A critical challenge in glioma treatment is detecting tumour infiltration during surgery to achieve safe maximal resection^1,2,3. Unfortunately, safely resectable residual tumour is found in the majority of patients with glioma after surgery, causing early recurrence and decreased survival^4,5,6. Here we present FastGlioma, a visual foundation model for fast (<10 s) and accurate detection of glioma infiltration in fresh, unprocessed surgical tissue. FastGlioma was pretrained using large-scale self-supervision (around 4 million images) on rapid, label-free optical microscopy, and fine-tuned to output a normalized score that indicates the degree of tumour infiltration within whole-slide optical images. In a prospective, multicentre, international testing cohort of patients with diffuse glioma (n = 220), FastGlioma was able to detect and quantify the degree of tumour infiltration with an average area under the receiver operating characteristic curve of 92.1 ± 0.9%. FastGlioma outperformed image-guided and fluorescence-guided adjuncts for detecting tumour infiltration during surgery by a wide margin in a head-to-head, prospective study (n = 129). The performance of FastGlioma remained high across diverse patient demographics, medical centres and diffuse glioma molecular subtypes as defined by the World Health Organization. FastGlioma shows zero-shot generalization to other adult and paediatric brain tumour diagnoses, demonstrating the potential for our foundation model to be used as a general-purpose adjunct for guiding brain tumour surgeries. These findings represent the transformative potential of medical foundation models to unlock the role of artificial intelligence in the care of patients with cancer.

The prediction of the individual prognosis of low-grade glioma (LGG) patients is limited in routine clinical practice. Nowadays, 5-aminolevulinic acid (5-ALA) fluorescence is primarily applied for improved intraoperative visualization of high-grade gliomas. However, visible fluorescence is also observed in rare cases despite LGG histopathology and might be an indicator for aggressive tumor behavior. The aim of this study was thus to investigate the value of intraoperative 5-ALA fluorescence for prognosis in LGG patients. We performed a retrospective analysis of patients with newly diagnosed histopathologically confirmed LGG and preoperative 5-ALA administration at two independent specialized centers. In this cohort, we correlated the visible intraoperative fluorescence status with progression-free survival (PFS), malignant transformation-free survival (MTFS) and overall survival (OS). Altogether, visible fluorescence was detected in 7 (12%) of 59 included patients in focal intratumoral areas. At a mean follow-up time of 5.3 ± 2.9 years, patients with fluorescing LGG had significantly shorter PFS (2.3 ± 0.7 vs. 5.0 ± 0.4 years; p = 0.01), MTFS (3.9 ± 0.7 vs. 8.0 ± 0.6 years; p = 0.03), and OS (5.4 ± 1.0 vs. 10.3 ± 0.5 years; p = 0.01) than non-fluorescing tumors. Our data indicate that visible 5-ALA fluorescence during surgery of pure LGG might be an already intraoperatively available marker of unfavorable patient outcome and thus close imaging follow-up might be considered.

In this study, we assessed the prognostic relevance of temporal muscle thickness (TMT), likely reflecting patient’s frailty, in patients with primary central nervous system lymphoma (PCNSL). In 128 newly diagnosed PCNSL patients TMT was analyzed on cranial magnetic resonance images. Predefined sex-specific TMT cutoff values were used to categorize the patient cohort. Survival analyses, using a log-rank test as well as Cox models adjusted for further prognostic parameters, were performed. The risk of death was significantly increased for PCNSL patients with reduced muscle thickness (hazard ratio of 3.189, 95% CI: 2–097–4.848, p < 0.001). Importantly, the results confirmed that TMT could be used as an independent prognostic marker upon multivariate Cox modeling (hazard ratio of 2.504, 95% CI: 1.608–3.911, p < 0.001) adjusting for sex, age at time of diagnosis, deep brain involvement of the PCNSL lesions, Eastern Cooperative Oncology Group (ECOG) performance status, and methotrexate-based chemotherapy. A TMT value below the sex-related cutoff value at the time of diagnosis is an independent adverse marker in patients with PCNSL. Thus, our results suggest the systematic inclusion of TMT in further translational and clinical studies designed to help validate its role as a prognostic biomarker.

Fluorescence-guided surgery is a state-of-the-art approach for intraoperative imaging during neurosurgical removal of tumor tissue. While the visualization of high-grade gliomas is reliable, lower grade glioma often lack visible fluorescence signals. Here, we present a hybrid prototype combining visible light optical coherence microscopy (OCM) and high-resolution fluorescence imaging for assessment of brain tumor samples acquired by 5-aminolevulinic acid (5-ALA) fluorescence-guided surgery. OCM provides high-resolution information of the inherent tissue scattering and absorption properties of tissue. We here explore quantitative attenuation coefficients derived from volumetric OCM intensity data and quantitative high-resolution 5-ALA fluorescence as potential biomarkers for tissue malignancy including otherwise difficult-to-assess low-grade glioma. We validate our findings against the gold standard histology and use attenuation and fluorescence intensity measures to differentiate between tumor core, infiltrative zone and adjacent brain tissue. Using large field-of-view scans acquired by a near-infrared swept-source optical coherence tomography setup, we provide initial assessments of tumor heterogeneity. Finally, we use cross-sectional OCM images to train a convolutional neural network that discriminates tumor from non-tumor tissue with an accuracy of 97%. Collectively, the present hybrid approach offers potential to translate into an in vivo imaging setup for substantially improved intraoperative guidance of brain tumor surgeries.

Background: Accurate intraoperative diagnosis is crucial for differentiating between primary central nervous system (CNS) lymphoma (PCNSL) and other CNS entities, guiding surgical decision-making, but represents significant challenges due to overlapping histomorphological features, time constraints, and differing treatment strategies. We combined stimulated Raman histology (SRH) with deep learning to address this challenge. Methods: We imaged unprocessed, label-free tissue samples intraoperatively using a portable Raman scattering microscope, generating virtual H&E-like images within <3 min. We developed a deep learning pipeline called RapidLymphoma based on a self-supervised learning strategy to (1) detect PCNSL, (2) differentiate from other CNS entities, and (3) test the diagnostic performance in a prospective international multicenter cohort and 2 additional independent test cohorts. We trained on 54 000 SRH patch images sourced from surgical resections and stereotactic-guided biopsies, including various CNS neoplastic/nonneoplastic lesions. Training and test data were collected from 4 tertiary international medical centers. The final histopathological diagnosis served as ground truth. Results: In the prospective test cohort of PCNSL and non-PCNSL entities (n = 160), RapidLymphoma achieved an overall balanced accuracy of 97.81% ± 0.91, non-inferior to frozen section analysis in detecting PCNSL (100% vs. 77.77%). The additional test cohorts (n = 420, n = 59) reached balanced accuracy rates of 95.44% ± 0.74 and 95.57% ± 2.47 in differentiating IDH-wildtype diffuse gliomas and various brain metastasis from PCNSL. Visual heatmaps revealed RapidLymphoma’s capabilities to detect class-specific histomorphological key features. Conclusions: RapidLymphoma proves reliable and valid for intraoperative PCNSL detection and differentiation from other CNS entities. It provides visual feedback within 3 min, enabling fast clinical decision-making and subsequent treatment strategy planning.

Accurate pathological diagnosis is crucial in guiding personalized treatments for patients with central nervous system cancers. Distinguishing glioblastoma and primary central nervous system lymphoma is particularly challenging due to their overlapping pathology features, despite the distinct treatments required. To address this challenge, we establish the Pathology Image Characterization Tool with Uncertainty-aware Rapid Evaluations (PICTURE) system using 2141 pathology slides collected worldwide. PICTURE employs Bayesian inference, deep ensemble, and normalizing flow to account for the uncertainties in its predictions and training set labels. PICTURE accurately diagnoses glioblastoma and primary central nervous system lymphoma with an area under the receiver operating characteristic curve (AUROC) of 0.989, with the results validated in five independent cohorts (AUROC = 0.924-0.996). In addition, PICTURE identifies samples belonging to 67 types of rare central nervous system cancers that are neither gliomas nor lymphomas. Our approaches provide a generalizable framework for differentiating pathological mimics and enable rapid diagnoses for central nervous system cancer patients.