multilevel modeling across 4 different time points (baseline, 3, 6, and 12 months after RT) between minimal, generalized, and isolated verbal memory impairment phenotypes taken from a previous study's latent profile analysis. Results The minimal impairment phenotype showed no regions of significant change in cortical thickness across the 4 time points analyzed. The generalized impairment phenotype showed bihemispheric, multilobar cortical atrophy, with the greatest changes observed in the left precentral gyrus, lateral occipital cortex, and frontal pole. The verbal memory impairment phenotype demonstrated less extensive atrophy focally in the left superior temporal and superior parietal cortices. Annualized atrophy rates in regions of significant cortical thinning in the generalized and verbal memory impairment phenotypes ranged from 3.5% to 8.7% compared with baseline. Conclusions In patients with primary brain tumors, the cognitive profile before RT informs the pattern and risk for RT-related cortical atrophy. Cognitive profiles may aid in identifying patients at greater risk for RT-related atrophy and accelerated cognitive decline.