Super-resolution of biomedical volumes with 2D supervision.

Volumetric biomedical microscopy has the potential to increase the diagnostic information extracted from clinical tissue specimens and improve the diagnostic accuracy of both human pathologists and computational pathology models. Unfortunately, barriers to integrating 3-dimensional (3D) volumetric microscopy into clinical medicine include long imaging times, poor depth/z-axis resolution, and an insufficient amount of high-quality volumetric data. Leveraging the abundance of high-resolution 2D microscopy data, we introduce masked slice diffusion for super-resolution (MSDSR), which exploits the inherent equivalence in the data-generating distribution across all spatial dimensions of biological specimens. This intrinsic characteristic allows for super-resolution models trained on high-resolution images from one plane (e.g., XY) to effectively generalize to others (XZ, YZ), overcoming the traditional dependency on orientation. We focus on the application of MSDSR to stimulated Raman histology (SRH), an optical imaging modality for biological specimen analysis and intraoperative diagnosis, characterized by its rapid acquisition of high-resolution 2D images but slow and costly optical z-sectioning. To evaluate MSDSR's efficacy, we introduce a new performance metric, SliceFID, and demonstrate MSDSR's superior performance over baseline models through extensive evaluations. Our findings reveal that MSDSR not only significantly enhances the quality and resolution of 3D volumetric data, but also addresses major obstacles hindering the broader application of 3D volumetric microscopy in clinical diagnostics and biomedical research.

Fast intraoperative detection of primary CNS lymphoma and differentiation from common CNS tumors using stimulated Raman histology and deep learning.

Accurate intraoperative diagnosis is crucial for differentiating between primary 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. We imaged unprocessed, label-free tissue samples intraoperatively using a portable Raman scattering microscope, generating virtual H&E-like images within less than three minutes. 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 two additional independent test cohorts. We trained on 54,000 SRH patch images sourced from surgical resections and stereotactic-guided biopsies, including various CNS tumor/non-tumor lesions. Training and test data were collected from four tertiary international medical centers. The final histopathological diagnosis served as ground-truth. 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. 78.94%). 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. RapidLymphoma is valid and reliable in detecting PCNSL and differentiating from other CNS entities within three minutes, as well as visual feedback in an intraoperative setting. This leads to fast clinical decision-making and further treatment strategy planning.

Localization of protoporphyrin IX during glioma-resection surgery via paired stimulated Raman histology and fluorescence microscopy.

The most widely used fluorophore in glioma-resection surgery, 5-aminolevulinic acid (5-ALA), is thought to cause the selective accumulation of fluorescent protoporphyrin IX (PpIX) in tumour cells. Here we show that the clinical detection of PpIX can be improved via a microscope that performs paired stimulated Raman histology and two-photon excitation fluorescence microscopy (TPEF). We validated the technique in fresh tumour specimens from 115 patients with high-grade gliomas across four medical institutions. We found a weak negative correlation between tissue cellularity and the fluorescence intensity of PpIX across all imaged specimens. Semi-supervised clustering of the TPEF images revealed five distinct patterns of PpIX fluorescence, and spatial transcriptomic analyses of the imaged tissue showed that myeloid cells predominate in areas where PpIX accumulates in the intracellular space. Further analysis of external spatially resolved metabolomics, transcriptomics and RNA-sequencing datasets from glioblastoma specimens confirmed that myeloid cells preferentially accumulate and metabolize PpIX. Our findings question 5-ALA-induced fluorescence in glioma cells and show how 5-ALA and TPEF imaging can provide a window into the immune microenvironment of gliomas.

Development and validation of an artificial intelligence model to accurately predict spinopelvic parameters.

OBJECTIVE: Achieving appropriate spinopelvic alignment has been shown to be associated with improved clinical symptoms. However, measurement of spinopelvic radiographic parameters is time-intensive and interobserver reliability is a concern. Automated measurement tools have the promise of rapid and consistent measurements, but existing tools are still limited to some degree by manual user-entry requirements. This study presents a novel artificial intelligence (AI) tool called SpinePose that automatically predicts spinopelvic parameters with high accuracy without the need for manual entry. METHODS: SpinePose was trained and validated on 761 sagittal whole-spine radiographs to predict the sagittal vertical axis (SVA), pelvic tilt (PT), pelvic incidence (PI), sacral slope (SS), lumbar lordosis (LL), T1 pelvic angle (T1PA), and L1 pelvic angle (L1PA). A separate test set of 40 radiographs was labeled by four reviewers, including fellowship-trained spine surgeons and a fellowship-trained radiologist with neuroradiology subspecialty certification. Median errors relative to the most senior reviewer were calculated to determine model accuracy on test images. Intraclass correlation coefficients (ICCs) were used to assess interrater reliability. RESULTS: SpinePose exhibited the following median (interquartile range) parameter errors: SVA 2.2 mm (2.3 mm) (p = 0.93), PT 1.3° (1.2°) (p = 0.48), SS 1.7° (2.2°) (p = 0.64), PI 2.2° (2.1°) (p = 0.24), LL 2.6° (4.0°) (p = 0.89), T1PA 1.1° (0.9°) (p = 0.42), and L1PA 1.4° (1.6°) (p = 0.49). Model predictions also exhibited excellent reliability at all parameters (ICC 0.91-1.0). CONCLUSIONS: SpinePose accurately predicted spinopelvic parameters with excellent reliability comparable to that of fellowship-trained spine surgeons and neuroradiologists. Utilization of predictive AI tools in spinal imaging can substantially aid in patient selection and surgical planning.

Combined cytotoxic and immune-stimulatory gene therapy for primary adult high-grade glioma: a phase 1, first-in-human trial.

BACKGROUND: High-grade gliomas have a poor prognosis and do not respond well to treatment. Effective cancer immune responses depend on functional immune cells, which are typically absent from the brain. This study aimed to evaluate the safety and activity of two adenoviral vectors expressing HSV1-TK (Ad-hCMV-TK) and Flt3L (Ad-hCMV-Flt3L) in patients with high-grade glioma. METHODS: In this dose-finding, first-in-human trial, treatment-naive adults aged 18-75 years with newly identified high-grade glioma that was evaluated per immunotherapy response assessment in neuro-oncology criteria, and a Karnofsky Performance Status score of 70 or more, underwent maximal safe resection followed by injections of adenoviral vectors expressing HSV1-TK and Flt3L into the tumour bed. The study was conducted at the University of Michigan Medical School, Michigan Medicine (Ann Arbor, MI, USA). The study included six escalating doses of viral particles with starting doses of 1×1010 Ad-hCMV-TK viral particles and 1×109 Ad-hCMV-Flt3L viral particles (cohort A), and then 1×1011 Ad-hCMV-TK viral particles and 1×109 Ad-hCMV-Flt3L viral particles (cohort B), 1×1010 Ad-hCMV-TK viral particles and 1×1010 Ad-hCMV-Flt3L viral particles (cohort C), 1×1011 Ad-hCMV-TK viral particles and 1×1010 Ad-hCMV-Flt3L viral particles (cohort D), 1×1010 Ad-hCMV-TK viral particles and 1×1011 Ad-hCMV-Flt3L viral particles (cohort E), and 1×1011 Ad-hCMV-TK viral particles and 1×1011 Ad-hCMV-Flt3L viral particles (cohort F) following a 3+3 design. Two 1 mL tuberculin syringes were used to deliver freehand a mix of Ad-hCMV-TK and Ad-hCMV-Flt3L vectors into the walls of the resection cavity with a total injection of 2 mL distributed as 0·1 mL per site across 20 locations. Subsequently, patients received two 14-day courses of valacyclovir (2 g orally, three times per day) at 1-3 days and 10-12 weeks after vector administration and standad upfront chemoradiotherapy. The primary endpoint was the maximum tolerated dose of Ad-hCMV-Flt3L and Ad-hCMV-TK. Overall survival was a secondary endpoint. Recruitment is complete and the trial is finished. The trial is registered with ClinicalTrials.gov, NCT01811992. FINDINGS: Between April 8, 2014, and March 13, 2019, 21 patients were assessed for eligibility and 18 patients with high-grade glioma were enrolled and included in the analysis (three patients in each of the six dose cohorts); eight patients were female and ten were male. Neuropathological examination identified 14 (78%) patients with glioblastoma, three (17%) with gliosarcoma, and one (6%) with anaplastic ependymoma. The treatment was well-tolerated, and no dose-limiting toxicity was observed. The maximum tolerated dose was not reached. The most common serious grade 3-4 adverse events across all treatment groups were wound infection (four events in two patients) and thromboembolic events (five events in four patients). One death due to an adverse event (respiratory failure) occurred but was not related to study treatment. No treatment-related deaths occurred during the study. Median overall survival was 21·3 months (95% CI 11·1-26·1). INTERPRETATION: The combination of two adenoviral vectors demonstrated safety and feasibility in patients with high-grade glioma and warrants further investigation in a phase 1b/2 clinical trial. FUNDING: Funded in part by Phase One Foundation, Los Angeles, CA, The Board of Governors at Cedars-Sinai Medical Center, Los Angeles, CA, and The Rogel Cancer Center at The University of Michigan.

Hierarchical discriminative learning improves visual representations of biomedical microscopy.

Learning high-quality, self-supervised, visual representations is essential to advance the role of computer vision in biomedical microscopy and clinical medicine. Previous work has focused on self-supervised representation learning (SSL) methods developed for instance discrimination and applied them directly to image patches, or fields-of-view, sampled from gigapixel whole-slide images (WSIs) used for cancer diagnosis. However, this strategy is limited because it (1) assumes patches from the same patient are independent, (2) neglects the patient-slide-patch hierarchy of clinical biomedical microscopy, and (3) requires strong data augmentations that can degrade downstream performance. Importantly, sampled patches from WSIs of a patient's tumor are a diverse set of image examples that capture the same underlying cancer diagnosis. This motivated HiDisc, a data-driven method that leverages the inherent patient-slide-patch hierarchy of clinical biomedical microscopy to define a hierarchical discriminative learning task that implicitly learns features of the underlying diagnosis. HiDisc uses a self-supervised contrastive learning framework in which positive patch pairs are defined based on a common ancestry in the data hierarchy, and a unified patch, slide, and patient discriminative learning objective is used for visual SSL. We benchmark HiDisc visual representations on two vision tasks using two biomedical microscopy datasets, and demonstrate that (1) HiDisc pretraining outperforms current state-of-the-art self-supervised pretraining methods for cancer diagnosis and genetic mutation prediction, and (2) HiDisc learns high-quality visual representations using natural patch diversity without strong data augmentations.

Unlocking glioma genetics with deep learning.

The AI era in medicine has ushered in new opportunities to improve the diagnosis and treatment of human disease. CHARM, an AI algorithm described in this issue,1 has the potential to streamline molecular classification, intraoperative diagnosis, surgical decision making, and trial enrollment for glioma patients.

A comparison of ventricular volume and linear indices in predicting shunt dependence in aneurysmal subarachnoid hemorrhage.

Background: Guidelines for determining shunt dependence after aneurysmal subarachnoid hemorrhage (aSAH) remain unclear. We previously demonstrated change in ventricular volume (VV) between head CT scans taken pre- and post-EVD clamping was predictive of shunt dependence in aSAH. We sought to compare the predictive value of this measure to more commonly used linear indices. Methods: We retrospectively analyzed images of 68 patients treated for aSAH who required EVD placement and underwent one EVD weaning trial, 34 of whom underwent shunt placement. We utilized an in-house MATLAB program to analyze VV and supratentorial VV (sVV) in head CT scans obtained before and after EVD clamping. Evans' index (EI), frontal and occipital horn ratio (FOHR), Huckman's measurement, minimum lateral ventricular width (LV-Min.), and lateral ventricle body span (LV-Body) were measured using digital calipers in PACS. Receiver operating curves (ROC) were generated. Results: Area under the ROC curves (AUC) for the change in VV, sVV, EI, FOHR, Huckman's, LV-Min., and LV-Body with clamping were 0.84, 0.84, 0.65, 0.71.0.69, 0.67, and 0.66, respectively. AUC for post-clamp scan measurements were 0.75, 0.75, 0.74, 0.72, 0.72, 0.70, and 0.75, respectively. Conclusion: VV change with EVD clamping was more predictive of shunt dependence in aSAH than change in linear measurements with clamping and all post-clamp measurements. Measurement of ventricular size on serial imaging with volumetrics or linear indices utilizing multidimensional data points may therefore be a more robust metric than unidimensional linear indices in predicting shunt dependence in this cohort. Prospective studies are needed for validation.

102 AI-Based Molecular Classification of Diffuse Gliomas using Rapid, Label-Free Optical Imaging.

INTRODUCTION: Molecular classification has transformed the management of brain tumors by enabling more accurate prognostication and personalized treatment. Access to timely molecular diagnostic testing for brain tumor patients is limited, complicating surgical and adjuvant treatment and obstructing clinical trial enrollment. METHODS: By combining stimulated Raman histology (SRH), a rapid, label-free, non-consumptive, optical imaging method, and deep learning-based image classification, we are able to predict the molecular genetic features used by the World Health Organization (WHO) to define the adult-type diffuse glioma taxonomy, including IDH-1/2, 1p19q-codeletion, and ATRX loss. We developed a multimodal deep neural network training strategy that uses both SRH images and large-scale, public diffuse glioma genomic data (i.e. TCGA, CGGA, etc.) in order to achieve optimal molecular classification performance. RESULTS: One institution was used for model training (University of Michigan) and four institutions (NYU, UCSF, Medical University of Vienna, and University Hospital Cologne) were included for patient enrollment in the prospective testing cohort. Using our system, called DeepGlioma, we achieved an average molecular genetic classification accuracy of 93.2% and identified the correct diffuse glioma molecular subgroup with 91.5% accuracy within 2 minutes in the operating room. DeepGlioma outperformed conventional IDH1-R132H immunohistochemistry (94.2% versus 91.4% accuracy) as a first-line molecular diagnostic screening method for diffuse gliomas and can detect canonical and non-canonical IDH mutations. CONCLUSIONS: Our results demonstrate how artificial intelligence and optical histology can be used to provide a rapid and scalable alternative to wet lab methods for the molecular diagnosis of brain tumor patients during surgery.

Artificial-intelligence-based molecular classification of diffuse gliomas using rapid, label-free optical imaging.

Molecular classification has transformed the management of brain tumors by enabling more accurate prognostication and personalized treatment. However, timely molecular diagnostic testing for patients with brain tumors is limited, complicating surgical and adjuvant treatment and obstructing clinical trial enrollment. In this study, we developed DeepGlioma, a rapid (<90 seconds), artificial-intelligence-based diagnostic screening system to streamline the molecular diagnosis of diffuse gliomas. DeepGlioma is trained using a multimodal dataset that includes stimulated Raman histology (SRH); a rapid, label-free, non-consumptive, optical imaging method; and large-scale, public genomic data. In a prospective, multicenter, international testing cohort of patients with diffuse glioma (n = 153) who underwent real-time SRH imaging, we demonstrate that DeepGlioma can predict the molecular alterations used by the World Health Organization to define the adult-type diffuse glioma taxonomy (IDH mutation, 1p19q co-deletion and ATRX mutation), achieving a mean molecular classification accuracy of 93.3 ± 1.6%. Our results represent how artificial intelligence and optical histology can be used to provide a rapid and scalable adjunct to wet lab methods for the molecular screening of patients with diffuse glioma.

Generation of 3D ex vivo mouse- and patient-derived glioma explant slice model for integration of confocal time-lapse imaging and spatial analysis.

Development of spatial-integrative pre-clinical models is needed for glioblastoma, which are heterogenous tumors with poor prognosis. Here, we present an optimized protocol to generate three-dimensional ex vivo explant slice glioma model from orthotopic tumors, genetically engineered mouse models, and fresh patient-derived specimens. We describe a step-by-step workflow for tissue acquisition, dissection, and sectioning of 300-µm tumor slices maintaining cell viability. The explant slice model allows the integration of confocal time-lapse imaging with spatial analysis for studying migration, invasion, and tumor microenvironment, making it a valuable platform for testing effective treatment modalities. For complete details on the use and execution of this protocol, please refer to Comba et al. (2022).1.

Subclonal evolution and expansion of spatially distinct THY1-positive cells is associated with recurrence in glioblastoma.

PURPOSE: Glioblastoma(GBM) is a lethal disease characterized by inevitable recurrence. Here we investigate the molecular pathways mediating resistance, with the goal of identifying novel therapeutic opportunities. EXPERIMENTAL DESIGN: We developed a longitudinal in vivo recurrence model utilizing patient-derived explants to produce paired specimens(pre- and post-recurrence) following temozolomide(TMZ) and radiation(IR). These specimens were evaluated for treatment response and to identify gene expression pathways driving treatment resistance. Findings were clinically validated using spatial transcriptomics of human GBMs. RESULTS: These studies reveal in replicate cohorts, a gene expression profile characterized by upregulation of mesenchymal and stem-like genes at recurrence. Analyses of clinical databases revealed significant association of this transcriptional profile with worse overall survival and upregulation at recurrence. Notably, gene expression analyses identified upregulation of TGFß signaling, and more than one-hundred-fold increase in THY1 levels at recurrence. Furthermore, THY1-positive cells represented <10% of cells in treatment-naïve tumors, compared to 75-96% in recurrent tumors. We then isolated THY1-positive cells from treatment-naïve patient samples and determined that they were inherently resistant to chemoradiation in orthotopic models. Additionally, using image-guided biopsies from treatment-naïve human GBM, we conducted spatial transcriptomic analyses. This revealed rare THY1+ regions characterized by mesenchymal/stem-like gene expression, analogous to our recurrent mouse model, which co-localized with macrophages within the perivascular niche. We then inhibited TGFBRI activity in vivo which decreased mesenchymal/stem-like protein levels, including THY1, and restored sensitivity to TMZ/IR in recurrent tumors. CONCLUSIONS: These findings reveal that GBM recurrence may result from tumor repopulation by pre-existing, therapy-resistant, THY1-positive, mesenchymal cells within the perivascular niche.

Methods and Impact for Using Federated Learning to Collaborate on Clinical Research.

BACKGROUND: The development of accurate machine learning algorithms requires sufficient quantities of diverse data. This poses a challenge in health care because of the sensitive and siloed nature of biomedical information. Decentralized algorithms through federated learning (FL) avoid data aggregation by instead distributing algorithms to the data before centrally updating one global model. OBJECTIVE: To establish a multicenter collaboration and assess the feasibility of using FL to train machine learning models for intracranial hemorrhage (ICH) detection without sharing data between sites. METHODS: Five neurosurgery departments across the United States collaborated to establish a federated network and train a convolutional neural network to detect ICH on computed tomography scans. The global FL model was benchmarked against a standard, centrally trained model using a held-out data set and was compared against locally trained models using site data. RESULTS: A federated network of practicing neurosurgeon scientists was successfully initiated to train a model for predicting ICH. The FL model achieved an area under the ROC curve of 0.9487 (95% CI 0.9471-0.9503) when predicting all subtypes of ICH compared with a benchmark (non-FL) area under the ROC curve of 0.9753 (95% CI 0.9742-0.9764), although performance varied by subtype. The FL model consistently achieved top three performance when validated on any site's data, suggesting improved generalizability. A qualitative survey described the experience of participants in the federated network. CONCLUSION: This study demonstrates the feasibility of implementing a federated network for multi-institutional collaboration among clinicians and using FL to conduct machine learning research, thereby opening a new paradigm for neurosurgical collaboration.

Perigeniculate Giant Cell Tumor of Temporal Bone.

Giant cell tumors of bone (GCTBs) are benign osteolytic neoplasms that can be treated with either gross-total resection or subtotal resection with adjuvant radiotherapy. For the rare GCTB of the temporal bone, close proximity to critical structures can produce functional deficits and make gross-total resection difficult to achieve without significant morbidity. We present the case of a 28-year-old woman with progressive facial paresis, otalgia, neck pain, imbalance, and subjective hearing loss. She was found to have a facial nerve mass centered at the geniculate ganglion extending into the labyrinthine segment and vestibule. We achieved gross-total resection with preserved facial nerve function as the tumor did not originate from the facial nerve and could be dissected free from the nerve. Final pathology was consistent with GCTB.

Generating novel pituitary datasets from open-source imaging data and deep volumetric segmentation.

PURPOSE: The estimated incidence of pituitary adenomas in the general population is 10-30%, yet radiographic diagnosis remains a challenge. Diagnosis is complicated by the heterogeneity of radiographic features in both normal (e.g. complex anatomy, pregnancy) and pathologic states (e.g. primary endocrinopathy, hypophysitis). Clinical symptoms and laboratory testing are often equivocal, which can result in misdiagnosis or unnecessary specialist referrals. Computer vision models can aid in pituitary adenoma diagnosis; however, a major challenge to model development is the lack of dedicated pituitary imaging datasets. We hypothesized that deep volumetric segmentation models trained to extract the sellar and parasellar region from existing whole-brain MRI scans could be used to generate a novel dataset of pituitary imaging. METHODS: Six open-source whole-brain MRI datasets, created for research purposes, were included for model development. Deep learning-based volumetric segmentation models were trained using 318 manually annotated MRI scans from a single open-source MRI dataset. Out-of-distribution volumetric segmentation performance was then tested on 418 MRIs from five held-out research datasets. RESULTS: On our annotated images, agreement between manual and model volumetric segmentations was high. Dice scores (a measure of overlap) ranged 0.76-0.82 for both in-distribution and out-of-distribution model testing. In total, 6,755 MRIs from six data sources were included in the final generated pituitary dataset. CONCLUSIONS: We present the first and largest dataset of pituitary imaging constructed using existing MRI data and deep volumetric segmentation models trained to identify sellar and parasellar anatomy. The model generalizes well across patient populations and MRI scanner types. We hope our pituitary dataset will be an integral part of future machine learning research on pituitary pathologies.

Spatiotemporal analysis of glioma heterogeneity reveals COL1A1 as an actionable target to disrupt tumor progression.

Intra-tumoral heterogeneity is a hallmark of glioblastoma that challenges treatment efficacy. However, the mechanisms that set up tumor heterogeneity and tumor cell migration remain poorly understood. Herein, we present a comprehensive spatiotemporal study that aligns distinctive intra-tumoral histopathological structures, oncostreams, with dynamic properties and a specific, actionable, spatial transcriptomic signature. Oncostreams are dynamic multicellular fascicles of spindle-like and aligned cells with mesenchymal properties, detected using ex vivo explants and in vivo intravital imaging. Their density correlates with tumor aggressiveness in genetically engineered mouse glioma models, and high grade human gliomas. Oncostreams facilitate the intra-tumoral distribution of tumoral and non-tumoral cells, and potentially the collective invasion of the normal brain. These fascicles are defined by a specific molecular signature that regulates their organization and function. Oncostreams structure and function depend on overexpression of COL1A1. Col1a1 is a central gene in the dynamic organization of glioma mesenchymal transformation, and a powerful regulator of glioma malignant behavior. Inhibition of Col1a1 eliminates oncostreams, reprograms the malignant histopathological phenotype, reduces expression of the mesenchymal associated genes, induces changes in the tumor microenvironment and prolongs animal survival. Oncostreams represent a pathological marker of potential value for diagnosis, prognosis, and treatment.

Extent of Resection Research in Skull Base Neurosurgery: Previous Studies and Future Directions.

Surgery is the first-line therapy for most benign and malignant skull base tumors. Extent of resection (EOR) is a metric commonly used for preoperative surgical planning and to predict risk of postoperative tumor recurrence. Therefore, understanding the evidence on EOR in skull base neurosurgery is essential to providing optimal care for each patient. Several studies from the skull base neurosurgery literature have presented investigations of various topics related to EOR, including 1) preoperative EOR scoring systems, 2) intraoperative EOR scoring systems, 3) EOR and tumor recurrence, and 4) EOR and functional outcomes. We propose that future investigations should focus on the following elements to improve EOR research in skull base neurosurgery: 1) multi-institutional collaboratives with treatment propensity matching; 2) expert consensus and mixed-methods study design; and 3) predictive analytics/machine learning. We believe that these methods offer several advantages that have been described in the literature and that they address limitations of previous studies. The aim of this review was to inform future study design and improve the overall quality of subsequent investigations on EOR in skull base neurosurgery.

Laminectomy at T4 and T5 for Resection of Symptomatic Cavernous Malformation.

Although rare, intramedullary spinal cavernous malformations have a 1.4%-6.8% annual hemorrhage risk and can cause significant morbidity.1 Prior hemorrhage and size >1 cm are risk factors for future hemorrhage that, in addition to notable or progressive symptoms, may justify early surgical intervention.1,2 In this video, we present key steps in surgical management of a large, symptomatic thoracic cavernous malformation. A 56-year-old woman presented with worsening lower extremity weakness, imbalance, and difficulty ambulating. Strength was 3/5 in her right lower extremity and 4/5 in her left lower extremity. She had an incomplete T4 sensory level and hyperreflexia. Magnetic resonance imaging demonstrated a heterogeneous "popcorn"-appearing expansile intradural intramedullary 2.2- × 1.2-cm lesion at T4-5, consistent with a cavernous malformation. Angiography was deferred given the characteristic magnetic resonance imaging appearance. Given her progressive symptoms (including weakness), lesion size, and good health, resection was recommended. Using neurological monitoring, a T4-5 laminectomy, midline myelotomy, and piecemeal microsurgical resection of the lesion was performed, clearly identifying the cavernoma-spinal cord interface and avoiding spinal cord retraction. Histopathology confirmed a cavernoma. Postoperatively, the patient had improved left lower extremity strength and stable right lower extremity strength but worsened dorsiflexion (1/5), which improved with rehabilitation. At 1-year follow-up, she had full strength in her left lower extremity and 4/5 in her right lower extremity, with mild paresthesias below T10. Consistent with prior series demonstrating low complication rates and good long-term neurological outcomes,2 microsurgical resection of selected symptomatic intramedullary spinal cavernous malformations can halt neurological decline and potentially improve neurological function.

Rapid Automated Analysis of Skull Base Tumor Specimens Using Intraoperative Optical Imaging and Artificial Intelligence.

BACKGROUND: Accurate specimen analysis of skull base tumors is essential for providing personalized surgical treatment strategies. Intraoperative specimen interpretation can be challenging because of the wide range of skull base pathologies and lack of intraoperative pathology resources. OBJECTIVE: To develop an independent and parallel intraoperative workflow that can provide rapid and accurate skull base tumor specimen analysis using label-free optical imaging and artificial intelligence. METHODS: We used a fiber laser-based, label-free, nonconsumptive, high-resolution microscopy method (<60 seconds per 1 × 1 mm2), called stimulated Raman histology (SRH), to image a consecutive, multicenter cohort of patients with skull base tumor. SRH images were then used to train a convolutional neural network model using 3 representation learning strategies: cross-entropy, self-supervised contrastive learning, and supervised contrastive learning. Our trained convolutional neural network models were tested on a held-out, multicenter SRH data set. RESULTS: SRH was able to image the diagnostic features of both benign and malignant skull base tumors. Of the 3 representation learning strategies, supervised contrastive learning most effectively learned the distinctive and diagnostic SRH image features for each of the skull base tumor types. In our multicenter testing set, cross-entropy achieved an overall diagnostic accuracy of 91.5%, self-supervised contrastive learning 83.9%, and supervised contrastive learning 96.6%. Our trained model was able to segment tumor-normal margins and detect regions of microscopic tumor infiltration in meningioma SRH images. CONCLUSION: SRH with trained artificial intelligence models can provide rapid and accurate intraoperative analysis of skull base tumor specimens to inform surgical decision-making.