While immune checkpoint inhibitors (ICIs) in combination with radiotherapy (RT) are widely used for patients with brain metastasis (BM), markers that predict treatment response for combined RT and ICI (RT-ICI) and their optimal dosing and sequence for the best immunogenic effects are still under investigation. The aim of this study was to evaluate prognostic factors for therapeutic outcome and to compare effects of concurrent and non-concurrent RT-ICI. We retrospectively analyzed data of 93 patients with 319 BMs of different cancer types who received PD-1 inhibitors and RT at the University Hospital Cologne between September/2014 and November/2020. Primary study endpoints were overall survival (OS), progression-free survival (PFS), and local control (LC). We included 66.7% melanoma, 22.8% lung, and 5.5% other cancer types with a mean follow-up time of 23.8 months. Median OS time was 12.19 months. LC at 6 months was 95.3% (concurrent) vs. 69.2% (non-concurrent; p = 0.008). Univariate Cox regression analysis detected following prognostic factors for OS: neutrophil-to-lymphocyte ratio NLR favoring <3 (low; HR 2.037 (1.184–3.506), p = 0.010), lactate dehydrogenase (LDH) favoring ≤ULN (HR 1.853 (1.059–3.241), p = 0.031), absence of neurological symptoms (HR 2.114 (1.285–3.478), p = 0.003), RT concept favoring SRS (HR 1.985 (1.112–3.543), p = 0.019), RT dose favoring ≥60 Gy (HR 0.519 (0.309–0.871), p = 0.013), and prior anti-CTLA4 treatment (HR 0.498 (0.271–0.914), p = 0.024). Independent prognostic factors for OS were concurrent RT-ICI application (HR 0.539 (0.299–0.971), p = 0.024) with a median OS of 17.61 vs. 6.83 months (non-concurrent), ECOG performance status favoring 0 (HR 7.756 (1.253–6.061), p = 0.012), cancer type favoring melanoma (HR 0.516 (0.288–0.926), p = 0.026), BM volume (PTV) favoring ≤3 cm³ (HR 1.947 (1.007–3.763), p = 0.048). Subgroups with the following factors showed significantly longer OS when being treated concurrently: RT dose 3 cm³ (p = 0.007), other cancer types than melanoma (p = 0.006), anti-CTLA4-naïve patients (p < 0.001), low NLR (p = 0.039), steroid intake ≤4 mg (p = 0.042). Specific immune responses, such as abscopal effects (AbEs), pseudoprogression (PsP), or immune-related adverse events (IrAEs), occurred more frequently with concurrent RT-ICI and resulted in better OS. Other toxicities, including radionecrosis, were not statistically different in both groups. The concurrent application of RT and ICI, the ECOG-PS, cancer type, and PTV had an independently prognostic impact on OS. In concurrently treated patients, treatment response (LC) was delayed and specific immune responses (AbE, PsP, IrAE) occurred more frequently with longer OS rates. Our results suggest that concurrent RT-ICI application is more beneficial than sequential treatment in patients with low pretreatment inflammatory status, more and larger BMs, and with other cancer types than melanoma.

Background: The expression level of the programmed cell death ligand 1 (PD-L1) appears to be a predictor for response to immunotherapy using checkpoint inhibitors in patients with non-small cell lung cancer (NSCLC). As differences in terms of PD-L1 expression levels in the extracranial primary tumor and the brain metastases may occur, a reliable method for the non-invasive assessment of the intracranial PD-L1 expression is, therefore of clinical value. Here, we evaluated the potential of radiomics for a non-invasive prediction of PD-L1 expression in patients with brain metastases secondary to NSCLC. Patients and methods: Fifty-three NSCLC patients with brain metastases from two academic neuro-oncological centers (group 1, n = 36 patients; group 2, n = 17 patients) underwent tumor resection with a subsequent immunohistochemical evaluation of the PD-L1 expression. Brain metastases were manually segmented on preoperative T1-weighted contrast-enhanced MRI. Group 1 was used for model training and validation, group 2 for model testing. After image pre-processing and radiomics feature extraction, a test-retest analysis was performed to identify robust features prior to feature selection. The radiomics model was trained and validated using random stratified cross-validation. Finally, the best-performing radiomics model was applied to the test data. Diagnostic performance was evaluated using receiver operating characteristic (ROC) analyses. Results: An intracranial PD-L1 expression (i.e., staining of at least 1% or more of tumor cells) was present in 18 of 36 patients (50%) in group 1, and 7 of 17 patients (41%) in group 2. Univariate analysis identified the contrast-enhancing tumor volume as a significant predictor for PD-L1 expression (area under the ROC curve (AUC), 0.77). A random forest classifier using a four-parameter radiomics signature, including tumor volume, yielded an AUC of 0.83 ± 0.18 in the training data (group 1), and an AUC of 0.84 in the external test data (group 2). Conclusion: The developed radiomics classifiers allows for a non-invasive assessment of the intracranial PD-L1 expression in patients with brain metastases secondary to NSCLC with high accuracy.

Purpose: Recent artificial intelligence algorithms aided intraoperative decision-making via stimulated Raman histology (SRH) during craniotomy. This study assesses deep learning algorithms for rapid intraoperative diagnosis from SRH images in small stereotactic-guided brain biopsies. It defines a minimum tissue sample size threshold to ensure diagnostic accuracy. Experimental Design: A prospective single-center study examined 121 SRH images from 84 patients with unclear intracranial lesions undergoing stereotactic brain biopsy. Unprocessed, label-free samples were imaged using a portable fiber laser Raman scattering microscope. Three deep learning models were tested to (i) identify tumorous/nontumorous tissue as qualitative biopsy control; (ii) subclassify into high-grade glioma (central nervous system World Health Organization grade 4), diffuse low-grade glioma (central nervous system World Health Organization grades 2–3), metastases, lymphoma, or gliosis; and (iii) molecularly subtype IDH and 1p/19q statuses of adult-type diffuse gliomas. Model predictions were evaluated against frozen section analysis and final neuropathologic diagnoses. Results: The first model identified tumorous/nontumorous tissue with 91.7% accuracy. Sample size on slides impacted accuracy in brain tumor subclassification (81.6%, κ = 0.72 frozen section; 73.9%, κ = 0.61 second model), with SRH images being smaller than hematoxylin and eosin images (4.1 ± 2.5 mm2 vs. 16.7 ± 8.2 mm2, P < 0.001). SRH images with more than 140 high-quality patches and a mean squeezed sample of 5.26 mm2 yielded 89.5% accuracy in subclassification and 93.9% in molecular subtyping of adult-type diffuse gliomas. Conclusions: Artificial intelligence–based SRH image analysis is non-inferior to frozen section analysis in detecting and subclassifying brain tumors during small stereotactic-guided biopsies once a critical squeezed sample size is reached. Beyond frozen section analysis, it enables valid molecular glioma subtyping, allowing faster treatment decisions in the future; however, refinement is needed for long-term application.

The treatment of symptomatic, progressive or recurrent acquired intracerebral cysts is challenging, especially when they are localized in eloquent structures. In addition to resection, endoscopic fenestration, or stereotactic puncture, the implantation of a cystoventricular shunt by stereotactic guidance (SCVS) has been reported as a minimally invasive procedure; however, only scarce data are available regarding its feasibility and efficacy. Here, the authors evaluated the feasibility and efficacy of frame-based SCVS in patients with acquired intracranial cysts. In this single-center retrospective analysis, the authors included all patients with acquired intracerebral cysts treated by SCVS following a standardized prospective protocol between 2012 and 2020. They analyzed clinical symptoms, complications, and radiological outcome with regard to cyst volume reduction by 3D volumetry. Thirty-four patients (17 females and 17 males; median age 44 years, range 5–77 years) were identified. The median initial cyst volume was 11.5 cm³ (range 1.6–71.6 cm³), and the mean follow-up was 20 months (range 1–82 months). At the last follow-up, 27 of 34 patients (79%) showed a cyst volume reduction of more than 50%. Initial symptoms improved or resolved in 74% (n = 25) and remained stable in 24% (n = 8). No permanent clinical deterioration after treatment was observed. The total complication rate was 5.9%, comprising transient neurological deterioration (n = 1) and ventriculitis (n = 1). There were no deaths. The overall recurrence rate was 11.8%. In this study, SCVS proved to be a safe, minimally invasive, and effective treatment with reliable long-term volume reduction, resulting in clinical improvement and a minor complication rate.

Background: De- and acceleration traumata can cause diffuse axonal injury (DAI) in patients with traumatic brain injury (TBI). The diagnosis of DAI on CT is challenging due to the lack of structural abnormalities. Radiomics, a method from the field of artificial intelligence (AI) offers the opportunity to extract additional information from imaging data. The purpose of this work was the evaluation of the feasibility of radiomics for an improved diagnosis of DAI in comparison to conventional radiological image assessment. Methods: CT and MR imaging was performed in 42 patients suspicious of DAI due to the clinical state, and two control groups (n = 44;42). DAI was diagnosed by experienced neuroradiologists. Radiomics features were extracted using a standardized MRI-based atlas of the predilection areas for DAI. Different MRI and CT based models were trained and validated by five-fold cross validation. Diagnostic performance was compared to the reading of two experienced radiologists and further validated in an external test dataset. Results: The MRI and CT models showed significant differences in radiomics features between patients with DAI and controls. The developed MRI based random forest classifier yielded an accuracy of 80–90%. The best performing CT model yielded an accuracy of 88% in the training data and 70% in the external test data. The results were comparable to conventional image analysis which achieved an accuracy of 70–81% for CT-based diagnosis. Conclusion: MRI- and CT-based radiomics analysis is feasible for the assessment of DAI. The radiomics classifier achieved equivalent performance rates as visual radiological image diagnosis. Especially a radiomics based CT classifier can be of clinical value as a screening and AI-based decision support tool for patients with TBI.

Background: Stimulated Raman histology (SRH) is a label-free optical imaging method for rapid intraoperative analysis of fresh tissue samples. Analysis of SRH images using Convolutional Neural Networks (CNN) has shown promising results for predicting the main histopathological classes of neurooncological tumors. Due to the relatively low number of rare tumor representations in CNN training datasets, a valid prediction of rarer entities remains limited. To develop new reliable analysis tools, larger datasets and greater tumor variety are crucial. One way to accomplish this is through research biobanks storing frozen tumor tissue samples. However, there is currently no data available regarding the pertinency of previously frozen tissue samples for SRH analysis. The aim of this study was to assess image quality and perform a comparative reliability analysis of artificial intelligence-based tumor classification using SRH in fresh and frozen tissue samples. Methods: In a monocentric prospective study, tissue samples from 25 patients undergoing brain tumor resection were obtained. SRH was acquired in fresh and defrosted samples of the same specimen after varying storage durations at −80 °C. Image quality was rated by an experienced neuropathologist, and prediction of histopathological diagnosis was performed using two established CNNs. Results: The image quality of SRH in fresh and defrosted tissue samples was high, with a mean image quality score of 1.96 (range 1–5) for both groups. CNN analysis showed high internal consistency for histo-(Cα 0.95) and molecular (Cα 0.83) pathological tumor classification. The results were confirmed using a dataset with samples from the local tumor biobank (Cα 0.91 and 0.53). Conclusions: Our results showed that SRH appears comparably reliable in fresh and frozen tissue samples, enabling the integration of tumor biobank specimens to potentially improve the diagnostic range and reliability of CNN prediction tools.