Real-time intraoperative confocal laser endomicroscopy with on-site neuropathologist consultation for glioma margin assessment: illustrative case

Abstract

BACKGROUND

Achieving maximal safe resection in glioma surgery requires accurate real-time margin assessment, yet existing technologies have limitations. Fluorescence-guided surgery shows limitations at infiltrative margins, while frozen section analysis requires 20–45 minutes per specimen. Confocal laser endomicroscopy (CLE) is an emerging in vivo imaging technology that enables real-time cellular-resolution tissue examination in situ.

OBSERVATIONS

The authors present a case demonstrating integration of CLE with on-site neuropathologist collaboration for real-time intraoperative diagnosis. A 53-year-old woman with left frontotemporal high-grade glioma underwent surgery with systematic examination of five locations using CLE. The collaborative workflow enabled rapid tissue assessment (mean 95 seconds per site) with immediate expert interpretation. Critically, high-grade glioma was detected at a site with complete absence of 5-aminolevulinic acid fluorescence (spot 4), demonstrating CLE’s capability to identify tumor infiltration beyond fluorescence-guided boundaries and enabling extended resection.

LESSONS

Integrating CLE with on-site pathologist consultation creates an effective paradigm for real-time margin assessment, addressing the 30%–40% false-negative rate of fluorescence guidance. This collaborative approach is immediately implementable at centers with neuropathology expertise and establishes a foundation for future telepathology and artificial intelligence–assisted diagnostic platforms.

https://thejns.org/doi/10.3171/CASE25890

The primary goal in the surgical management of glioma is maximal safe resection, yet the diffuse infiltrative nature of these tumors into adjacent brain parenchyma makes intraoperative distinction between tumor margins and normal tissue a significant challenge.^1,2 While technologies such as fluorescence-guided surgery or intraoperative MRI have improved the extent of resection, there remains an unmet need for a definitive method to assess tissue at a cellular level in real time.^2,3 Confocal laser endomicroscopy (CLE) is an emerging technology that provides surgeons with real-time microscopic visualization of tissue in vivo, potentially addressing this gap. However, accurate interpretation of CLE images requires expertise in correlating microscopic cellular patterns with histopathological features—a challenge that may limit adoption in routine surgical practice.⁴

This report presents an illustrative case demonstrating a practical collaborative workflow that integrates CLE with on-site neuropathologist consultation for real-time margin assessment during glioma resection. We used the cCeLL-In vivo system (VPIX Medical Inc.), a handheld CLE device with US Food and Drug Administration 510(k) clearance for microscopic tissue visualization.^5,6 The case highlights how this collaborative approach addresses both the technical capabilities and the interpretive challenges of intraoperative CLE implementation.

Illustrative Case

Patient and Surgical Procedure

A 53-year-old woman with a history of well-controlled epilepsy presented with seizure recurrence and progressive memory deficits. MRI revealed a heterogeneously enhancing mass in the left medial temporal lobe and hippocampus, with surrounding T2 fluid-attenuated inversion recovery (FLAIR) hyperintensity extending to the anterior temporal lobe, orbitofrontal region, and insula, consistent with high-grade glioma (Fig. 1). The patient underwent left frontotemporal craniotomy with tumor resection under general anesthesia. 5-Aminolevulinic acid (5-ALA) was administered orally at a standard dose (20 mg/kg) approximately 3 hours before surgery for fluorescence guidance.⁷ The procedure was performed as part of an institutional review board–approved study evaluating CLE for intraoperative margin assessment. The patient acknowledged that the technology enabled more complete tumor resection without requiring additional surgical procedures.

FIG. 1.

Preoperative MR images. A: Axial T1-weighted images with gadolinium enhancement demonstrating a heterogeneously enhancing mass in the left frontotemporal region, centered in the medial temporal lobe including the hippocampus and lateral temporal lobe. The tumor shows irregular enhancement patterns with central necrosis and perilesional edema. B: Corresponding axial FLAIR images showing extensive T2 FLAIR hyperintensity involving the left frontal and temporal lobes, with infiltrative signal changes extending to the orbitofrontal region and surrounding white matter.

System Configuration and Intraoperative Workflow

The cCeLL-In vivo system utilizes a 775-nm laser with 12-μm imaging depth and provides cellular-resolution images at 1024 × 1024 pixel resolution. To optimize surgical ergonomics, the CLE video output was integrated into the surgical microscope’s (KINEVO 900, Carl Zeiss Meditec AG) multivision display, creating a picture-in-picture view that allowed simultaneous visualization of the macroscopic surgical field and microscopic CLE images on a single monitor.⁸

A collaborative protocol was established between the operating neurosurgeon (C.K.P., 29 years of experience) and an on-site neuropathologist (J.K.W., 24 years of experience). At each examination site, the workflow consisted of four sequential steps: 1) indocyanine green (ICG) was applied topically for fluorescent contrast (staining time: 22 ± 5.8 seconds); 2) the surgeon positioned the sterile-sheathed CLE probe on target tissue; 3) the neuropathologist provided immediate interpretation based on cellular morphology visible in the live CLE image; and 4) a tissue biopsy was obtained from the identical location for histopathological validation. This systematic approach enabled direct correlation between CLE findings and permanent section diagnosis at each site. The integrated workflow is illustrated in Fig. 2. Video 1 demonstrates the complete surgical workflow and system integration.

FIG. 2.

Intraoperative workflow integrating CLE with on-site neuropathologist consultation. A: Timeline of intraoperative examination showing five sequential spots with corresponding consultation times: spot 1 (normal cortex, 180 seconds), spot 2 (nontumor white matter, 86 seconds), spot 3 (ambiguous/reactive gliosis, 81 seconds), spot 4 (5-ALA–negative high-grade glioma, 87 seconds), and spot 5 (5-ALA–positive high-grade glioma, 42 seconds). Color gradient indicates diagnostic certainty from normal (light red) to malignant (dark red). B: Schematic representation of the real-time collaborative workflow. The surgical layer encompasses tissue-specific resection guided by CLE imaging (775-nm laser, 12-μm depth, 1024 × 1024 pixels) and microscopy with 5-ALA fluorescence mode and white light imaging. The diagnostic layer involves on-site neuropathologist consultation and pathological confirmation through real-time interpretation of CLE images. Bidirectional arrows indicate continuous communication between surgical and diagnostic teams, enabling immediate decision-making for maximal safe resection. Modified from our previously published illustration (Nam SM, et al. J Korean Neurosurg Soc. 2025;68(2):137-149).²⁹

VIDEO 1.Complete surgical workflow of real-time intraoperative CLE. The video demonstrates the integration of the CLE system with the surgical microscope, the step-by-step process of image acquisition at five distinct spots (including ICG application and probe positioning), and the simultaneous on-site consultation with the neuropathologist. Key findings, including the detection of high-grade glioma in a 5-ALA–negative region (spot 4), are highlighted. bx = biopsy; HGG = high-grade glioma; LGG = low-grade glioma; ts = tissue; V = vein. Click here to view.

Intraoperative Findings

Five anatomical locations were systematically examined using CLE, progressing from peripheral normal-appearing tissue toward the tumor core over a period of approximately 78 minutes. The mean imaging time per site was 95 ± 51 seconds. The examination timeline and spatial sequence are illustrated in Fig. 2A, and CLE findings with corresponding hematoxylin and eosin (H&E) histopathology for each spot are shown in Fig. 3.

FIG. 3.

Intraoperative CLE and corresponding H&E findings from five spots. A: Spot 1 (normal cortex, inferior frontal lobe). CLE demonstrates sparse cellularity with normal neuronal architecture and minimal vascular structures (red arrowheads indicate neurons). Corresponding H&E staining confirms normal cortical tissue with preserved neuronal layers and no evidence of tumor infiltration. B: Spot 2 (nontumor white matter, orbitofrontal region). CLE shows linear arrangements of myelinated fibers with low cellularity (red arrowheads indicate fiber tracts). H&E staining reveals normal white matter architecture with parallel myelinated axons and scattered oligodendroglial cells without tumor cells. This area corresponded to the orbitofrontal region showing FLAIR hyperintensity on preoperative MR images (Fig. 1B), confirming vasogenic edema rather than tumor infiltration. C: Spot 3 (ambiguous findings, superior temporal lobe). CLE displays moderately increased cellularity with irregular cellular architecture (red arrowheads), raising suspicion for tumor infiltration. H&E staining demonstrates reactive gliosis with astrocytic proliferation and increased cellularity, lacking definitive features of high-grade glioma. D: Spot 4 (5-ALA–negative high-grade glioma, hippocampal head). CLE reveals significantly increased cellularity with dense, disorganized cellular patterns and prominent nuclear pleomorphism (red arrowheads), suspicious for high-grade tumor despite absence of 5-ALA fluorescence. H&E staining confirms high-grade glioma with hypercellularity, nuclear atypia, and mitotic figures. This finding led to extended resection beyond the 5-ALA–positive margin. E: Spot 5 (5-ALA–positive high-grade glioma). CLE shows marked cellular density with chaotic architecture and prominent vascular proliferation (red arrowheads), consistent with high-grade glioma. H&E staining demonstrates densely cellular high-grade glioma with significant nuclear pleomorphism, increased mitotic activity, and microvascular proliferation. Scale bars are embedded in the original images (CLE: 50 μm; H&E: 50 μm for spots 1–3, 100 μm for spots 4 and 5).

Spot 1 (normal cortex)

Following frontotemporal craniotomy, the initial examination targeted frontal cortex remote from the tumor. CLE imaging demonstrated low cellular density with regular architecture and intact fine capillary networks, consistent with normal brain parenchyma. The pathologist interpreted this as nontumorous tissue, subsequently validated by permanent section showing normal cortical architecture without tumor infiltration (Fig. 3A).

Spot 2 (peripheral margin, nontumor)

The second site examined the orbitofrontal region corresponding to T2 FLAIR hyperintensity. CLE revealed vertically oriented fibrous structures characteristic of white matter tracts. The pathologist’s interpretation identified these features as nontumorous white matter with mild reactive changes. Permanent histopathology confirmed white matter with reactive gliosis and no tumor infiltration (Fig. 3B).

Spot 3 (peripheral margin, ambiguous)

An edematous region at the temporal margin without 5-ALA fluorescence was examined next. Based on location and imaging characteristics, differential diagnosis included infiltrative low-grade tumor or reactive gliosis. CLE imaging revealed moderately increased cellular density with irregular distribution and entangled cellular processes. The pathologist interpreted these features as ambiguous—low-grade glial tumor versus reactive gliosis—with definitive diagnosis deferred pending permanent section. Subsequent H&E analysis confirmed reactive gliosis without definitive tumor (Fig. 3C).

Spot 4 (tumor margin, 5-ALA negative)

A region adjacent to the tumor core demonstrated complete absence of 5-ALA fluorescence despite the surgeon’s clinical suspicion for residual tumor based on tissue consistency and anatomical location. CLE examination revealed markedly increased cellular density, nuclear pleomorphism, and disorganized architecture. The pathologist’s interpretation was high-grade glioma. Permanent section confirmed high-grade glial tumor at this 5-ALA–negative location (Fig. 3D), demonstrating CLE’s capability to detect tumor infiltration beyond fluorescence-guided boundaries. This finding proved critical for extending resection margins.

Spot 5 (tumor core, 5-ALA positive)

The final examination targeted tissue with intense 5-ALA fluorescence within the tumor core. CLE features definitively demonstrated high-grade glioma characteristics, including markedly enlarged pleomorphic cells, disorganized architecture, and prominent irregular vascular structures. Permanent histopathology confirmed high-grade glial tumor (Fig. 3E).

Postoperative MRI demonstrated gross-total resection of the enhancing tumor. Immediately postoperatively, the patient developed motor aphasia with impaired speech fluency, corresponding to a Karnofsky Performance Status (KPS) score of 70. Following the diagnosis of WHO grade 4 glioma, the patient underwent standard adjuvant chemoradiation therapy according to international protocols. At the 3-month follow-up, while speech function had improved, mild deterioration in right hand fine motor function was noted, maintaining the KPS score at 70. The patient remains under close surveillance. The patient expressed satisfaction with the surgical outcome and appreciated how the real-time tissue assessment technology enabled more complete tumor removal without requiring additional procedures.

Discussion

Observations

Clinical Significance of Real-Time Collaborative Workflow

The primary observation from this case is the establishment of a highly efficient and accurate workflow supporting real-time resection margin decision-making. Traditional intraoperative assessment relies on the surgeon’s visual and tactile judgment, supplemented by fluorescence guidance and frozen section analysis when available. However, each modality has recognized limitations. Fluorescence-guided surgery, while valuable for identifying bulk tumor, demonstrates visible fluorescence in only 49%–80% of infiltrative glioblastoma margins.^9,10 Notably, a recent study demonstrated that CLE detected tumor in 5-ALA–negative marginal areas with 73% sensitivity, compared to only 38% for fluorescence guidance alone, highlighting the complementary value of CLE at infiltrative margins.¹¹ Frozen section analysis, accepted as the gold standard for tissue diagnosis, requires 20–45 minutes per specimen and necessitates tissue excision for processing.^6,12

The integration of CLE with real-time neuropathologist consultation addresses both limitations simultaneously. CLE provides rapid assessment (1–2 minutes per site) through in situ tissue examination without requiring excision, while on-site pathologist consultation ensures accurate interpretation with diagnostic accuracy comparable to permanent histopathology.⁴ In our experience, the complete workflow—from probe positioning to pathologist interpretation—averaged 95 seconds per site, with image interpretation itself requiring less than 1 minute. Although initial system setup required additional time for workflow optimization, this is expected to decrease with institutional experience. This workflow augments surgical judgment with an additional layer of histopathological information, which is particularly valuable in ambiguous margins where clinical decision-making is most challenging.

The detection of high-grade glioma at spot 4—a region with complete absence of 5-ALA fluorescence—illustrates the complementary value of this approach. The surgeon’s clinical suspicion based on tissue consistency and anatomical context prompted immediate CLE examination, enabling real-time confirmation and extended resection that might otherwise have been deferred or missed entirely.

Comparison With Emerging Technologies

Since the initial feasibility study by Sanai et al. in 2011,¹³ CLE has evolved from a proof of concept to a clinically validated modality, with multiple centers establishing real-time pathological consultation workflows^14–17 (Table 1). Published series report diagnostic accuracy ranging from 85% to 94%, with total added surgical time of approximately 10–15 minutes per case.^18–21 Our experience aligns with these benchmarks, demonstrating that on-site collaboration achieves comparable efficiency to established telepathology protocols.

TABLE 1.

Summary of in vivo CLE studies for brain tumor surgery

Authors & Year	System/Agent*	No. of Patients†	Pathological Interpretation	Workflow Integration	Key Findings
Sanai et al., 201113	Optiscan/FNa	33	None (retrospective)	Demonstrated safety of integration; w/o adverse events	1st feasibility; proof of concept for in vivo cellular visualization
Martirosyan et al., 201619	Optiscan/FNa	72	None (retrospective)	Mean duration 15.7 mins; 1st diagnostic image w/in ~17 sec; yield: 46% usable images	High diagnostic accuracy for gliomas (sensitivity 91%, specificity 94%)
Höhne et al., 202114	CONVIVO/FNa	12	None (surgeon only)	Seamless application w/o altering surgical strategy	Achieved 100% tumor visualization w/ intuitive handling
Xu et al., 202215	CONVIVO/FNa	30	None (quantitative)	Established SNR metrics for image quality control	In vivo images showed superior brightness/contrast to ex vivo
Abramov et al., 202318	CONVIVO/FNa	30	Telepathology (real-time)	Validated the workflow for remote expert review	1st telepathology study; 94% diagnostic concordance w/ FS
Wagner et al., 202421	CONVIVO/FNa	203	Telepathology (real-time)	Speed: 10× faster (3 mins) than FS (27 mins)	Accuracy 87% (failed noninferiority vs FS 91%)
Abramov et al., 202511	CONVIVO/FNa	33	Telepathology (real-time)	Complementary use in 5-ALA–negative marginal areas	CLE (sensitivity 73%, specificity 41%) vs 5-ALA (sensitivity 38%, specificity 82%)
Reichenbach et al., 202416	CONVIVO/label free	3	None (post hoc)	Image quality analysis in context of routine surgical procedure	Label-free technical limit: only 22% usable images due to motion artifacts (42%)
Restelli et al., 202520	CONVIVO/FNa	75	Telepathology (real-time)	Time: total added ~10.5 mins; interpretation (~41 sec) < acquisition per site (~2.1 mins)	Accuracy 85.8% for routine screening at resection margins
Xu et al., 202526	CONVIVO/FNa	50 (27 telepathology)	Hybrid (real-time & retrospective)	Telepathology added ~3.8-min delay per site	Specificity drops significantly at margins (93% core → 50% margin)
Sistiaga et al., 202617	CONVIVO/FNa	7	None (surgeon only)	Integration into tubular retractor & endoscopic settings	Feasibility study; confirmed probe maneuverability in restricted corridors
Present case	cCeLL-In vivo/ICG	1	On-site (real-time)	Seamless on-site integration during surgery	Confirmed immediate feedback utility for real-time decision-making

FNa = fluorescein sodium; FS = frozen section; IV = intravenous; SNR = signal-to-noise ratio.

* FNa administration (intravenous, 2–5 mg/kg) protocols varied: at induction,^14,17 at dural opening,¹⁵ or just prior to imaging.¹³ Redosing permitted in prolonged cases.^11,15,18,26

† Represents the number of patients who underwent in vivo imaging. Martirosyan et al. (2016)¹⁹ reported 74 total cases (2 ex vivo excluded). Xu et al. (2025)²⁶ included 27 cases evaluated via telepathology.

This capability for rapid intraoperative diagnosis parallels broader developments in real-time tissue characterization. Stimulated Raman histology (SRH), exemplified by systems such as FastGlioma, has shown similar diagnostic accuracy with processing times of approximately 10–100 seconds per specimen.²² Both CLE and SRH share the fundamental advantage of eliminating tissue processing delays inherent to frozen section analysis.

However, each approach presents distinct characteristics that influence implementation. SRH enables label-free imaging and generates images more closely resembling traditional H&E histology, which may facilitate pathologist interpretation.²³ In contrast, CLE requires fluorescent contrast agents (such as fluorescein sodium or, as used in this case, ICG) but offers the unique advantage of in situ tissue examination without requiring tissue excision.^18,24 In scenarios where clinical suspicion exists but fluorescence guidance is absent, CLE allows immediate assessment without committing to tissue removal, enabling the surgeon to survey multiple areas before deciding where resection should be extended.

Educational Benefits Through Bidirectional Learning

An important benefit of this workflow is the acceleration of expertise development for both pathologists and surgeons through immediate bidirectional feedback. For pathologists, observing the surgical field in real time—including tissue handling, macroscopic appearance, anatomical location, fluorescence patterns, and tissue consistency—provides critical contextual information that enhances microscopic interpretation. This immediate correlation between macroscopic surgical findings and microscopic cellular architecture refines diagnostic accuracy over successive cases.

For surgeons, this collaboration enables the development of an integrated decision-making framework. In addition to visual inspection, tactile feedback, and anatomical knowledge, surgeons gain immediate histopathological correlation for each tissue they encounter. Over time, this allows surgeons to validate their intraoperative impressions against histopathological confirmation immediately rather than waiting days for permanent section. The result is not just better decision-making in the current case, but progressive refinement of surgical expertise that carries forward to subsequent procedures, potentially enabling more confident and accurate margin assessment.

Limitations

Several important limitations warrant acknowledgment. First, this report represents a single patient experience at a single institution, and generalizability across diverse tumor types, grades, and anatomical locations remains to be established. Second, the requirement for on-site neuropathologist expertise represents a significant barrier to widespread adoption. This case demonstrates the value of immediate expert consultation; however, such expertise is not universally available. Third, interobserver variability among pathologists with different levels of CLE experience remains a practical consideration that affects real-world implementation. Finally, eventual beneficial effects on survival or disease progression remain to be determined in larger prospective studies.

Future Directions: Telepathology, Artificial Intelligence Integration, and Operating Room Evolution

The on-site collaborative model demonstrated here establishes a practical framework that can be adapted for important future developments. First, telepathology systems enabling remote expert interpretation could extend the benefits of this workflow to centers lacking immediate access to specialized neuropathology expertise. Several groups have reported successful telepathology implementations for intraoperative use,^{4,17,18,25,26} and the real-time communication protocols validated through on-site collaboration provide a template for remote consultation scenarios.

Second, the continued advancement of artificial intelligence (AI) in medical imaging may further enhance the accessibility and efficiency of CLE-based workflows.^27,28 Machine learning algorithms trained on datasets where CLE images are correlated with expert interpretation could eventually provide standardized assessment of tissue characteristics. Such AI-assisted systems could serve as decision support tools, particularly at centers lacking immediate access to expert neuropathologists, providing preliminary assessment that guides sampling decisions and highlights areas requiring expert review.

Beyond these specific technical capabilities, this case illustrates a broader evolution in the informational architecture of the operating room. Traditional neurosurgical procedures have operated primarily within a single operational layer where the surgical team makes decisions based on visual inspection, tactile feedback, and anatomical knowledge. The integration of real-time CLE with immediate pathological consultation creates an additional diagnostic layer that operates simultaneously with surgical decision-making. This workflow integrates diagnostic and surgical layers, establishing the operating room as a multilayered information environment.²⁹ As telepathology and AI-assisted interpretation advance, such integration will become more feasible and widespread.

Lessons

This case demonstrates that integrating CLE with on-site neuropathologist collaboration creates an effective workflow for real-time margin assessment during glioma resection. The collaborative approach enabled rapid tissue characterization (mean 95 seconds per site) with diagnostic accuracy comparable to permanent histopathology. The most significant finding was detecting high-grade glioma in a region completely lacking 5-ALA fluorescence, enabling extended resection beyond fluorescence-guided boundaries. This capability to identify tumor at fluorescence-negative margins represents a critical advancement for achieving maximal safe resection.

Beyond immediate diagnostic value, this integration of diagnostic and surgical layers expands the informational landscape of the operating room, establishing a foundation for future telepathology systems and AI-assisted platforms that may further enhance real-time intraoperative decision-making.

Acknowledgments

Funding is acknowledged from the Korea Health Industry Development Institute (KHIDI) (RS-2025-23872969 to C.K.P.). Claude (Anthropic), a large language model, was used to improve grammar, refine sentence structure, correct idiomatic expressions (phrasing), and enhance the overall readability and clarity of the writing.

Disclosures

Dr. Kim reported personal fees from VPIX Medical Inc. outside the submitted work.

Author Contributions

Conception and design: Park, Nam. Acquisition of data: all authors. Analysis and interpretation of data: Park, Nam, Won. Drafting the article: Park, Nam, Kim. Critically revising the article: Park, Nam. Reviewed submitted version of manuscript: Park. Approved the final version of the manuscript on behalf of all authors: Park. Statistical analysis: Nam. Administrative/technical/material support: all authors. Study supervision: Park.

Supplemental Information

Videos

  Video 1. https://vimeo.com/1156529919.

Correspondence

Chul-Kee Park: Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea. nsckpark@snu.ac.kr.